1
00:00:05,239 --> 00:00:07,259
[music]

2
00:00:09,825 --> 00:00:11,845
[music]

3
00:00:16,000 --> 00:00:18,080
We love more bin chickens. That's what

4
00:00:18,080 --> 00:00:21,920
we're here for. Uh we welcome. We are

5
00:00:21,920 --> 00:00:23,920
here for a few things. It's the

6
00:00:23,920 --> 00:00:28,165
education showcase. Woo!

7
00:00:28,165 --> 00:00:28,720
[applause]

8
00:00:28,720 --> 00:00:31,599
We are joined by some fabulous students

9
00:00:31,599 --> 00:00:33,840
to hear about all about the amazing

10
00:00:33,840 --> 00:00:36,025
things that they've been doing.

11
00:00:36,025 --> 00:00:36,559
[laughter]

12
00:00:36,559 --> 00:00:39,360
We will probably not set fire to

13
00:00:39,360 --> 00:00:43,440
anything today. I know this is intensely

14
00:00:43,440 --> 00:00:44,960
disappointing for some, but we will

15
00:00:44,960 --> 00:00:46,960
maybe light a flame of hope in your

16
00:00:46,960 --> 00:00:50,640
hearts for the youth. They are okay. Um

17
00:00:50,640 --> 00:00:53,199
we are we are joined in person by a

18
00:00:53,199 --> 00:00:54,480
bunch of students but there is one

19
00:00:54,480 --> 00:00:56,079
student who couldn't join us in person

20
00:00:56,079 --> 00:00:59,920
today. Uh apparently Perth is far away.

21
00:00:59,920 --> 00:01:05,360
I don't know. Um uh Pyon Extreme West

22
00:01:05,360 --> 00:01:09,600
soon something. Um okay. So first up we

23
00:01:09,600 --> 00:01:15,920
have um a video from Alla and uh she is

24
00:01:15,920 --> 00:01:20,000
in year 11. Uh she's going to talk about

25
00:01:20,000 --> 00:01:22,640
Stallic Tits. Uh and I've realized I

26
00:01:22,640 --> 00:01:24,479
don't have her favorite emoji up on my

27
00:01:24,479 --> 00:01:26,320
spreadsheet, which is an oversight on my

28
00:01:26,320 --> 00:01:29,200
part. So at the end of this, you get to

29
00:01:29,200 --> 00:01:30,560
find out how long she's been programming

30
00:01:30,560 --> 00:01:32,799
and what her favorite emoji is. But in

31
00:01:32,799 --> 00:01:36,479
the meantime, uh we're going to hear

32
00:01:36,479 --> 00:01:37,840
directly from her.

33
00:01:37,840 --> 00:01:39,439
If we plug the audio in, we

34
00:01:39,439 --> 00:01:41,040
Oh,

35
00:01:41,040 --> 00:01:42,799
sorry. This is all my fault.

36
00:01:42,799 --> 00:01:46,119
That's all right.

37
00:01:46,799 --> 00:01:51,600
This one. No, thank you. I I my uh I

38
00:01:51,600 --> 00:01:54,320
would do a terrible voice over for it as

39
00:01:54,320 --> 00:01:57,880
well. Okay.

40
00:01:58,159 --> 00:02:00,880
Stalmmites and a drone, a Python arcade

41
00:02:00,880 --> 00:02:03,360
game.

42
00:02:03,360 --> 00:02:05,840
I was handed out an assignment from

43
00:02:05,840 --> 00:02:08,239
school to build a Python game during my

44
00:02:08,239 --> 00:02:11,280
twoe term break and that led me to build

45
00:02:11,280 --> 00:02:14,000
cable drone. It involves a yellow drone

46
00:02:14,000 --> 00:02:16,239
moving inside a cave avoiding obstacles

47
00:02:16,239 --> 00:02:18,400
of mineral formations called stallic

48
00:02:18,400 --> 00:02:21,280
mites and stallctites. Here is the game

49
00:02:21,280 --> 00:02:24,280
play.

50
00:02:49,599 --> 00:02:52,599
Oo.

51
00:03:11,760 --> 00:03:14,000
The objective of the game is for the

52
00:03:14,000 --> 00:03:15,680
drone to explode the cave without

53
00:03:15,680 --> 00:03:17,680
colliding with the stalkites and

54
00:03:17,680 --> 00:03:20,560
stallites. And if the drone does happen

55
00:03:20,560 --> 00:03:22,720
to collide, it crashes to the ground and

56
00:03:22,720 --> 00:03:24,800
a game over text is displayed to the

57
00:03:24,800 --> 00:03:26,319
player.

58
00:03:26,319 --> 00:03:29,040
And the up and down arrow keys help the

59
00:03:29,040 --> 00:03:30,640
player navigate between the mineral

60
00:03:30,640 --> 00:03:35,319
formations without crashing into them.

61
00:03:35,440 --> 00:03:38,400
A bit of backstory first. After getting

62
00:03:38,400 --> 00:03:40,560
the school project, I was stuck on two

63
00:03:40,560 --> 00:03:42,959
things during the ideation process. What

64
00:03:42,959 --> 00:03:44,959
type of game to create? Should it be

65
00:03:44,959 --> 00:03:47,440
text based or graphical? And what the

66
00:03:47,440 --> 00:03:50,000
game play should be. I wanted to create

67
00:03:50,000 --> 00:03:52,560
a 2D graphical game using the Python

68
00:03:52,560 --> 00:03:55,519
open-source library called Pygame. This

69
00:03:55,519 --> 00:03:59,120
was my first attempt to make a Python

70
00:03:59,120 --> 00:04:01,519
graphical game without with sound

71
00:04:01,519 --> 00:04:04,560
effects. I knew that this was not easy,

72
00:04:04,560 --> 00:04:07,519
so I got right to work.

73
00:04:07,519 --> 00:04:10,720
I have seen videos of how the popular

74
00:04:10,720 --> 00:04:13,200
Flappy Bird game is, and I know that

75
00:04:13,200 --> 00:04:15,200
beginner programmers usually create a

76
00:04:15,200 --> 00:04:17,600
Flappy Bird clone as their first Python

77
00:04:17,600 --> 00:04:20,560
game project. The simplicity of the game

78
00:04:20,560 --> 00:04:22,800
and the addictive nature of the gameplay

79
00:04:22,800 --> 00:04:25,840
was what drew me to it. However, I

80
00:04:25,840 --> 00:04:27,600
wanted to personalize the game in a

81
00:04:27,600 --> 00:04:31,120
unique setting. Thus, I picked a terrain

82
00:04:31,120 --> 00:04:34,000
in my game as a cave with mineral

83
00:04:34,000 --> 00:04:37,040
formations. And the drone is exploring

84
00:04:37,040 --> 00:04:41,960
the cave and trying to avoid obstacles.

85
00:04:42,639 --> 00:04:45,040
One of the first things I learned was

86
00:04:45,040 --> 00:04:47,280
what a stallicmite and stallctides

87
00:04:47,280 --> 00:04:50,400
actually are. Stallctides are mineral

88
00:04:50,400 --> 00:04:52,320
formations that hang from the ceilings

89
00:04:52,320 --> 00:04:55,040
of caves. And stallic mides are mineral

90
00:04:55,040 --> 00:04:57,120
formations that rise from the cave

91
00:04:57,120 --> 00:05:00,960
floor. The C in stallctite stands for

92
00:05:00,960 --> 00:05:04,800
celic. They are always found in pairs as

93
00:05:04,800 --> 00:05:07,120
stallic mites are the accumulation of

94
00:05:07,120 --> 00:05:11,199
materials that drip from stallctites. So

95
00:05:11,199 --> 00:05:13,520
the G in the stallicmite stands for

96
00:05:13,520 --> 00:05:16,639
ground. These formations are the

97
00:05:16,639 --> 00:05:18,960
obstacles in the game which the drone

98
00:05:18,960 --> 00:05:21,680
must avoid to prevent a collision from

99
00:05:21,680 --> 00:05:24,680
happening.

100
00:05:25,360 --> 00:05:26,880
Some of the things that I learned from

101
00:05:26,880 --> 00:05:30,639
the game is how to detect key presses

102
00:05:30,639 --> 00:05:33,680
and the drawn movement physics and

103
00:05:33,680 --> 00:05:36,320
blading images on transparent layers.

104
00:05:36,320 --> 00:05:38,880
Also had to learn how to draw the

105
00:05:38,880 --> 00:05:41,840
obstacles in and save the best score as

106
00:05:41,840 --> 00:05:46,039
well as add the sound effects.

107
00:05:46,400 --> 00:05:48,880
For the key press, I used a building

108
00:05:48,880 --> 00:05:52,800
function called bygame.key get pressed.

109
00:05:52,800 --> 00:05:55,360
It continuously checks for identifying

110
00:05:55,360 --> 00:05:59,360
any keys. The pressing the up arrow

111
00:05:59,360 --> 00:06:01,680
makes the Y value of the drone be

112
00:06:01,680 --> 00:06:04,560
decremented so that the drone goes up by

113
00:06:04,560 --> 00:06:07,520
a factor of 10 pixels. The down arrow

114
00:06:07,520 --> 00:06:09,919
increments the drone's Y value by 10

115
00:06:09,919 --> 00:06:14,039
pixels moving it down.

116
00:06:14,080 --> 00:06:17,039
The game shows 30 frames per second

117
00:06:17,039 --> 00:06:21,360
using clawframe tick 30. The drone in a

118
00:06:21,360 --> 00:06:23,840
game moves only up and down. Its

119
00:06:23,840 --> 00:06:27,039
movement in the X-axis is locked. The

120
00:06:27,039 --> 00:06:29,039
only user controlled object in the game

121
00:06:29,039 --> 00:06:31,759
is the drone. The up and down arrow key

122
00:06:31,759 --> 00:06:33,759
presses move the drone along the

123
00:06:33,759 --> 00:06:36,560
vertical Y-axis. The up and down arrow

124
00:06:36,560 --> 00:06:39,039
keys also need to be constantly pressed

125
00:06:39,039 --> 00:06:42,080
to keep the drone afloat since it will

126
00:06:42,080 --> 00:06:46,919
fall down due to gravity by 5 pixels.

127
00:06:46,960 --> 00:06:49,440
The obstacles move towards the drone

128
00:06:49,440 --> 00:06:51,919
from right to left of the screen 10

129
00:06:51,919 --> 00:06:55,360
pixels per frame. This stimulates the

130
00:06:55,360 --> 00:06:57,919
forward drone movement. And unlike the

131
00:06:57,919 --> 00:07:00,240
drone, the stallic mites and stallctides

132
00:07:00,240 --> 00:07:03,840
change their xaxis while the yaxis is

133
00:07:03,840 --> 00:07:06,840
locked.

134
00:07:07,199 --> 00:07:09,280
Similar to the obstacles, the cave

135
00:07:09,280 --> 00:07:12,479
background with cobwebs, bats, and

136
00:07:12,479 --> 00:07:15,120
spiders also move from right to left of

137
00:07:15,120 --> 00:07:18,240
the screen, but it is slower at 5 pixels

138
00:07:18,240 --> 00:07:21,599
per frame. This relative speed between

139
00:07:21,599 --> 00:07:23,599
the obstacles and the cave background

140
00:07:23,599 --> 00:07:27,199
create a parallax effect, giving an

141
00:07:27,199 --> 00:07:32,440
illusion of a 3D depth of the cave.

142
00:07:33,199 --> 00:07:37,039
In Pygame, blit or block image transfer

143
00:07:37,039 --> 00:07:39,840
means to draw, copy this image onto

144
00:07:39,840 --> 00:07:42,880
another surface. The drone layer has the

145
00:07:42,880 --> 00:07:45,599
transparent PNG image of the drone. And

146
00:07:45,599 --> 00:07:47,919
using the blit function, we position the

147
00:07:47,919 --> 00:07:50,240
drone layer at a specific X Y

148
00:07:50,240 --> 00:07:52,560
coordinate.

149
00:07:52,560 --> 00:07:55,759
Similarly, we blit the polygon layer and

150
00:07:55,759 --> 00:07:58,879
cave layer. Once all the layers are

151
00:07:58,879 --> 00:08:01,680
updated, we call pygame.d display.flip.

152
00:08:01,680 --> 00:08:05,759
flip. The function is used to update the

153
00:08:05,759 --> 00:08:08,240
actual game window to show everything we

154
00:08:08,240 --> 00:08:12,919
have changed since the previous update.

155
00:08:14,080 --> 00:08:16,240
The game has different shapes of

156
00:08:16,240 --> 00:08:18,639
stallmides and stallctites with random

157
00:08:18,639 --> 00:08:21,680
heights. The height is generated using

158
00:08:21,680 --> 00:08:25,039
the standard random function in Pygame.

159
00:08:25,039 --> 00:08:27,280
The obstacles are drawn using the

160
00:08:27,280 --> 00:08:30,720
built-in polygon, rectangle, and ellipse

161
00:08:30,720 --> 00:08:33,919
functions. First, the stallmite on the

162
00:08:33,919 --> 00:08:36,800
ground is drawn. And then after adding a

163
00:08:36,800 --> 00:08:40,240
constant gap of 300 pixels to its

164
00:08:40,240 --> 00:08:42,719
height, the stallctites are drawn by

165
00:08:42,719 --> 00:08:46,440
calculating the coordinates.

166
00:08:48,240 --> 00:08:50,560
Three steps are involved in the saving

167
00:08:50,560 --> 00:08:53,920
of the best score. When the game loads,

168
00:08:53,920 --> 00:08:57,120
we first look for the file score.txt in

169
00:08:57,120 --> 00:09:00,000
the drone game folder at the app data

170
00:09:00,000 --> 00:09:02,320
local location.

171
00:09:02,320 --> 00:09:05,040
If it doesn't exist, this is the first

172
00:09:05,040 --> 00:09:08,000
run of the game. So, the folder and the

173
00:09:08,000 --> 00:09:10,640
file are created and the variable high

174
00:09:10,640 --> 00:09:14,000
score is loaded with zero. When the game

175
00:09:14,000 --> 00:09:16,000
is finished, if the current high score

176
00:09:16,000 --> 00:09:19,760
is larger than the high score in the txt

177
00:09:19,760 --> 00:09:22,320
file, then it is written with the new

178
00:09:22,320 --> 00:09:25,320
score.

179
00:09:26,720 --> 00:09:29,440
While the game is being loaded, a 34

180
00:09:29,440 --> 00:09:32,080
second eerie background music with water

181
00:09:32,080 --> 00:09:34,640
flowing sound and an explosion sound

182
00:09:34,640 --> 00:09:37,920
effect clip are loaded. Their volumes

183
00:09:37,920 --> 00:09:40,080
can be independently controlled using

184
00:09:40,080 --> 00:09:43,680
Pygame DOB mixer settings and the

185
00:09:43,680 --> 00:09:45,839
background music is played in loop

186
00:09:45,839 --> 00:09:48,399
during the game play. When the drone

187
00:09:48,399 --> 00:09:51,200
crashes into an obstacle, the explosion

188
00:09:51,200 --> 00:09:54,000
sound is played once and the background

189
00:09:54,000 --> 00:09:58,040
music is stopped immediately.

190
00:09:58,880 --> 00:10:01,680
I'm proud of the ca drone game that I

191
00:10:01,680 --> 00:10:04,080
could do during my term break. I used

192
00:10:04,080 --> 00:10:06,399
Chad GPT to learn the ropes of game

193
00:10:06,399 --> 00:10:09,760
design. I overcame many challenges, but

194
00:10:09,760 --> 00:10:12,560
one cute bug in the game pointed out by

195
00:10:12,560 --> 00:10:14,880
my little sister during her testing was

196
00:10:14,880 --> 00:10:16,800
about the text showing the current score

197
00:10:16,800 --> 00:10:20,399
and best score. The score.txt is to stay

198
00:10:20,399 --> 00:10:22,880
visible as a top layer all the time in

199
00:10:22,880 --> 00:10:25,680
the game, but Stallides overlap it

200
00:10:25,680 --> 00:10:28,680
repeatedly.

201
00:10:30,720 --> 00:10:33,680
But I decided to keep the bug. This

202
00:10:33,680 --> 00:10:36,720
effect adds a depth to the field of the

203
00:10:36,720 --> 00:10:39,839
game and so I kept it as an imperfect

204
00:10:39,839 --> 00:10:42,839
perfection.

205
00:10:43,440 --> 00:10:46,000
I used many free assets for designing

206
00:10:46,000 --> 00:10:50,000
this game. Images are from magnafic.com

207
00:10:50,000 --> 00:10:52,399
and I edited some of the images in Adobe

208
00:10:52,399 --> 00:10:55,440
Photoshop. Sound effects are downloaded

209
00:10:55,440 --> 00:10:58,720
from pixabay.com and the audio editing

210
00:10:58,720 --> 00:11:01,760
was done in Audacity. The game's title

211
00:11:01,760 --> 00:11:04,959
font is downloaded from Google Fonts and

212
00:11:04,959 --> 00:11:06,640
this PowerPoint template is from

213
00:11:06,640 --> 00:11:09,640
slidescope.com.

214
00:11:10,880 --> 00:11:13,040
That's all for today. Thank you very

215
00:11:13,040 --> 00:11:15,279
much for your time and kindly connect

216
00:11:15,279 --> 00:11:17,440
with me on LinkedIn. You can either scan

217
00:11:17,440 --> 00:11:20,000
the QR code here or find me at

218
00:11:20,000 --> 00:11:23,000
linkedin.com/in.

219
00:11:27,102 --> 00:11:28,560
[applause]

220
00:11:28,560 --> 00:11:31,440
Massive round of applause. [screaming]

221
00:11:31,440 --> 00:11:34,720
That was the fantastic

222
00:11:34,720 --> 00:11:37,760
Alla. Next up, we have Dev. While I'm

223
00:11:37,760 --> 00:11:41,760
stalling, unplug this. Gosh, I have two.

224
00:11:41,760 --> 00:11:45,519
Yeah, there you go. Good luck. Um, so,

225
00:11:45,519 --> 00:11:47,040
uh,

226
00:11:47,040 --> 00:11:49,120
that was Alla her she has been

227
00:11:49,120 --> 00:11:51,839
programming for four years. Her favorite

228
00:11:51,839 --> 00:11:54,880
emoji I will act out and you all have to

229
00:11:54,880 --> 00:11:57,839
guess it. [snorts]

230
00:11:57,839 --> 00:12:02,800
Yeah, nice. Uh, and she recently learned

231
00:12:02,800 --> 00:12:04,079
that she really loves the latest

232
00:12:04,079 --> 00:12:06,880
framework laptop 13 Pro, which can be

233
00:12:06,880 --> 00:12:08,480
customized and upgraded. This sounds

234
00:12:08,480 --> 00:12:10,880
like a plug. uh in the future based on

235
00:12:10,880 --> 00:12:12,880
my growing demands. It really celebrates

236
00:12:12,880 --> 00:12:14,639
my right to repair. Okay, good.

237
00:12:14,639 --> 00:12:16,720
Modularity and continued support for

238
00:12:16,720 --> 00:12:19,279
many years. Excellent. Sponsor next year

239
00:12:19,279 --> 00:12:24,000
maybe. Um okay, so up on deck we have

240
00:12:24,000 --> 00:12:26,160
Dev. Dev is going to talk to us about

241
00:12:26,160 --> 00:12:28,959
Connect 4, which I assume is how you

242
00:12:28,959 --> 00:12:30,480
procrastinate in class. Don't answer

243
00:12:30,480 --> 00:12:33,920
that. Um uh Dev has been programming for

244
00:12:33,920 --> 00:12:37,120
3 years. Uh, you started in year 8 when

245
00:12:37,120 --> 00:12:40,480
you were approximately 13 years old. Uh,

246
00:12:40,480 --> 00:12:45,000
your favorite emoji is

247
00:12:45,519 --> 00:12:49,519
Yeah. blush emoji. Yes.

248
00:12:49,519 --> 00:12:52,399
Um, and something you recently learned

249
00:12:52,399 --> 00:12:55,600
about computers is that AI has stolen

250
00:12:55,600 --> 00:12:57,440
more data than quantum computing

251
00:12:57,440 --> 00:13:00,000
breaches ever will. And with that, can

252
00:13:00,000 --> 00:13:01,519
we get a massive round of applause for

253
00:13:01,519 --> 00:13:04,098
Dev [applause]

254
00:13:07,680 --> 00:13:10,160
Hi everyone. Thank you for coming to see

255
00:13:10,160 --> 00:13:11,839
our education talks. Today I'll be

256
00:13:11,839 --> 00:13:14,480
talking to you guys about an AI made to

257
00:13:14,480 --> 00:13:18,079
play Connect 4. So

258
00:13:18,079 --> 00:13:21,440
firstly, who am I? Hi, my name is Dev.

259
00:13:21,440 --> 00:13:23,519
My I'm a year 11 student and I go to

260
00:13:23,519 --> 00:13:25,120
Bergman Anglican School and I'm also

261
00:13:25,120 --> 00:13:28,079
studying cyber security at UNSW. My

262
00:13:28,079 --> 00:13:30,800
favorite colors are blue and purple and

263
00:13:30,800 --> 00:13:35,279
uh Oh, hi I like to win.

264
00:13:35,279 --> 00:13:36,240
So, [gasps]

265
00:13:36,240 --> 00:13:38,079
the problem I've had is that all my

266
00:13:38,079 --> 00:13:40,399
friends suck at Connect 4 and I really

267
00:13:40,399 --> 00:13:43,360
wanted to challenge myself. The solution

268
00:13:43,360 --> 00:13:45,519
was to build an AI that I could play

269
00:13:45,519 --> 00:13:48,079
with and have it powered by the miniax

270
00:13:48,079 --> 00:13:51,519
algorithm with alpha beta pruning.

271
00:13:51,519 --> 00:13:52,720
So, raise your hand if you've played

272
00:13:52,720 --> 00:13:55,040
Connect 4 before, guys. Nice. Okay,

273
00:13:55,040 --> 00:13:57,519
perfect. So, we all know Connect 4 is a

274
00:13:57,519 --> 00:13:59,199
six row by seven column game where

275
00:13:59,199 --> 00:14:01,360
players take turns to drop counters in a

276
00:14:01,360 --> 00:14:02,800
grid trying to make four in a We

277
00:14:02,800 --> 00:14:04,639
understand that. I'll explain my

278
00:14:04,639 --> 00:14:05,920
algorithm first in a little bit of

279
00:14:05,920 --> 00:14:07,920
detail in the context of chess because I

280
00:14:07,920 --> 00:14:09,199
feel like everyone knows chess. It's

281
00:14:09,199 --> 00:14:11,440
pretty popular, right? Um, basically how

282
00:14:11,440 --> 00:14:13,600
this works is my algorithm is a search

283
00:14:13,600 --> 00:14:16,639
algorithm. Basically, you foresee the

284
00:14:16,639 --> 00:14:19,360
future going forward several moves until

285
00:14:19,360 --> 00:14:21,440
you reach a point where you perform an

286
00:14:21,440 --> 00:14:24,800
evaluation. So, let's assume in our

287
00:14:24,800 --> 00:14:27,120
chess example, right, that we have a

288
00:14:27,120 --> 00:14:29,760
white and a black player. The white

289
00:14:29,760 --> 00:14:31,279
player is trying to maximize their

290
00:14:31,279 --> 00:14:33,680
evaluation. hence pick numbers or

291
00:14:33,680 --> 00:14:35,839
evaluations that are higher and black is

292
00:14:35,839 --> 00:14:37,600
trying to minimize or pick smaller

293
00:14:37,600 --> 00:14:38,959
numbers. So you can see in the number

294
00:14:38,959 --> 00:14:43,120
line is five, black can be neg five. So

295
00:14:43,120 --> 00:14:45,360
if we figure out that the two positions

296
00:14:45,360 --> 00:14:48,160
in the bottom are valued at -4 and

297
00:14:48,160 --> 00:14:50,399
three. White as a maximizing player

298
00:14:50,399 --> 00:14:52,399
picks the best of the two options or the

299
00:14:52,399 --> 00:14:55,040
larger number it picks three. If we do

300
00:14:55,040 --> 00:14:57,040
that for the other side of the branch

301
00:14:57,040 --> 00:14:59,199
five and one, white also picks five.

302
00:14:59,199 --> 00:15:01,360
Black is a minimizing player. So, black

303
00:15:01,360 --> 00:15:03,760
will pick three. Now, if we do the same

304
00:15:03,760 --> 00:15:05,680
logic on the other side, we get4 for

305
00:15:05,680 --> 00:15:08,480
white. We get 9 for white again. Black

306
00:15:08,480 --> 00:15:10,639
picks4 again, the smallest number. So,

307
00:15:10,639 --> 00:15:12,639
what do you guys think white will pick

308
00:15:12,639 --> 00:15:16,560
in this scenario? Three org4?

309
00:15:16,560 --> 00:15:18,639
You guys are geniuses. It's three. Great

310
00:15:18,639 --> 00:15:22,160
job. Okay, now that we've uh covered the

311
00:15:22,160 --> 00:15:23,920
algorithm that I'm going to be using, uh

312
00:15:23,920 --> 00:15:25,360
I have to actually figure out how I'm

313
00:15:25,360 --> 00:15:27,519
going to implement it. And I ran into a

314
00:15:27,519 --> 00:15:30,160
little problem. I didn't actually know

315
00:15:30,160 --> 00:15:31,760
what I was going to use to visualize the

316
00:15:31,760 --> 00:15:34,000
brain of the AI. So, at first I was

317
00:15:34,000 --> 00:15:36,079
doing some research and I thought, hm, I

318
00:15:36,079 --> 00:15:37,920
can use this library called Manom, which

319
00:15:37,920 --> 00:15:39,680
is a Python library that you can use to

320
00:15:39,680 --> 00:15:42,720
programmatically create math animations.

321
00:15:42,720 --> 00:15:44,240
Here are some photos of what it looks

322
00:15:44,240 --> 00:15:46,560
like. This this is a photo of some

323
00:15:46,560 --> 00:15:48,959
circular geometry. And this is an

324
00:15:48,959 --> 00:15:51,279
example of uh visualizing a neural

325
00:15:51,279 --> 00:15:54,720
network. However, as I would soon come

326
00:15:54,720 --> 00:15:56,160
to discover after reading the

327
00:15:56,160 --> 00:15:59,040
documentation is that there's no way to

328
00:15:59,040 --> 00:16:01,279
actually visualize data in real time

329
00:16:01,279 --> 00:16:04,720
using manom with the exception of OpenGL

330
00:16:04,720 --> 00:16:06,480
which is a variant of manom but that

331
00:16:06,480 --> 00:16:08,639
only works on your desktop and there's

332
00:16:08,639 --> 00:16:10,560
no actual way to port that to the web

333
00:16:10,560 --> 00:16:13,199
which is very unfortunate. So uh yeah,

334
00:16:13,199 --> 00:16:15,600
no madam.

335
00:16:15,600 --> 00:16:17,279
However, after doing some more research,

336
00:16:17,279 --> 00:16:20,000
I decided to use D3JS, which is a

337
00:16:20,000 --> 00:16:21,360
JavaScript library that lets you

338
00:16:21,360 --> 00:16:23,680
visualize data and it's really pretty. I

339
00:16:23,680 --> 00:16:25,199
picked this because it works really well

340
00:16:25,199 --> 00:16:26,639
with the front end, which I've coded in

341
00:16:26,639 --> 00:16:29,440
TypeScript. And um yeah, so after

342
00:16:29,440 --> 00:16:31,040
figuring out how to solve that problem,

343
00:16:31,040 --> 00:16:32,560
I had another small problem which was

344
00:16:32,560 --> 00:16:34,399
that if I were to use the algorithm I

345
00:16:34,399 --> 00:16:36,560
just explained to you guys, then uh it

346
00:16:36,560 --> 00:16:38,800
would be pretty inefficient. So the time

347
00:16:38,800 --> 00:16:40,399
complexity which is basically like how

348
00:16:40,399 --> 00:16:43,360
good or bad a efficiency of algorithm is

349
00:16:43,360 --> 00:16:45,920
the time complexity for this is O of the

350
00:16:45,920 --> 00:16:47,440
branching factor to the power of the

351
00:16:47,440 --> 00:16:49,839
depth of the tree which uh in layman

352
00:16:49,839 --> 00:16:51,920
terms just means that uh it's not

353
00:16:51,920 --> 00:16:54,800
efficient. We need to fix it. So the

354
00:16:54,800 --> 00:16:58,880
solution is if I use alpha beta pruning.

355
00:16:58,880 --> 00:17:00,880
This is a variant of the same algorithm

356
00:17:00,880 --> 00:17:02,320
that lets you ignore the branches that

357
00:17:02,320 --> 00:17:05,199
don't influence your final decision. So

358
00:17:05,199 --> 00:17:07,760
um what I have up here now are the time

359
00:17:07,760 --> 00:17:09,600
complexities with the new variant.

360
00:17:09,600 --> 00:17:12,799
Basically what it means is um at best we

361
00:17:12,799 --> 00:17:15,439
can perform a search twice as deep at

362
00:17:15,439 --> 00:17:17,600
the exact same time as vanilla miniax.

363
00:17:17,600 --> 00:17:19,600
So it's basically twice as good. At the

364
00:17:19,600 --> 00:17:22,160
average case you can probably go about

365
00:17:22,160 --> 00:17:25,360
33% deeper than the vanilla miniax. And

366
00:17:25,360 --> 00:17:27,839
um yeah if we continue on with our

367
00:17:27,839 --> 00:17:29,919
visualization we have this tree. This is

368
00:17:29,919 --> 00:17:31,120
the same tree as before. We have

369
00:17:31,120 --> 00:17:33,200
negative 1 and three. So, white picks

370
00:17:33,200 --> 00:17:35,600
three, the maximum value. We have five.

371
00:17:35,600 --> 00:17:37,760
But instead of evaluating the final

372
00:17:37,760 --> 00:17:40,160
black node, we can simply realize that

373
00:17:40,160 --> 00:17:41,760
white is obviously going to pick a value

374
00:17:41,760 --> 00:17:44,400
that's five or greater. Henceforth, we

375
00:17:44,400 --> 00:17:47,440
genuinely don't need to know what that

376
00:17:47,440 --> 00:17:49,039
uh node is. We can just leave it as a

377
00:17:49,039 --> 00:17:50,559
question mark and we can effectively cut

378
00:17:50,559 --> 00:17:52,880
the branch getting rid of it. And uh

379
00:17:52,880 --> 00:17:54,080
that's really good because we don't have

380
00:17:54,080 --> 00:17:55,520
to calculate these types of nodes that

381
00:17:55,520 --> 00:17:57,440
are useless to the final result. So, we

382
00:17:57,440 --> 00:17:58,960
can basically cut that off. If we

383
00:17:58,960 --> 00:18:00,320
continue with our visualization with

384
00:18:00,320 --> 00:18:03,360
four and negative6, we get neg4. And um

385
00:18:03,360 --> 00:18:05,360
we know that black at the top here is

386
00:18:05,360 --> 00:18:07,600
going to pick a value that is -4 or

387
00:18:07,600 --> 00:18:10,640
lower. And if we look to the um other

388
00:18:10,640 --> 00:18:12,400
side, which has three, we know that

389
00:18:12,400 --> 00:18:13,919
white obviously needs to pick a value

390
00:18:13,919 --> 00:18:15,520
that is large. So it's not going to pick

391
00:18:15,520 --> 00:18:17,520
neg4, it's going to pick three. So we

392
00:18:17,520 --> 00:18:19,120
can effectively get rid of that side of

393
00:18:19,120 --> 00:18:21,760
the branch. And uh yeah, leaving us with

394
00:18:21,760 --> 00:18:24,240
three. I'll go through a uh quick demo

395
00:18:24,240 --> 00:18:29,440
now just to show what it looks like.

396
00:18:32,160 --> 00:18:34,559
Yeah. So, this is what it looks like. We

397
00:18:34,559 --> 00:18:38,360
have an animation

398
00:18:40,480 --> 00:18:45,960
that's using the uh Gap library and

399
00:18:46,000 --> 00:18:47,679
I'll play it on hard difficulty just so

400
00:18:47,679 --> 00:18:49,360
you guys can see it. You can see at the

401
00:18:49,360 --> 00:18:53,039
top uh top right we have the search tree

402
00:18:53,039 --> 00:18:56,000
visualization using the D3JS which I had

403
00:18:56,000 --> 00:18:58,320
to pivot to uh the architecture change.

404
00:18:58,320 --> 00:19:00,799
Below that you can see the stats for the

405
00:19:00,799 --> 00:19:03,840
AI that's uh how good it thinks each

406
00:19:03,840 --> 00:19:05,679
move is. You can check how long it takes

407
00:19:05,679 --> 00:19:08,240
to evaluate each position as well. And

408
00:19:08,240 --> 00:19:11,360
um below that is the AI column scores

409
00:19:11,360 --> 00:19:13,280
which is basically calculating how good

410
00:19:13,280 --> 00:19:15,760
it thinks each column is uh for it to

411
00:19:15,760 --> 00:19:18,000
place. Um, so yeah, that's basically the

412
00:19:18,000 --> 00:19:21,640
gist of the demo.

413
00:19:23,293 --> 00:19:24,799
[applause]

414
00:19:24,799 --> 00:19:26,880
All right, so

415
00:19:26,880 --> 00:19:28,720
I'll talk about the uh main libraries I

416
00:19:28,720 --> 00:19:30,880
used in a little bit of detail. So D3JS

417
00:19:30,880 --> 00:19:32,240
obviously that's what I used for the

418
00:19:32,240 --> 00:19:33,840
visualization when you saw the node

419
00:19:33,840 --> 00:19:36,720
tree. Um, we use recharts here for the

420
00:19:36,720 --> 00:19:39,120
chart you saw at the bottom. that um

421
00:19:39,120 --> 00:19:41,360
helps because it gives us a clear easy

422
00:19:41,360 --> 00:19:42,720
visualization that humans actually

423
00:19:42,720 --> 00:19:44,720
understand and it complements the tree

424
00:19:44,720 --> 00:19:46,640
diagram pretty well in my opinion and

425
00:19:46,640 --> 00:19:48,559
fast API is my bridge between the front

426
00:19:48,559 --> 00:19:50,559
end and the back end and I use it to

427
00:19:50,559 --> 00:19:53,520
expose endpoints like the /ai move uh

428
00:19:53,520 --> 00:19:56,799
state and reset um endpoints that the

429
00:19:56,799 --> 00:19:59,440
function uh front end will call on each

430
00:19:59,440 --> 00:20:02,640
turn. Here's a laundry list of other

431
00:20:02,640 --> 00:20:05,120
stuff. Uh the main things uh from this

432
00:20:05,120 --> 00:20:08,320
list are I use shad CN for the pre-built

433
00:20:08,320 --> 00:20:10,240
um pre-built front-end components.

434
00:20:10,240 --> 00:20:12,000
They're styled with tailwind and CSS.

435
00:20:12,000 --> 00:20:13,679
For the animation you guys saw that said

436
00:20:13,679 --> 00:20:16,480
Python 2026, that's using Gap and I used

437
00:20:16,480 --> 00:20:20,240
frame motion for the board drop. Okay,

438
00:20:20,240 --> 00:20:22,400
so this is a little bit on workflow

439
00:20:22,400 --> 00:20:24,480
because I made this for a school project

440
00:20:24,480 --> 00:20:27,840
and I had to learn how to use GitHub. So

441
00:20:27,840 --> 00:20:29,440
fun little learning experience here. I

442
00:20:29,440 --> 00:20:31,679
used two branches and I had like roughly

443
00:20:31,679 --> 00:20:34,080
55 commits and it took me like two to

444
00:20:34,080 --> 00:20:36,880
four weeks to make this. Um, see my

445
00:20:36,880 --> 00:20:38,320
teacher forced us to use more than one

446
00:20:38,320 --> 00:20:41,239
branch but like I didn't really

447
00:20:41,239 --> 00:20:41,760
[laughter]

448
00:20:41,760 --> 00:20:45,840
like I didn't really need to cuz like I

449
00:20:45,840 --> 00:20:48,320
didn't actually um I had to um I used

450
00:20:48,320 --> 00:20:49,760
one branch for front end and the other

451
00:20:49,760 --> 00:20:51,360
one for the back end but I feel like it

452
00:20:51,360 --> 00:20:52,640
would have been simplistic just to use

453
00:20:52,640 --> 00:20:55,120
one but all right so what did I learn

454
00:20:55,120 --> 00:20:57,600
really quickly? So, uh, making UIs are

455
00:20:57,600 --> 00:20:59,440
brutal, and I think this project just

456
00:20:59,440 --> 00:21:00,960
made me realize how much respect I have

457
00:21:00,960 --> 00:21:02,240
for a lot of front-end developers out

458
00:21:02,240 --> 00:21:03,280
there, especially making things

459
00:21:03,280 --> 00:21:05,200
accessible, making things look pretty.

460
00:21:05,200 --> 00:21:07,039
Uh, make your algorithms better, faster,

461
00:21:07,039 --> 00:21:08,640
stronger, and research what it should

462
00:21:08,640 --> 00:21:10,240
look like instead of just making a bad

463
00:21:10,240 --> 00:21:12,240
one. Uh, read the docs before you marry

464
00:21:12,240 --> 00:21:14,080
a library. And don't be stupid like I

465
00:21:14,080 --> 00:21:16,000
did. Uh, and yeah, my friends won't play

466
00:21:16,000 --> 00:21:17,919
me because I'm the best at Connect 4.

467
00:21:17,919 --> 00:21:18,880
[laughter]

468
00:21:18,880 --> 00:21:20,240
Here are some socials if you want to

469
00:21:20,240 --> 00:21:22,159
take a look at the code. The the code is

470
00:21:22,159 --> 00:21:23,840
on the the GitHub QR code and my

471
00:21:23,840 --> 00:21:25,200
LinkedIn's on the other side. And uh

472
00:21:25,200 --> 00:21:28,399
yeah, thank you guys for listening.

473
00:21:28,399 --> 00:21:30,419
[applause]

474
00:21:32,799 --> 00:21:36,320
Thank you. Next up, we have the fabulous

475
00:21:36,320 --> 00:21:40,480
Tun and Anoo. And while they connect, oh

476
00:21:40,480 --> 00:21:43,200
gosh, there's setup. Uh while they

477
00:21:43,200 --> 00:21:46,240
connect and get ready, I'm going to tell

478
00:21:46,240 --> 00:21:50,240
you about them. Uh they have been

479
00:21:50,240 --> 00:21:54,320
programming for seven and 3 years

480
00:21:54,320 --> 00:21:57,440
respectively and their favorite emoji

481
00:21:57,440 --> 00:22:00,400
are

482
00:22:00,400 --> 00:22:02,159
yawning and I really don't want to do

483
00:22:02,159 --> 00:22:05,120
this one.

484
00:22:05,120 --> 00:22:07,679
Yeah, that kind of Oh gosh. Yeah, my

485
00:22:07,679 --> 00:22:10,159
brain is definitely shaking now. Um

486
00:22:10,159 --> 00:22:12,000
Terrone recently learned that compilers

487
00:22:12,000 --> 00:22:14,799
don't read code line by line. They first

488
00:22:14,799 --> 00:22:16,400
pass it into a tree structure that

489
00:22:16,400 --> 00:22:18,000
represents the program before checking

490
00:22:18,000 --> 00:22:20,960
it and generating machine code and

491
00:22:20,960 --> 00:22:24,880
recently learned uh was ASA path finding

492
00:22:24,880 --> 00:22:27,120
algorithm uh which uses a heruristic to

493
00:22:27,120 --> 00:22:29,039
help determine the fastest path in a

494
00:22:29,039 --> 00:22:30,720
graph and they are going to teach

495
00:22:30,720 --> 00:22:31,840
they're going to here to talk about

496
00:22:31,840 --> 00:22:34,799
neither of those two things but instead

497
00:22:34,799 --> 00:22:37,919
um from vision to braille massive round

498
00:22:37,919 --> 00:22:40,450
of applause.

499
00:22:40,450 --> 00:22:42,470
[applause]

500
00:22:49,919 --> 00:22:54,200
more applause. Okay.

501
00:22:54,860 --> 00:22:56,825
[laughter]

502
00:22:56,825 --> 00:22:58,845
[applause]

503
00:22:59,039 --> 00:23:01,840
Oh, wow. Did you know I recently learned

504
00:23:01,840 --> 00:23:04,480
that they called this a confidence

505
00:23:04,480 --> 00:23:05,919
monitor?

506
00:23:05,919 --> 00:23:08,720
Yeah. Uh, and this gave me confidence

507
00:23:08,720 --> 00:23:11,891
that may in fact have been misplaced.

508
00:23:11,891 --> 00:23:13,911
[laughter]

509
00:23:14,080 --> 00:23:16,320
Um, but I do like I do like the naming

510
00:23:16,320 --> 00:23:19,039
convention of it. Um, okay. We're we're

511
00:23:19,039 --> 00:23:20,400
going to be I'm just going to steal

512
00:23:20,400 --> 00:23:21,919
their thunder and steal their talk now.

513
00:23:21,919 --> 00:23:23,039
We're going to be learning about from

514
00:23:23,039 --> 00:23:26,240
vision to braille building a tactile

515
00:23:26,240 --> 00:23:28,880
reading system with Python, which I'm

516
00:23:28,880 --> 00:23:33,280
imagining will involve these things.

517
00:23:33,280 --> 00:23:35,760
AAA

518
00:23:35,760 --> 00:23:40,039
batteries don't cause fires, right?

519
00:23:40,711 --> 00:23:41,600
[laughter]

520
00:23:41,600 --> 00:23:42,880
Okay, one more massive round of

521
00:23:42,880 --> 00:23:45,296
applause. [cheering]

522
00:23:45,296 --> 00:23:47,316
[applause]

523
00:23:49,760 --> 00:23:52,559
Good afternoon everyone. My name is Tun

524
00:23:52,559 --> 00:23:54,000
and I'm Ambu.

525
00:23:54,000 --> 00:23:55,280
Today we're going to present our project

526
00:23:55,280 --> 00:23:57,919
Brail Vision.

527
00:23:57,919 --> 00:24:00,559
So what is Brail Vision? So back in year

528
00:24:00,559 --> 00:24:02,559
10, we wanted to experiment with OCR in

529
00:24:02,559 --> 00:24:05,360
Python. We're testing how Python could

530
00:24:05,360 --> 00:24:08,320
use a camera to extract and recognize

531
00:24:08,320 --> 00:24:10,320
information. From there, we started

532
00:24:10,320 --> 00:24:11,840
thinking about the actual uses that it

533
00:24:11,840 --> 00:24:14,320
could have. We looked at accessibility

534
00:24:14,320 --> 00:24:16,000
and we thought we could use it to help

535
00:24:16,000 --> 00:24:18,159
blind and visually impaired people read

536
00:24:18,159 --> 00:24:20,640
physically.

537
00:24:20,640 --> 00:24:22,640
How does our braille vision work? It

538
00:24:22,640 --> 00:24:24,799
takes an input from the camera and then

539
00:24:24,799 --> 00:24:28,000
that's converted for easy OCR into text.

540
00:24:28,000 --> 00:24:29,840
This is then corrected through our

541
00:24:29,840 --> 00:24:32,400
algorithms and converted into braille

542
00:24:32,400 --> 00:24:34,159
which is represented through our

543
00:24:34,159 --> 00:24:38,360
physical braille solenoids.

544
00:24:38,799 --> 00:24:41,440
So we use easy OCR to detect printed

545
00:24:41,440 --> 00:24:44,640
text from the camera. Um easy OCR gives

546
00:24:44,640 --> 00:24:46,559
us both the detected text and the

547
00:24:46,559 --> 00:24:49,200
confidence score. This confidence score

548
00:24:49,200 --> 00:24:51,919
is important because OCR isn't always a

549
00:24:51,919 --> 00:24:55,200
reliable source. Instead um treating

550
00:24:55,200 --> 00:24:56,880
instead of treating every detection the

551
00:24:56,880 --> 00:24:59,440
same, we use the confidence to decide

552
00:24:59,440 --> 00:25:02,960
how much correction should be allowed.

553
00:25:02,960 --> 00:25:05,919
Easy OCR can be unreliable at times

554
00:25:05,919 --> 00:25:08,080
based on lighting and angles of the

555
00:25:08,080 --> 00:25:10,080
image.

556
00:25:10,080 --> 00:25:12,640
We tried using a grayscale and altering

557
00:25:12,640 --> 00:25:14,640
the contrast of the image, but it made

558
00:25:14,640 --> 00:25:16,799
the accuracy worse than using original

559
00:25:16,799 --> 00:25:19,440
images. We found out the reason why is

560
00:25:19,440 --> 00:25:21,840
because OCR was trained across a wide

561
00:25:21,840 --> 00:25:25,600
variety of natural images.

562
00:25:25,600 --> 00:25:27,919
As such, we've had to break down uh the

563
00:25:27,919 --> 00:25:30,000
confidence the results of the confidence

564
00:25:30,000 --> 00:25:32,640
scores so we can actually make use of

565
00:25:32,640 --> 00:25:35,760
this text. If it's below 20%, we

566
00:25:35,760 --> 00:25:38,400
disregard the detection altogether.

567
00:25:38,400 --> 00:25:40,960
Between 20 and 35%

568
00:25:40,960 --> 00:25:43,760
uh we dis we keep the detection but

569
00:25:43,760 --> 00:25:46,400
avoid general dictionary corrections.

570
00:25:46,400 --> 00:25:49,360
between 35 and 80% we allow dictionary

571
00:25:49,360 --> 00:25:52,480
corrections within two characters

572
00:25:52,480 --> 00:25:55,360
and if if it's above 80% we trust the

573
00:25:55,360 --> 00:25:59,400
OCR and move on.

574
00:26:00,159 --> 00:26:02,000
So small typos can greatly alter the

575
00:26:02,000 --> 00:26:03,520
meaning and affect understanding of

576
00:26:03,520 --> 00:26:06,000
text. So for general spelling errors we

577
00:26:06,000 --> 00:26:08,799
use Python module SIM spell. So Sim

578
00:26:08,799 --> 00:26:10,240
spell searches it dictionary for

579
00:26:10,240 --> 00:26:12,400
possible corrections. We use the max

580
00:26:12,400 --> 00:26:13,919
edit distance to control how much

581
00:26:13,919 --> 00:26:16,799
characters it's allowed to change.

582
00:26:16,799 --> 00:26:18,960
If it finds several suggestions, we take

583
00:26:18,960 --> 00:26:21,039
the highest ranked one. If it can't find

584
00:26:21,039 --> 00:26:23,360
any suitable, we keep the original word.

585
00:26:23,360 --> 00:26:25,200
But using the same edit distance for

586
00:26:25,200 --> 00:26:28,480
every word did have another problem.

587
00:26:28,480 --> 00:26:30,320
An edit distance of one had a much

588
00:26:30,320 --> 00:26:32,000
bigger impact on shorter words compared

589
00:26:32,000 --> 00:26:34,080
to longer words. So we had to make a new

590
00:26:34,080 --> 00:26:35,919
method to determine the edit distance

591
00:26:35,919 --> 00:26:38,480
where we combine both the confidence

592
00:26:38,480 --> 00:26:42,000
score and word length.

593
00:26:42,000 --> 00:26:44,640
A dictionary still has one major

594
00:26:44,640 --> 00:26:46,960
limitation. OCR can produce the wrong

595
00:26:46,960 --> 00:26:49,760
word, but that wrong word might still be

596
00:26:49,760 --> 00:26:52,240
a real English word. For example, the

597
00:26:52,240 --> 00:26:55,679
dog ran Rome contains only valid words.

598
00:26:55,679 --> 00:26:57,919
So, Sim spell has no reason to change

599
00:26:57,919 --> 00:27:00,080
Rome. Originally, we tried using the

600
00:27:00,080 --> 00:27:03,360
charge API to correct words using

601
00:27:03,360 --> 00:27:06,000
context. But without a stable Wi-Fi, it

602
00:27:06,000 --> 00:27:09,039
ran too slow and was ineffective for us.

603
00:27:09,039 --> 00:27:12,159
That's where distill belt distill came

604
00:27:12,159 --> 00:27:15,679
in. Our code selects a word um and to

605
00:27:15,679 --> 00:27:18,640
check and replace it with a mask. It

606
00:27:18,640 --> 00:27:20,400
then generates plausible alternatives

607
00:27:20,400 --> 00:27:23,600
such as sum and home.

608
00:27:23,600 --> 00:27:25,120
This tool belt uses a rest of the

609
00:27:25,120 --> 00:27:27,760
sentence to score which candidate fits

610
00:27:27,760 --> 00:27:30,159
the context better. In this case, the

611
00:27:30,159 --> 00:27:35,080
final correct output is dog ran home.

612
00:27:35,200 --> 00:27:37,760
So before distiller scores anything, we

613
00:27:37,760 --> 00:27:39,679
need to restrict the candidate set. We

614
00:27:39,679 --> 00:27:41,919
use simf spell to generate um possible

615
00:27:41,919 --> 00:27:43,919
alternatives with like a head distance

616
00:27:43,919 --> 00:27:46,720
of one. That keeps alternatives visually

617
00:27:46,720 --> 00:27:48,320
similar to what the OCR originally

618
00:27:48,320 --> 00:27:50,640
detected. We also limit dictionary

619
00:27:50,640 --> 00:27:53,039
suggestions to eight words. Those

620
00:27:53,039 --> 00:27:54,480
suggest suggestions are stored as

621
00:27:54,480 --> 00:27:56,799
candidates for the still book. This is

622
00:27:56,799 --> 00:27:58,240
done so the model doesn't freely just

623
00:27:58,240 --> 00:28:00,080
change a sentence only ranks the best

624
00:28:00,080 --> 00:28:03,080
candidate.

625
00:28:03,200 --> 00:28:05,360
Once we have the candidates, we run the

626
00:28:05,360 --> 00:28:07,760
mask sentence through distiller. The

627
00:28:07,760 --> 00:28:09,360
code finds the exact position of the

628
00:28:09,360 --> 00:28:12,000
mask. Distiller outputs raw scores

629
00:28:12,000 --> 00:28:13,919
called logits possible words at that

630
00:28:13,919 --> 00:28:16,559
position. We convert those scores into

631
00:28:16,559 --> 00:28:18,880
log probabilities using the log softmax

632
00:28:18,880 --> 00:28:21,679
function. They can be compared. A higher

633
00:28:21,679 --> 00:28:23,760
log probability means the model

634
00:28:23,760 --> 00:28:25,360
considered that candidate more likely in

635
00:28:25,360 --> 00:28:27,279
the sentence.

636
00:28:27,279 --> 00:28:29,600
So how are logits comp converted into

637
00:28:29,600 --> 00:28:32,159
log probabilities? So at first the

638
00:28:32,159 --> 00:28:34,000
program gives you a raw score which

639
00:28:34,000 --> 00:28:36,159
isn't really too helpful.

640
00:28:36,159 --> 00:28:38,480
Then we use the softmax function formula

641
00:28:38,480 --> 00:28:40,799
to convert them into

642
00:28:40,799 --> 00:28:42,720
probabilities. It does this by

643
00:28:42,720 --> 00:28:44,559
exponentiating each score and dividing

644
00:28:44,559 --> 00:28:47,200
it by the total. These probabilities

645
00:28:47,200 --> 00:28:48,880
then are then put on a logarithmic scale

646
00:28:48,880 --> 00:28:50,880
to enhance the differences making it

647
00:28:50,880 --> 00:28:54,640
easier for the program to compare.

648
00:28:54,640 --> 00:28:56,159
We still don't automatically accept

649
00:28:56,159 --> 00:28:58,880
whichever candidate distes highest.

650
00:28:58,880 --> 00:29:01,039
First, we calculate how much better the

651
00:29:01,039 --> 00:29:02,960
best alternative scored compared to the

652
00:29:02,960 --> 00:29:05,440
original. That improvement must pass the

653
00:29:05,440 --> 00:29:08,159
minimum threshold. If easy OCR was

654
00:29:08,159 --> 00:29:10,240
already very confident, we increase the

655
00:29:10,240 --> 00:29:12,240
required threshold. So, the context

656
00:29:12,240 --> 00:29:14,480
model acts like a cautious opinion

657
00:29:14,480 --> 00:29:18,600
rather than having complete control.

658
00:29:18,640 --> 00:29:20,720
Once the text is corrected, the process

659
00:29:20,720 --> 00:29:22,720
becomes deterministic.

660
00:29:22,720 --> 00:29:25,039
Each character is mapped to a six dot

661
00:29:25,039 --> 00:29:28,080
brail pattern. This six pattern the six

662
00:29:28,080 --> 00:29:31,039
positions are one numbered one to six.

663
00:29:31,039 --> 00:29:33,120
For example, different letters comp

664
00:29:33,120 --> 00:29:35,520
correspond to different combinations of

665
00:29:35,520 --> 00:29:38,399
raised dots. The character is converted

666
00:29:38,399 --> 00:29:40,080
into a pattern before being sent to the

667
00:29:40,080 --> 00:29:42,240
hardware.

668
00:29:42,240 --> 00:29:45,360
This is the full physical circuit. Um

669
00:29:45,360 --> 00:29:47,360
the main parts of the ESP 32, the

670
00:29:47,360 --> 00:29:49,760
solenoids, the power source and the

671
00:29:49,760 --> 00:29:53,760
solenoid pistons and the relay module.

672
00:29:53,760 --> 00:29:56,320
The ESP32 access the control of our

673
00:29:56,320 --> 00:29:59,440
system. It hosts its own Wi-Fi which

674
00:29:59,440 --> 00:30:01,360
that code connects and sends over

675
00:30:01,360 --> 00:30:03,520
individual characters. It then

676
00:30:03,520 --> 00:30:05,120
determines which braille dots are

677
00:30:05,120 --> 00:30:08,000
necessary. The ESP32 itself does not

678
00:30:08,000 --> 00:30:10,880
power the solenoids.

679
00:30:10,880 --> 00:30:12,880
The relay module acts as a sort of

680
00:30:12,880 --> 00:30:15,200
electrical switch

681
00:30:15,200 --> 00:30:17,039
channels. It channels the power to the

682
00:30:17,039 --> 00:30:19,760
solenoids. The ESP32 tells individual

683
00:30:19,760 --> 00:30:21,840
relays when to switch. Each relay

684
00:30:21,840 --> 00:30:24,799
controls power to one solenoid. This

685
00:30:24,799 --> 00:30:27,360
means each braille dot can be controlled

686
00:30:27,360 --> 00:30:29,440
independently.

687
00:30:29,440 --> 00:30:31,600
The solenoids create physical braille

688
00:30:31,600 --> 00:30:34,399
dots. When powered, the solenoids rise

689
00:30:34,399 --> 00:30:35,679
to represent different braille

690
00:30:35,679 --> 00:30:37,679
characters. This is where the digital

691
00:30:37,679 --> 00:30:41,520
characters become something tactile.

692
00:30:41,520 --> 00:30:43,360
We will now visually demonstrate our

693
00:30:43,360 --> 00:30:45,440
project on video one.

694
00:30:45,440 --> 00:30:47,440
So now we'll go over a quick um demo of

695
00:30:47,440 --> 00:30:49,919
our project. On the laptop over here, we

696
00:30:49,919 --> 00:30:53,120
have the OCR running scanning for text.

697
00:30:53,120 --> 00:30:54,880
So now we're like pick up on the text on

698
00:30:54,880 --> 00:30:58,320
the notebook.

699
00:30:58,320 --> 00:31:00,480
The laptop individually sends characters

700
00:31:00,480 --> 00:31:02,799
to the ESP32.

701
00:31:02,799 --> 00:31:05,520
As you can see over here, the ESP32

702
00:31:05,520 --> 00:31:07,919
sends signals to the relay module and

703
00:31:07,919 --> 00:31:09,919
the relay module in turn powers the

704
00:31:09,919 --> 00:31:12,159
solenoids which represents individual

705
00:31:12,159 --> 00:31:15,640
braid patterns.

706
00:31:24,456 --> 00:31:26,476
[applause]

707
00:31:26,640 --> 00:31:29,041
Thank you so much for your time.

708
00:31:29,041 --> 00:31:33,039
[applause] Thank you for your time.

709
00:31:33,039 --> 00:31:36,640
Woo. That's amazing. Before you unplug

710
00:31:36,640 --> 00:31:38,960
this, you must hold this up so that the

711
00:31:38,960 --> 00:31:41,360
audience can see as well. I don't want

712
00:31:41,360 --> 00:31:43,120
to I I don't want to break that. That

713
00:31:43,120 --> 00:31:48,679
was amazing. Yes. Proof.

714
00:31:50,002 --> 00:31:51,039
[applause]

715
00:31:51,039 --> 00:31:55,120
Up on deck. Now we have Tara. And Tara's

716
00:31:55,120 --> 00:31:57,919
going to be talking to us about rocket

717
00:31:57,919 --> 00:32:01,840
science in the shower. Uh, which I guess

718
00:32:01,840 --> 00:32:04,000
if it does catch fire, you've got this.

719
00:32:04,000 --> 00:32:05,840
Yeah. Okay. Um, Tara has been

720
00:32:05,840 --> 00:32:10,880
programming for 1.5 years. And, um, her

721
00:32:10,880 --> 00:32:14,000
favorite emoji is

722
00:32:14,000 --> 00:32:16,480
I should have checked this.

723
00:32:16,480 --> 00:32:20,000
The rockstar one. Yes. Thank you. I'm so

724
00:32:20,000 --> 00:32:21,919
glad you all know what these emoji are

725
00:32:21,919 --> 00:32:25,760
named cuz I Yeah. Um she recently found

726
00:32:25,760 --> 00:32:29,440
out that Samsung was founded in 1938

727
00:32:29,440 --> 00:32:32,480
as a grocery trading company that

728
00:32:32,480 --> 00:32:36,000
exported dried Korean fish, vegetables,

729
00:32:36,000 --> 00:32:38,080
and fruit directly to cities in

730
00:32:38,080 --> 00:32:40,880
northeast China. I didn't know that

731
00:32:40,880 --> 00:32:42,960
either.

732
00:32:42,960 --> 00:32:46,320
Um, and she's also now discovering the

733
00:32:46,320 --> 00:32:49,519
complexities of HDMI cables. There we

734
00:32:49,519 --> 00:32:53,679
go. Um,

735
00:32:53,679 --> 00:32:56,000
now I have to revisit my earlier

736
00:32:56,000 --> 00:33:00,039
citation about confidence.

737
00:33:01,120 --> 00:33:02,640
While we have some confidence, can I get

738
00:33:02,640 --> 00:33:05,327
a massive round of applause for Tara?

739
00:33:05,327 --> 00:33:07,347
[applause]

740
00:33:09,679 --> 00:33:11,760
Now, despite what Nikki said, my project

741
00:33:11,760 --> 00:33:14,000
isn't exactly about rocket science. It's

742
00:33:14,000 --> 00:33:15,760
actually about Wordle, but rocket

743
00:33:15,760 --> 00:33:17,840
science is still part of the picture.

744
00:33:17,840 --> 00:33:20,159
So, I'm going to be playing an animation

745
00:33:20,159 --> 00:33:22,880
that sort of explains how I actually

746
00:33:22,880 --> 00:33:26,480
figured out the idea. So, let's begin.

747
00:33:26,480 --> 00:33:29,480
Okay.

748
00:33:33,840 --> 00:33:36,240
Yep. Ideas rarely start in a neat

749
00:33:36,240 --> 00:33:38,000
planning space. They hit you when you

750
00:33:38,000 --> 00:33:39,919
are completely vulnerable. Like standing

751
00:33:39,919 --> 00:33:42,000
under scalding water in the shower,

752
00:33:42,000 --> 00:33:43,840
staring blankly at the grid of bathroom

753
00:33:43,840 --> 00:33:46,320
war tiles. I was planning drafting a

754
00:33:46,320 --> 00:33:48,799
Python telemetry script to calculate

755
00:33:48,799 --> 00:33:51,360
orbital velocities and fuel burn curves

756
00:33:51,360 --> 00:33:53,919
for interstellar spacecraft. Rocket

757
00:33:53,919 --> 00:33:56,960
physics is clean and predictable. But as

758
00:33:56,960 --> 00:33:59,840
the water ran, my eyes locked onto the

759
00:33:59,840 --> 00:34:01,840
intersecting grout lines of the tiles

760
00:34:01,840 --> 00:34:04,399
and my brain flipped. I didn't want to

761
00:34:04,399 --> 00:34:06,480
launch rockets anymore. I wanted to

762
00:34:06,480 --> 00:34:08,560
build an engine to solve the daily

763
00:34:08,560 --> 00:34:10,879
Wordle. I leaped out of the shower,

764
00:34:10,879 --> 00:34:12,960
leaving wet footprints across the floor,

765
00:34:12,960 --> 00:34:15,440
booting up a blank IDE.

766
00:34:15,440 --> 00:34:17,440
Today's presentation will be providing a

767
00:34:17,440 --> 00:34:20,653
glimpse of what I did after.

768
00:34:20,653 --> 00:34:21,520
[laughter]

769
00:34:21,520 --> 00:34:23,040
For those of you who are wondering, I

770
00:34:23,040 --> 00:34:25,520
did that animation completely on HTML.

771
00:34:25,520 --> 00:34:29,639
It was really hard.

772
00:34:29,839 --> 00:34:32,960
A bit about me. Hi everyone. I'm Tara.

773
00:34:32,960 --> 00:34:35,119
I'm 13 years old. And a fun fact about

774
00:34:35,119 --> 00:34:36,639
me is I'm kind of obsessed with the

775
00:34:36,639 --> 00:34:41,013
periodic table. So yeah, [laughter]

776
00:34:41,919 --> 00:34:43,919
so this presentation will be following

777
00:34:43,919 --> 00:34:45,839
this structure starting with a context

778
00:34:45,839 --> 00:34:48,639
on Wordle entropy, a demo, and a little

779
00:34:48,639 --> 00:34:50,639
reflection about stuff I learned and the

780
00:34:50,639 --> 00:34:53,280
things I could improve.

781
00:34:53,280 --> 00:34:55,599
All right, Wordle is a simple guessing

782
00:34:55,599 --> 00:34:57,920
game. You have six attempts to find a

783
00:34:57,920 --> 00:35:00,160
fiveletter word. And after each guess,

784
00:35:00,160 --> 00:35:02,079
the game gives you information about

785
00:35:02,079 --> 00:35:04,880
three colors, green, yellow, and gray.

786
00:35:04,880 --> 00:35:06,720
But out of that, feedback is doing

787
00:35:06,720 --> 00:35:08,720
something much more important than

788
00:35:08,720 --> 00:35:10,480
simply telling us whether a letter is

789
00:35:10,480 --> 00:35:12,400
right or wrong. It is reducing

790
00:35:12,400 --> 00:35:15,040
uncertainty. We begin with thousands of

791
00:35:15,040 --> 00:35:17,280
possible words. Before our first guess,

792
00:35:17,280 --> 00:35:19,440
we have a huge amount of uncertainty

793
00:35:19,440 --> 00:35:21,760
about which one is correct. Then we make

794
00:35:21,760 --> 00:35:24,000
a guess. The five tiles return a pattern

795
00:35:24,000 --> 00:35:26,160
of colors, and that pattern allows us to

796
00:35:26,160 --> 00:35:28,240
eliminate words that could no longer be

797
00:35:28,240 --> 00:35:30,160
the answer. So, our goal isn't

798
00:35:30,160 --> 00:35:31,920
necessarily to guess the answer

799
00:35:31,920 --> 00:35:34,560
immediately. Our goal is to gain as much

800
00:35:34,560 --> 00:35:37,280
information as possible. But that

801
00:35:37,280 --> 00:35:39,839
creates a question, how can we measure

802
00:35:39,839 --> 00:35:42,480
information? This is where we need

803
00:35:42,480 --> 00:35:45,359
entropy. In information theory, entropy

804
00:35:45,359 --> 00:35:47,040
is the qualitative measurement of

805
00:35:47,040 --> 00:35:50,240
uncertainty or randomness in a data set.

806
00:35:50,240 --> 00:35:52,240
If a system is completely predictable,

807
00:35:52,240 --> 00:35:54,000
its entropy is zero. If [snorts] a

808
00:35:54,000 --> 00:35:56,160
system is completely random, its entropy

809
00:35:56,160 --> 00:35:58,560
is at its maximum. In the context of

810
00:35:58,560 --> 00:36:00,400
word orle, entropy represents the

811
00:36:00,400 --> 00:36:03,200
slicing power of a word. A high entropy

812
00:36:03,200 --> 00:36:05,839
guess is one that is guaranteed to split

813
00:36:05,839 --> 00:36:07,760
our remaining word pool into small

814
00:36:07,760 --> 00:36:10,000
manageable groups, no matter what the

815
00:36:10,000 --> 00:36:12,560
colors of the game returns. This is the

816
00:36:12,560 --> 00:36:17,079
idea that will drive our solver.

817
00:36:17,280 --> 00:36:19,040
Let's look at entropy and how it's

818
00:36:19,040 --> 00:36:21,200
calculated in our function. First, the

819
00:36:21,200 --> 00:36:23,359
word handles a boundary condition. If

820
00:36:23,359 --> 00:36:25,520
the pool of remaining words is empty,

821
00:36:25,520 --> 00:36:28,000
there is zero uncertainty left. So, it

822
00:36:28,000 --> 00:36:30,720
returns an entropy of zero. Here, the

823
00:36:30,720 --> 00:36:33,040
code runs a simulation. It takes a

824
00:36:33,040 --> 00:36:35,200
potential guess and simulates playing it

825
00:36:35,200 --> 00:36:37,200
against every single word left in our

826
00:36:37,200 --> 00:36:39,200
current pool. For every word, it

827
00:36:39,200 --> 00:36:41,040
calculates what the word oral feedback

828
00:36:41,040 --> 00:36:43,280
colors would be. It uses Python's

829
00:36:43,280 --> 00:36:45,280
counter to group the words into these

830
00:36:45,280 --> 00:36:47,280
color buckets, we'll call them, and

831
00:36:47,280 --> 00:36:50,160
counts them.

832
00:36:50,160 --> 00:36:52,079
This is the formula that calculates

833
00:36:52,079 --> 00:36:54,400
standard Shannon entropy. For every

834
00:36:54,400 --> 00:36:56,400
color bucket, the code calculates the

835
00:36:56,400 --> 00:36:59,040
probability P of hitting that bucket by

836
00:36:59,040 --> 00:37:01,359
dividing the count of words in that

837
00:37:01,359 --> 00:37:03,520
bucket by the total number of words in

838
00:37:03,520 --> 00:37:06,720
the P pool. If P is greater than zero,

839
00:37:06,720 --> 00:37:08,720
it multiplies that probability by the

840
00:37:08,720 --> 00:37:11,920
base 2 logarithm of its inverse. It adds

841
00:37:11,920 --> 00:37:14,240
these values together for all buckets to

842
00:37:14,240 --> 00:37:17,040
return the total expected entropy. If a

843
00:37:17,040 --> 00:37:19,760
word clumps all remaining words into one

844
00:37:19,760 --> 00:37:22,240
giant bucket, P is equal to 1, the

845
00:37:22,240 --> 00:37:24,320
answer to the logarithm is zero. So

846
00:37:24,320 --> 00:37:27,040
entropy is zero. If it distributes words

847
00:37:27,040 --> 00:37:30,960
evenly across many buckets is small, the

848
00:37:30,960 --> 00:37:33,280
answer to the logarithm is large and the

849
00:37:33,280 --> 00:37:36,160
resulting entropy is high. While entropy

850
00:37:36,160 --> 00:37:38,720
is our primary tool, our solver also

851
00:37:38,720 --> 00:37:41,440
uses three other functions that act as

852
00:37:41,440 --> 00:37:43,839
huristics to analyze the physical

853
00:37:43,839 --> 00:37:47,040
properties of human language.

854
00:37:47,040 --> 00:37:49,440
The first function evaluates positional

855
00:37:49,440 --> 00:37:52,160
topology. It uses a numerate word to

856
00:37:52,160 --> 00:37:54,480
loop through each letter position I from

857
00:37:54,480 --> 00:37:56,960
1 to four. It counts how many words in

858
00:37:56,960 --> 00:37:59,280
our remaining pool share that exact same

859
00:37:59,280 --> 00:38:01,680
letter at that exact same position. It

860
00:38:01,680 --> 00:38:03,440
divides this by the total number of

861
00:38:03,440 --> 00:38:05,599
remaining words to create a probability

862
00:38:05,599 --> 00:38:08,720
score. This ensures that the solver

863
00:38:08,720 --> 00:38:11,280
favors words containing letters placed

864
00:38:11,280 --> 00:38:15,119
in highly common positions.

865
00:38:15,119 --> 00:38:17,280
This function scores words based on

866
00:38:17,280 --> 00:38:20,240
their phonetic rhythm. It sets up a set

867
00:38:20,240 --> 00:38:22,560
of vowels and counts how many are in a

868
00:38:22,560 --> 00:38:25,359
word. The ideal amount of vowels to

869
00:38:25,359 --> 00:38:28,880
constants is 2 to three. The second

870
00:38:28,880 --> 00:38:31,359
measures informational variance.

871
00:38:31,359 --> 00:38:33,520
Converting the fiveletter word into a

872
00:38:33,520 --> 00:38:36,480
set removes duplicate letters. If all

873
00:38:36,480 --> 00:38:39,839
letters are unique, lenset word is five,

874
00:38:39,839 --> 00:38:42,079
yielding a perfect diversity score of

875
00:38:42,079 --> 00:38:45,280
one. A word like robot drops to 0.8

876
00:38:45,280 --> 00:38:47,760
because O is repeated. This pushes the

877
00:38:47,760 --> 00:38:50,720
solver to test five unique dimensions.

878
00:38:50,720 --> 00:38:53,440
On the very first turn where Len is

879
00:38:53,440 --> 00:38:56,960
zero, the code sort of shortcircuits the

880
00:38:56,960 --> 00:39:00,320
heavy math and immediately plays Tara's

881
00:39:00,320 --> 00:39:03,200
kind of sounds like my name, [laughter]

882
00:39:03,200 --> 00:39:05,200
which is a mathematically proven

883
00:39:05,200 --> 00:39:08,000
preomputed optimal opening word because

884
00:39:08,000 --> 00:39:10,320
it fulfills all these three checkpoints.

885
00:39:10,320 --> 00:39:12,400
One, it has two vowels and three

886
00:39:12,400 --> 00:39:14,720
consonants. Two, it has all different

887
00:39:14,720 --> 00:39:16,720
letters. And three, it has its letters

888
00:39:16,720 --> 00:39:19,359
at the most common places for them to

889
00:39:19,359 --> 00:39:22,880
be. Now, that was entropy and the

890
00:39:22,880 --> 00:39:26,079
calculation behind the solver. So, let's

891
00:39:26,079 --> 00:39:28,320
actually see if it works in actuality.

892
00:39:28,320 --> 00:39:31,782
Let's demo time.

893
00:39:31,782 --> 00:39:33,802
[applause]

894
00:39:45,280 --> 00:39:46,960
All right,

895
00:39:46,960 --> 00:39:48,240
while we're getting that out, who else

896
00:39:48,240 --> 00:39:52,599
has word first words?

897
00:39:55,920 --> 00:39:59,320
We have audio.

898
00:40:00,000 --> 00:40:03,240
It's you.

899
00:40:10,240 --> 00:40:14,680
Uh, has anyone ever

900
00:40:22,720 --> 00:40:25,720
seconded?

901
00:40:26,320 --> 00:40:27,680
Have you ever gotten a word on your

902
00:40:27,680 --> 00:40:28,320
first go?

903
00:40:28,320 --> 00:40:29,359
No.

904
00:40:29,359 --> 00:40:33,040
No. Y. [cheering]

905
00:40:33,040 --> 00:40:35,839
So, as you can probably tell, oh, as you

906
00:40:35,839 --> 00:40:38,960
can probably tell, the UI, like the user

907
00:40:38,960 --> 00:40:40,400
interface isn't the best. So, that's one

908
00:40:40,400 --> 00:40:41,359
of the things that I'm going to improve

909
00:40:41,359 --> 00:40:44,880
on. But, nonetheless, it still works.

910
00:40:44,880 --> 00:40:47,440
So, let's test it out. So, obviously,

911
00:40:47,440 --> 00:40:52,800
we're going to put in Tara into Wordle.

912
00:40:52,800 --> 00:40:53,119
Oh,

913
00:40:53,119 --> 00:40:54,400
give me give me a second.

914
00:40:54,400 --> 00:40:55,680
Thank you. Sorry.

915
00:40:55,680 --> 00:40:57,200
Here we go. You talk. I'll hold it.

916
00:40:57,200 --> 00:40:58,720
All right. So, we're going to put Tara

917
00:40:58,720 --> 00:41:02,800
into it. Let's do it. Okay. So, in my

918
00:41:02,800 --> 00:41:04,720
word, I forgot to mention this in the

919
00:41:04,720 --> 00:41:07,520
presentation, but the gray letter will

920
00:41:07,520 --> 00:41:10,000
be used as zero, yellow will be one, and

921
00:41:10,000 --> 00:41:14,000
green will be two. So, here two 0 0 two

922
00:41:14,000 --> 00:41:17,119
0. Let's put it in then.

923
00:41:17,119 --> 00:41:20,480
Two zero zero two zero. Thanks. I kind

924
00:41:20,480 --> 00:41:23,280
of forget sometimes. All right. Now, we

925
00:41:23,280 --> 00:41:25,680
have we're going to put in tweet. Can we

926
00:41:25,680 --> 00:41:29,028
see that? Cool.

927
00:41:29,028 --> 00:41:31,048
[laughter]

928
00:41:31,920 --> 00:41:34,400
Okay.

929
00:41:34,400 --> 00:41:36,800
Oh, so close.

930
00:41:36,800 --> 00:41:39,839
I was like, "Oh, oh, no."

931
00:41:39,839 --> 00:41:40,319
Okay.

932
00:41:40,319 --> 00:41:41,680
So close. [laughter]

933
00:41:41,680 --> 00:41:45,040
Okay. What was it? Two. Two. Two. Two.

934
00:41:45,040 --> 00:41:49,560
Zero. Let's see what it does.

935
00:41:50,480 --> 00:41:59,000
Tweed is I think that's our word. E E D

936
00:41:59,520 --> 00:42:03,640
Oh no.

937
00:42:04,240 --> 00:42:06,560
Okay.

938
00:42:06,560 --> 00:42:08,640
Two. Two. Two. Two.

939
00:42:08,640 --> 00:42:10,800
If you if you don't want your worldle

940
00:42:10,800 --> 00:42:11,839
spoiled today,

941
00:42:11,839 --> 00:42:15,280
look away now. [laughter]

942
00:42:15,280 --> 00:42:17,119
Oh no.

943
00:42:17,119 --> 00:42:19,359
The feedback seems to be inconsist

944
00:42:19,359 --> 00:42:22,079
continuing with current possibilities.

945
00:42:22,079 --> 00:42:23,839
That's never happened before. I think we

946
00:42:23,839 --> 00:42:25,520
need to update the dictionary. Let's try

947
00:42:25,520 --> 00:42:26,480
tween.

948
00:42:26,480 --> 00:42:30,040
Tween. All right.

949
00:42:35,839 --> 00:42:37,119
[screaming]

950
00:42:37,119 --> 00:42:40,800
Now, going back. [applause]

951
00:42:40,800 --> 00:42:43,040
Obviously, it needs some improvements.

952
00:42:43,040 --> 00:42:45,520
Tween isn't in the dictionary. So, as

953
00:42:45,520 --> 00:42:50,079
soon as I as soon as I get home, that's

954
00:42:50,079 --> 00:42:52,160
what I'm adding into the dictionary.

955
00:42:52,160 --> 00:42:53,920
Let's look at some of my future

956
00:42:53,920 --> 00:42:57,280
ambitions of this project. So, the first

957
00:42:57,280 --> 00:42:59,359
thing I want to do is obviously create a

958
00:42:59,359 --> 00:43:01,520
much better user interface. It's there's

959
00:43:01,520 --> 00:43:02,960
lots of words and I don't really like

960
00:43:02,960 --> 00:43:05,200
that that much. I want it to look nice.

961
00:43:05,200 --> 00:43:07,599
The next thing I want to do is expand my

962
00:43:07,599 --> 00:43:09,280
solver to different languages like

963
00:43:09,280 --> 00:43:12,640
Italian, Chinese, Hindi. And my

964
00:43:12,640 --> 00:43:14,319
short-term goal, however, is to create

965
00:43:14,319 --> 00:43:16,400
an Italian word or solver. But if there

966
00:43:16,400 --> 00:43:18,000
are any other languages that you guys

967
00:43:18,000 --> 00:43:20,800
want to see, please do tell me. But

968
00:43:20,800 --> 00:43:23,839
anyway, that's it from me. Bye. Thank

969
00:43:23,839 --> 00:43:25,619
you. [cheering]

970
00:43:25,619 --> 00:43:27,640
[applause]

971
00:43:30,480 --> 00:43:34,135
Massive round of applause for Tara.

972
00:43:34,135 --> 00:43:35,445
[cheering]

973
00:43:35,445 --> 00:43:37,040
[applause]

974
00:43:37,040 --> 00:43:38,640
Okay,

975
00:43:38,640 --> 00:43:44,615
I hope you all like words and also maths

976
00:43:44,615 --> 00:43:45,680
[cheering]

977
00:43:45,680 --> 00:43:48,640
because next up we have Isa who's going

978
00:43:48,640 --> 00:43:52,560
to talk about polomial long division.

979
00:43:52,560 --> 00:43:55,040
Now while she sets up I have it on good

980
00:43:55,040 --> 00:43:57,280
authority that it was her birthday

981
00:43:57,280 --> 00:44:00,240
earlier in the week and she was stuck in

982
00:44:00,240 --> 00:44:03,200
exams and then prep for pyon. So, can we

983
00:44:03,200 --> 00:44:04,480
all do the most embarrassing thing

984
00:44:04,480 --> 00:44:06,000
possible and sing her the first line of

985
00:44:06,000 --> 00:44:07,599
happy birthday ridiculously quickly?

986
00:44:07,599 --> 00:44:11,440
Ready? Happy birthday to you. Cut. Thank

987
00:44:11,440 --> 00:44:13,435
you.

988
00:44:13,435 --> 00:44:15,839
[screaming] Um, she has been programming

989
00:44:15,839 --> 00:44:17,920
for an unknown length of time depending

990
00:44:17,920 --> 00:44:19,920
on what our definition of programming

991
00:44:19,920 --> 00:44:22,319
is. And I never clarified

992
00:44:22,319 --> 00:44:25,440
so we'll never know.

993
00:44:25,440 --> 00:44:26,640
No, but how long?

994
00:44:26,640 --> 00:44:27,920
Uh, I don't know.

995
00:44:27,920 --> 00:44:30,960
Nobody knows. [laughter]

996
00:44:30,960 --> 00:44:33,760
He's saying two uh two two years. Two

997
00:44:33,760 --> 00:44:39,000
years. Uh her favorite emoji is

998
00:44:41,920 --> 00:44:44,720
swan. Swan. I was worried about that

999
00:44:44,720 --> 00:44:46,240
one. Um [laughter]

1000
00:44:46,240 --> 00:44:49,599
and recently she learned how CC cyclic

1001
00:44:49,599 --> 00:44:51,440
redundancy check can detect errors in

1002
00:44:51,440 --> 00:44:53,599
data transmission by using polomial

1003
00:44:53,599 --> 00:44:55,680
division to generate a check sum. It

1004
00:44:55,680 --> 00:44:58,079
helps ensure data hasn't been corrupted

1005
00:44:58,079 --> 00:45:00,560
during transfer.

1006
00:45:00,560 --> 00:45:02,560
One massive round of applause. Let's

1007
00:45:02,560 --> 00:45:05,181
hear all about it.

1008
00:45:05,181 --> 00:45:07,201
[applause]

1009
00:45:07,680 --> 00:45:10,079
Um hello everyone. My name is Isa and my

1010
00:45:10,079 --> 00:45:12,480
project is the CRC checker. So it uses

1011
00:45:12,480 --> 00:45:14,880
polomial long division. So just a bit

1012
00:45:14,880 --> 00:45:17,040
about CRC. It stands for the cyclic

1013
00:45:17,040 --> 00:45:19,280
redundancy check. Um and it checks to

1014
00:45:19,280 --> 00:45:21,200
see if data being sent and received on

1015
00:45:21,200 --> 00:45:23,040
two ends is the same data. So whether

1016
00:45:23,040 --> 00:45:24,560
it's been changed or altered in the

1017
00:45:24,560 --> 00:45:26,079
process, whether it's been corrupted or

1018
00:45:26,079 --> 00:45:27,760
not, it helps us make sure the

1019
00:45:27,760 --> 00:45:30,640
information is secure. Um, and so you

1020
00:45:30,640 --> 00:45:32,319
might be wondering how or like more

1021
00:45:32,319 --> 00:45:34,240
specifically why I decided to figure

1022
00:45:34,240 --> 00:45:36,319
this out or learn about it at all. So

1023
00:45:36,319 --> 00:45:38,960
CRC is very heavily maths based like in

1024
00:45:38,960 --> 00:45:40,400
the background is just lots and lots of

1025
00:45:40,400 --> 00:45:42,960
maths algorithm. And some of you

1026
00:45:42,960 --> 00:45:44,480
probably for that reason think it's

1027
00:45:44,480 --> 00:45:47,040
extremely boring. But let me explain.

1028
00:45:47,040 --> 00:45:49,280
You say yes. [laughter] Sorry. Let me

1029
00:45:49,280 --> 00:45:52,079
let me explain. Let me explain why I

1030
00:45:52,079 --> 00:45:54,480
care so much about it despite that. So,

1031
00:45:54,480 --> 00:45:56,640
just a bit of context about me. I'm in

1032
00:45:56,640 --> 00:45:58,800
year 11 and I go to Bergman Anglican

1033
00:45:58,800 --> 00:46:01,520
School located in the ACT and my skills

1034
00:46:01,520 --> 00:46:04,800
are maths.

1035
00:46:04,800 --> 00:46:06,960
Um, and given that my skills and passion

1036
00:46:06,960 --> 00:46:09,119
are in maths, um, I was accelerated to

1037
00:46:09,119 --> 00:46:12,480
year 11 spec when I was in year 10 and

1038
00:46:12,480 --> 00:46:14,640
our first exam was a polomial long

1039
00:46:14,640 --> 00:46:17,920
division exam. Um, however, I decided to

1040
00:46:17,920 --> 00:46:20,000
throw away a mark [snorts] because I

1041
00:46:20,000 --> 00:46:24,640
said that 180 - 75 - 60 is 55. And as

1042
00:46:24,640 --> 00:46:26,400
you can see, my teacher suggested that

1043
00:46:26,400 --> 00:46:29,680
it wasn't 55 and it was 45 instead. Um,

1044
00:46:29,680 --> 00:46:31,200
but what sucks about this is I actually

1045
00:46:31,200 --> 00:46:33,359
had two calculators in front of me, but

1046
00:46:33,359 --> 00:46:35,119
both my graphics and my scientific

1047
00:46:35,119 --> 00:46:38,192
calculator, and I decided not to use it.

1048
00:46:38,192 --> 00:46:40,960
[snorts] So, as you can see, and then a

1049
00:46:40,960 --> 00:46:42,960
couple months go by and my second exam

1050
00:46:42,960 --> 00:46:45,920
comes up. Um however once again polomial

1051
00:46:45,920 --> 00:46:48,960
long division was on our second exam but

1052
00:46:48,960 --> 00:46:51,947
I kind of did that on my exam.

1053
00:46:51,947 --> 00:46:52,880
[clears throat]

1054
00:46:52,880 --> 00:46:56,800
So as you can see 4 + 4 + 4 apparently

1055
00:46:56,800 --> 00:46:58,960
doesn't equal 16. I was told it equals

1056
00:46:58,960 --> 00:47:01,599
12. Um and what sucks more about this

1057
00:47:01,599 --> 00:47:04,000
exam is once again two calculators in

1058
00:47:04,000 --> 00:47:05,040
front of me right graphics and

1059
00:47:05,040 --> 00:47:07,200
scientific. But like I also had my

1060
00:47:07,200 --> 00:47:09,119
fingers like I could have counted and I

1061
00:47:09,119 --> 00:47:12,560
would have gotten it right. So like any

1062
00:47:12,560 --> 00:47:15,359
anyways moving on. So for the next 6

1063
00:47:15,359 --> 00:47:17,839
months my friends and my teacher Mr.

1064
00:47:17,839 --> 00:47:19,760
Griffin sitting in the front row there

1065
00:47:19,760 --> 00:47:21,280
kept making fun of me. They were like

1066
00:47:21,280 --> 00:47:23,119
whispering 12 in my ear every now and

1067
00:47:23,119 --> 00:47:26,640
then just reminding me of of what

1068
00:47:26,640 --> 00:47:29,200
happened. So like best friends by the

1069
00:47:29,200 --> 00:47:32,560
way. Um and so conveniently in computer

1070
00:47:32,560 --> 00:47:34,079
science we got to pick our own

1071
00:47:34,079 --> 00:47:36,720
algorithms for our project.

1072
00:47:36,720 --> 00:47:39,108
Guess what algorithm I picked?

1073
00:47:39,108 --> 00:47:39,359
[laughter and gasps]

1074
00:47:39,359 --> 00:47:41,599
I picked polomial long division, my

1075
00:47:41,599 --> 00:47:43,920
favorite algorithm by this point. And so

1076
00:47:43,920 --> 00:47:45,520
for the next three weeks, I continued to

1077
00:47:45,520 --> 00:47:47,760
hustle my IT teacher because I was like,

1078
00:47:47,760 --> 00:47:49,760
what the hell can polomial long division

1079
00:47:49,760 --> 00:47:51,680
do in computer science? Like my friends

1080
00:47:51,680 --> 00:47:53,040
picked these ones that like sounded

1081
00:47:53,040 --> 00:47:54,480
really complex, like they sounded like

1082
00:47:54,480 --> 00:47:56,240
they belonged in computer science, but

1083
00:47:56,240 --> 00:47:58,720
mine was like purely math space. And so

1084
00:47:58,720 --> 00:48:01,680
that's how I landed on CRC. And as a

1085
00:48:01,680 --> 00:48:03,680
direct consequence of the 4 plus 4 plus

1086
00:48:03,680 --> 00:48:06,240
4 incidents, here we are in Queensland 6

1087
00:48:06,240 --> 00:48:09,560
months later.

1088
00:48:10,000 --> 00:48:13,359
So I'll just go through the underlying

1089
00:48:13,359 --> 00:48:16,000
theory of CLC and how polomial long

1090
00:48:16,000 --> 00:48:17,680
division as a math algorithm can

1091
00:48:17,680 --> 00:48:19,920
actually be applied in computer science.

1092
00:48:19,920 --> 00:48:21,520
So when we're sending information we

1093
00:48:21,520 --> 00:48:23,599
have a sender and we have a receiver.

1094
00:48:23,599 --> 00:48:25,920
The sender message first gets converted

1095
00:48:25,920 --> 00:48:27,760
into a bitstream. So a bitstream is just

1096
00:48:27,760 --> 00:48:30,079
binary a bunch of zeros and ones. And so

1097
00:48:30,079 --> 00:48:31,920
every single digit is treated as the

1098
00:48:31,920 --> 00:48:34,160
coefficient of a degree in a polomial.

1099
00:48:34,160 --> 00:48:36,000
So once we've converted this bitstream

1100
00:48:36,000 --> 00:48:38,480
into a polomial, we divide it by what we

1101
00:48:38,480 --> 00:48:41,040
call a generator polomial. The generator

1102
00:48:41,040 --> 00:48:43,680
polomial is predefined. It's usually 32

1103
00:48:43,680 --> 00:48:45,599
bits. And what's key about this is it's

1104
00:48:45,599 --> 00:48:47,200
not a secret. It's not private. You can

1105
00:48:47,200 --> 00:48:49,200
search it up online. However, we divide

1106
00:48:49,200 --> 00:48:51,200
it by the generator polomial and we will

1107
00:48:51,200 --> 00:48:53,040
get a remainder. We append this

1108
00:48:53,040 --> 00:48:54,720
remainder to the end of our original

1109
00:48:54,720 --> 00:48:56,480
bitstream and we get what is called a

1110
00:48:56,480 --> 00:48:58,559
code word. So the sender sends the code

1111
00:48:58,559 --> 00:49:00,400
word to the receiver. Now on the

1112
00:49:00,400 --> 00:49:02,160
receiving end what happens is they take

1113
00:49:02,160 --> 00:49:03,599
this code word which is once again a

1114
00:49:03,599 --> 00:49:05,280
bunch of zeros and ones and once again

1115
00:49:05,280 --> 00:49:06,559
the digits are treated as the

1116
00:49:06,559 --> 00:49:09,040
coefficients of a polomial. We divide it

1117
00:49:09,040 --> 00:49:11,119
by the same generated polomial that was

1118
00:49:11,119 --> 00:49:13,200
used by the sender and we should get a

1119
00:49:13,200 --> 00:49:15,599
remainder of zero. So that's what should

1120
00:49:15,599 --> 00:49:17,440
happen theoretically. But let's say in

1121
00:49:17,440 --> 00:49:19,359
the middle a corruptor comes in a packet

1122
00:49:19,359 --> 00:49:21,440
gets lost or for whatever reason a bit

1123
00:49:21,440 --> 00:49:23,280
has been switched. In this case the

1124
00:49:23,280 --> 00:49:24,960
message is no longer the same message

1125
00:49:24,960 --> 00:49:26,720
and the integrity of the data has been

1126
00:49:26,720 --> 00:49:28,960
breached. However, this is okay because

1127
00:49:28,960 --> 00:49:30,800
when the receiver gets this message,

1128
00:49:30,800 --> 00:49:32,240
what's going to happen is they will

1129
00:49:32,240 --> 00:49:34,160
perform the same steps of division and

1130
00:49:34,160 --> 00:49:35,839
they will get a remainder that is not

1131
00:49:35,839 --> 00:49:37,599
zero. So that's how we can use the

1132
00:49:37,599 --> 00:49:39,119
remainder to determine whether or not

1133
00:49:39,119 --> 00:49:41,920
the message has been corrupted. My next

1134
00:49:41,920 --> 00:49:43,680
question was what are the likelihoods of

1135
00:49:43,680 --> 00:49:46,000
two different polomials being divided by

1136
00:49:46,000 --> 00:49:48,319
the generator polomial and getting the

1137
00:49:48,319 --> 00:49:51,200
same remainder. That is about 1 in 2 to

1138
00:49:51,200 --> 00:49:53,359
the^ of 32. And if you're good at math,

1139
00:49:53,359 --> 00:49:55,440
you'll know that's 4.3 billion. So it's

1140
00:49:55,440 --> 00:49:56,880
not very likely that that will happen

1141
00:49:56,880 --> 00:49:59,440
which proves to be very useful.

1142
00:49:59,440 --> 00:50:01,440
So that's the theory behind CRC and how

1143
00:50:01,440 --> 00:50:02,880
we can use the remainder to determine

1144
00:50:02,880 --> 00:50:04,319
whether or not the message is still

1145
00:50:04,319 --> 00:50:07,040
safe. Now I'll do a quick demo but what

1146
00:50:07,040 --> 00:50:09,119
I will have is I will have two senders a

1147
00:50:09,119 --> 00:50:11,359
receiver and a corruptor uh just to show

1148
00:50:11,359 --> 00:50:13,359
the process. One of the one of the

1149
00:50:13,359 --> 00:50:14,880
senders will send a message that does

1150
00:50:14,880 --> 00:50:16,400
get corrupted and one of the senders

1151
00:50:16,400 --> 00:50:18,160
will send a message that is untouched.

1152
00:50:18,160 --> 00:50:19,760
The receiver will be able to determine

1153
00:50:19,760 --> 00:50:21,280
whether the message is corrupted or not.

1154
00:50:21,280 --> 00:50:25,720
And I set this all up using sockets.

1155
00:50:26,079 --> 00:50:29,280
So here, [laughter] sorry I didn't make

1156
00:50:29,280 --> 00:50:31,760
a pretty, but um in the top row I have

1157
00:50:31,760 --> 00:50:33,680
two senders. One I'll just show you

1158
00:50:33,680 --> 00:50:36,000
later. Um in the bottom left corner I

1159
00:50:36,000 --> 00:50:38,160
have I have the receiver and in the

1160
00:50:38,160 --> 00:50:39,200
bottom right corner I have the

1161
00:50:39,200 --> 00:50:41,359
corruptor. So if I were to send a

1162
00:50:41,359 --> 00:50:43,599
message from the top right window, let's

1163
00:50:43,599 --> 00:50:46,559
say I decide to send hello.

1164
00:50:46,559 --> 00:50:48,559
Hi. Um it's [clears throat] green. Um,

1165
00:50:48,559 --> 00:50:50,240
and I've I've made it green so that it's

1166
00:50:50,240 --> 00:50:51,920
clear to show it's valid. Um, the

1167
00:50:51,920 --> 00:50:53,440
receiver has detected that the remainder

1168
00:50:53,440 --> 00:50:55,760
is zero. Everything is perfect. However,

1169
00:50:55,760 --> 00:50:58,400
if I were to decide to send hello from

1170
00:50:58,400 --> 00:51:00,800
this terminal, it is red. So, it has

1171
00:51:00,800 --> 00:51:02,319
been corrupted. Um, it goes through the

1172
00:51:02,319 --> 00:51:04,319
corruptor in the bottom right where it's

1173
00:51:04,319 --> 00:51:06,640
taken that um list of ones and zeros.

1174
00:51:06,640 --> 00:51:08,240
It's switched something and the receiver

1175
00:51:08,240 --> 00:51:10,079
has determined that. So, as you can see,

1176
00:51:10,079 --> 00:51:12,079
the second L is no longer an L. It's a

1177
00:51:12,079 --> 00:51:14,640
comma. I have two corruptors. I have a

1178
00:51:14,640 --> 00:51:16,559
one bit corruptor and a flip corruptor.

1179
00:51:16,559 --> 00:51:18,400
If the message is five characters or

1180
00:51:18,400 --> 00:51:20,480
lower, it is a one bit corruptor where a

1181
00:51:20,480 --> 00:51:22,319
random bit, it's completely random, will

1182
00:51:22,319 --> 00:51:24,079
get switched. Um, and this replicates

1183
00:51:24,079 --> 00:51:26,480
those uh tiny packet losses that may not

1184
00:51:26,480 --> 00:51:28,240
seem significant. Sometimes it's not

1185
00:51:28,240 --> 00:51:30,000
visible to humans and the message might

1186
00:51:30,000 --> 00:51:31,440
look the same. But what is really

1187
00:51:31,440 --> 00:51:33,040
important is that the computer can still

1188
00:51:33,040 --> 00:51:34,960
determine that the message is changed,

1189
00:51:34,960 --> 00:51:36,640
is altered, and it is not the same that

1190
00:51:36,640 --> 00:51:38,800
was being sent. My second type of

1191
00:51:38,800 --> 00:51:40,559
corruptor is a flip corruptor. If the

1192
00:51:40,559 --> 00:51:42,319
message exceeds five characters, every

1193
00:51:42,319 --> 00:51:44,079
single one and zero is flipped to the

1194
00:51:44,079 --> 00:51:46,640
opposite. This demonstrates the more

1195
00:51:46,640 --> 00:51:48,559
extreme like data manipulation when

1196
00:51:48,559 --> 00:51:50,160
everything gets corrupted or malicious

1197
00:51:50,160 --> 00:51:53,280
attempts in real life. If I were to send

1198
00:51:53,280 --> 00:51:55,760
hello again and again, you can see every

1199
00:51:55,760 --> 00:51:58,880
single time the corruption is different.

1200
00:51:58,880 --> 00:52:00,559
In the second time, the O is changed to

1201
00:52:00,559 --> 00:52:03,280
an M. And in the final time, we can't

1202
00:52:03,280 --> 00:52:05,200
see a difference. However, the receiver

1203
00:52:05,200 --> 00:52:06,559
has still been able to determine that

1204
00:52:06,559 --> 00:52:08,319
the message has been changed. That is

1205
00:52:08,319 --> 00:52:09,760
actually really lucky cuz um it's

1206
00:52:09,760 --> 00:52:11,440
random. And that's a perfect example of

1207
00:52:11,440 --> 00:52:13,040
when we can't see the change in the

1208
00:52:13,040 --> 00:52:14,640
message but it has been tampered with

1209
00:52:14,640 --> 00:52:16,240
and the integrity of the data has been

1210
00:52:16,240 --> 00:52:18,559
breached.

1211
00:52:18,559 --> 00:52:20,400
So this demo is just like really

1212
00:52:20,400 --> 00:52:22,960
important to show that although um

1213
00:52:22,960 --> 00:52:24,720
polomial long division doesn't seem

1214
00:52:24,720 --> 00:52:26,240
significant. It seems like a really

1215
00:52:26,240 --> 00:52:28,480
basic like maths thing but we use it in

1216
00:52:28,480 --> 00:52:30,000
our everyday lives. every time you send

1217
00:52:30,000 --> 00:52:31,280
a message online, those message

1218
00:52:31,280 --> 00:52:33,040
transmissions online and all of that

1219
00:52:33,040 --> 00:52:35,440
that it uses CRC to determine whether or

1220
00:52:35,440 --> 00:52:36,960
not that what you're receiving is

1221
00:52:36,960 --> 00:52:38,400
actually what's being sent and that the

1222
00:52:38,400 --> 00:52:42,119
message hasn't been corrupted.

1223
00:52:44,000 --> 00:52:45,200
Thank you for listening to my

1224
00:52:45,200 --> 00:52:47,616
presentation.

1225
00:52:47,616 --> 00:52:49,636
[applause and cheering]

1226
00:52:50,880 --> 00:52:53,520
That was amazing. [cheering]

1227
00:52:53,520 --> 00:52:55,520
Woo!

1228
00:52:55,520 --> 00:52:58,079
Massive round of applause. Next up, we

1229
00:52:58,079 --> 00:52:59,839
have Heidi.

1230
00:52:59,839 --> 00:53:04,480
And Heidi has been programming uh since

1231
00:53:04,480 --> 00:53:06,880
she was introduced to Scratch Junior in

1232
00:53:06,880 --> 00:53:10,200
year two.

1233
00:53:10,720 --> 00:53:13,760
Uh and she's going to talk about Oh

1234
00:53:13,760 --> 00:53:16,559
gosh. And I've forgotten Fabula and

1235
00:53:16,559 --> 00:53:19,920
Shette. Shette. Okay. I'm gonna find out

1236
00:53:19,920 --> 00:53:22,559
what that means soon. Uh her favorite

1237
00:53:22,559 --> 00:53:25,440
emoji is

1238
00:53:25,440 --> 00:53:28,160
salute. salute. Uh, and she recently

1239
00:53:28,160 --> 00:53:30,559
discovered that Cobalt,

1240
00:53:30,559 --> 00:53:32,400
that's it. No, that Cobalt, a

1241
00:53:32,400 --> 00:53:34,240
programming language developed nearly 70

1242
00:53:34,240 --> 00:53:38,160
years ago in 1959, is still being used

1243
00:53:38,160 --> 00:53:40,960
to manage 70 to 80% of business

1244
00:53:40,960 --> 00:53:43,200
transactions. That that keeps us

1245
00:53:43,200 --> 00:53:46,480
confident. And the average age of a

1246
00:53:46,480 --> 00:53:49,280
Cobalt programmer is around 55 years

1247
00:53:49,280 --> 00:53:51,200
old.

1248
00:53:51,200 --> 00:53:53,680
There's hope for us yet. Massive round

1249
00:53:53,680 --> 00:53:55,737
of applause.

1250
00:53:55,737 --> 00:53:57,047
[cheering]

1251
00:53:57,047 --> 00:53:59,067
[applause]

1252
00:53:59,200 --> 00:54:01,760
Um hello everyone. Have you as you have

1253
00:54:01,760 --> 00:54:04,880
heard um my name is Heidi and to

1254
00:54:04,880 --> 00:54:06,640
introduce myself I thought I'd share

1255
00:54:06,640 --> 00:54:08,079
three things I think you should know

1256
00:54:08,079 --> 00:54:10,240
about me. Uh the first one is that I

1257
00:54:10,240 --> 00:54:13,200
love Tiramisu.

1258
00:54:13,200 --> 00:54:16,000
The second one is that I like things to

1259
00:54:16,000 --> 00:54:18,559
go according to plan.

1260
00:54:18,559 --> 00:54:21,280
And number three is that I don't like to

1261
00:54:21,280 --> 00:54:23,760
make plans.

1262
00:54:23,760 --> 00:54:25,520
Now, that's enough about me. Let me

1263
00:54:25,520 --> 00:54:27,040
introduce you to the real star of the

1264
00:54:27,040 --> 00:54:30,480
show. This here is Tuti. Um, it's a web

1265
00:54:30,480 --> 00:54:32,000
app that I made that's sort of a mix

1266
00:54:32,000 --> 00:54:35,359
between Google Classroom and Flora. And

1267
00:54:35,359 --> 00:54:37,680
basically, teachers can create a

1268
00:54:37,680 --> 00:54:39,200
classroom where they can add their

1269
00:54:39,200 --> 00:54:41,280
students, set them tasks, and then

1270
00:54:41,280 --> 00:54:43,599
monitor their project on tasks. And then

1271
00:54:43,599 --> 00:54:45,440
students can join their teachers class

1272
00:54:45,440 --> 00:54:49,280
to do these um tasks and um complete

1273
00:54:49,280 --> 00:54:52,079
study sessions. So, um, as you can see

1274
00:54:52,079 --> 00:54:53,920
in the big screen, that's like the

1275
00:54:53,920 --> 00:54:55,599
student view, and there's like a list of

1276
00:54:55,599 --> 00:54:58,000
tasks to complete and a little calendar

1277
00:54:58,000 --> 00:54:59,920
that shows you when your tasks are due.

1278
00:54:59,920 --> 00:55:03,040
Um, in the top corner up there, um,

1279
00:55:03,040 --> 00:55:04,800
that's the student, uh, the teacher view

1280
00:55:04,800 --> 00:55:06,480
where you can see all your students. Um,

1281
00:55:06,480 --> 00:55:08,800
and below that, you can add their tasks.

1282
00:55:08,800 --> 00:55:11,359
And in the bottom corner, that's um, I

1283
00:55:11,359 --> 00:55:13,760
used a Google Charts API to like give

1284
00:55:13,760 --> 00:55:15,680
you some nice statistics about how your

1285
00:55:15,680 --> 00:55:18,960
students are going. And um in the middle

1286
00:55:18,960 --> 00:55:21,599
there, there's the timer function that

1287
00:55:21,599 --> 00:55:24,880
lets you do um study sessions. Now, the

1288
00:55:24,880 --> 00:55:27,359
origins of this project here um were

1289
00:55:27,359 --> 00:55:29,200
late November last year where our

1290
00:55:29,200 --> 00:55:30,480
teacher gave us our assessment

1291
00:55:30,480 --> 00:55:33,040
notification. Um and I'd like to draw

1292
00:55:33,040 --> 00:55:35,520
your attention to a word in it. It's

1293
00:55:35,520 --> 00:55:37,359
agile.

1294
00:55:37,359 --> 00:55:39,760
Yeah. Now when I learned about agile

1295
00:55:39,760 --> 00:55:41,359
along with waterfall and whatever

1296
00:55:41,359 --> 00:55:43,520
quantum state waggile decides to be in

1297
00:55:43,520 --> 00:55:45,778
any given day

1298
00:55:45,778 --> 00:55:46,640
[laughter]

1299
00:55:46,640 --> 00:55:48,480
um I just decided that this was a

1300
00:55:48,480 --> 00:55:50,559
syllabus prescribed over complication of

1301
00:55:50,559 --> 00:55:52,240
me just getting in there and getting it

1302
00:55:52,240 --> 00:55:54,880
done. But as always hindsight has done

1303
00:55:54,880 --> 00:55:58,079
me dirty and what I wanted to share with

1304
00:55:58,079 --> 00:56:00,000
you today is what I learned about the

1305
00:56:00,000 --> 00:56:02,720
importance of making a plan and um

1306
00:56:02,720 --> 00:56:05,839
agility I suppose. Um, but this isn't

1307
00:56:05,839 --> 00:56:07,280
going to be best illustrated by a

1308
00:56:07,280 --> 00:56:08,799
finished product. So, I'm going to take

1309
00:56:08,799 --> 00:56:11,200
you back to the beginning. Fabula and

1310
00:56:11,200 --> 00:56:13,839
Shuette. Now, I don't expect you to know

1311
00:56:13,839 --> 00:56:16,000
what these two words mean. Uh, they

1312
00:56:16,000 --> 00:56:18,240
originate in a literary period called

1313
00:56:18,240 --> 00:56:20,640
Russian formalism and are used to

1314
00:56:20,640 --> 00:56:22,720
describe the functions of a narrative.

1315
00:56:22,720 --> 00:56:25,280
Fabula is in as much words the plot of

1316
00:56:25,280 --> 00:56:27,359
the story. It's the linear progression

1317
00:56:27,359 --> 00:56:30,720
of events as they progress. Um, and

1318
00:56:30,720 --> 00:56:32,799
Shuette on the other hand is the story

1319
00:56:32,799 --> 00:56:35,839
as it is presented to the reader. Now,

1320
00:56:35,839 --> 00:56:37,920
you may be wondering, Heidi, what does

1321
00:56:37,920 --> 00:56:40,720
this have to do with programming? And I

1322
00:56:40,720 --> 00:56:42,559
think the best way to illustrate this is

1323
00:56:42,559 --> 00:56:44,799
with a story of my own. So, let's start

1324
00:56:44,799 --> 00:56:47,599
this one in media res. Uh, we're two

1325
00:56:47,599 --> 00:56:49,280
days out from submission and the

1326
00:56:49,280 --> 00:56:51,359
antagonist of our tale is rearing its

1327
00:56:51,359 --> 00:56:54,720
ugly head. It's this guy. And the reason

1328
00:56:54,720 --> 00:56:57,440
it's causing me so much grief is because

1329
00:56:57,440 --> 00:57:01,760
2 days out it looks like this.

1330
00:57:01,760 --> 00:57:04,640
Now the reason 2 days out I don't have a

1331
00:57:04,640 --> 00:57:06,880
calendar module is because something you

1332
00:57:06,880 --> 00:57:08,960
might have experienced before is I told

1333
00:57:08,960 --> 00:57:11,119
myself that it was going to be easy. So

1334
00:57:11,119 --> 00:57:14,240
I could just leave it till the end. But

1335
00:57:14,240 --> 00:57:16,880
I know that was pretty bad of me. But I

1336
00:57:16,880 --> 00:57:19,040
decided to come up with a little plan to

1337
00:57:19,040 --> 00:57:21,359
solve it. Step one, if there's ever a

1338
00:57:21,359 --> 00:57:24,739
complex but [laughter]

1339
00:57:25,119 --> 00:57:27,280
If there's ever a complex but ultimately

1340
00:57:27,280 --> 00:57:29,040
mundane unique thing you need to do in

1341
00:57:29,040 --> 00:57:31,359
Python, there's probably a library for

1342
00:57:31,359 --> 00:57:33,760
it. Step two was to just, you know,

1343
00:57:33,760 --> 00:57:36,480
package it up, sprinkle on some CSS, and

1344
00:57:36,480 --> 00:57:38,400
straight to step three where it'd be

1345
00:57:38,400 --> 00:57:40,559
finished. Now, this wouldn't be a good

1346
00:57:40,559 --> 00:57:43,280
story if it didn't have a complication.

1347
00:57:43,280 --> 00:57:45,920
Instead of packaging it up nicely, I

1348
00:57:45,920 --> 00:57:49,119
ended up sending 15 different lists

1349
00:57:49,119 --> 00:57:52,960
through to Ginger. And um that I know

1350
00:57:52,960 --> 00:57:54,640
there isn't any limit on how many you

1351
00:57:54,640 --> 00:57:56,480
can send through, but I think that's at

1352
00:57:56,480 --> 00:58:00,480
least 10 too many. On top of this, I

1353
00:58:00,480 --> 00:58:02,880
didn't realize it until recently, but I

1354
00:58:02,880 --> 00:58:05,040
was using two different kind of dates in

1355
00:58:05,040 --> 00:58:08,079
my SQL tables. [laughter]

1356
00:58:08,079 --> 00:58:11,119
And on top of this, I was moving between

1357
00:58:11,119 --> 00:58:13,040
two different datetime objects in my

1358
00:58:13,040 --> 00:58:15,680
Python, which caused me two days of

1359
00:58:15,680 --> 00:58:18,160
syntax error and non-type objects. And

1360
00:58:18,160 --> 00:58:21,359
perhaps my worst sin of them all is that

1361
00:58:21,359 --> 00:58:23,920
in an effort not to format but another

1362
00:58:23,920 --> 00:58:26,640
button on my HTML, I decided to move

1363
00:58:26,640 --> 00:58:28,559
between the different months in the

1364
00:58:28,559 --> 00:58:31,760
calendar with URL links. So that means I

1365
00:58:31,760 --> 00:58:34,000
had to send um the month through the

1366
00:58:34,000 --> 00:58:36,079
URL. And if my software teacher has

1367
00:58:36,079 --> 00:58:37,839
taught me anything is that you never

1368
00:58:37,839 --> 00:58:40,640
send anything through their URL. That's

1369
00:58:40,640 --> 00:58:42,630
a security breach.

1370
00:58:42,630 --> 00:58:43,200
[laughter]

1371
00:58:43,200 --> 00:58:45,359
Now,

1372
00:58:45,359 --> 00:58:47,520
this leaves you with no kind resolution

1373
00:58:47,520 --> 00:58:49,760
and actually an ultimately unanswered

1374
00:58:49,760 --> 00:58:51,680
question. What does this have to do with

1375
00:58:51,680 --> 00:58:54,240
Fabula and Shette? And I ask for your

1376
00:58:54,240 --> 00:58:56,720
patience as I digress, but a little bit

1377
00:58:56,720 --> 00:58:59,280
more. Now, one of the key features of

1378
00:58:59,280 --> 00:59:01,119
Russian formalism is that they chose to

1379
00:59:01,119 --> 00:59:03,760
approach poetics as a science rather

1380
00:59:03,760 --> 00:59:06,079
than an art. And this meant that they

1381
00:59:06,079 --> 00:59:07,680
were much more interested in how

1382
00:59:07,680 --> 00:59:09,920
literature achieved things rather than

1383
00:59:09,920 --> 00:59:12,160
what exactly it tried to achieve. think

1384
00:59:12,160 --> 00:59:15,599
like metaphors and the like. And I

1385
00:59:15,599 --> 00:59:17,680
suppose this is all a very roundabout ma

1386
00:59:17,680 --> 00:59:19,680
way of me saying that in and amongst

1387
00:59:19,680 --> 00:59:21,680
this confusion, what I learned is that

1388
00:59:21,680 --> 00:59:24,079
it's pro perhaps equally as important as

1389
00:59:24,079 --> 00:59:26,240
how your code achieves something as what

1390
00:59:26,240 --> 00:59:28,880
it achieves. And by extension of this,

1391
00:59:28,880 --> 00:59:31,280
it's both true that your first plan is

1392
00:59:31,280 --> 00:59:33,119
your most important and probably going

1393
00:59:33,119 --> 00:59:36,160
to be your worst. And yet the question

1394
00:59:36,160 --> 00:59:38,240
remains and I thank you for your

1395
00:59:38,240 --> 00:59:39,839
forbearance at the increasingly

1396
00:59:39,839 --> 00:59:42,880
nonlinear way I've gotten here. What

1397
00:59:42,880 --> 00:59:45,440
does this have to do with fabulous yet?

1398
00:59:45,440 --> 00:59:48,079
Well, I guess it can also be said that

1399
00:59:48,079 --> 00:59:50,240
no matter how way how well you plot out

1400
00:59:50,240 --> 00:59:52,319
your plan, how linear the sequence of

1401
00:59:52,319 --> 00:59:54,319
events, the story as it's going to be

1402
00:59:54,319 --> 00:59:57,359
presented to you can be very different.

1403
00:59:57,359 --> 00:59:59,359
But thankfully, we're using an agile

1404
00:59:59,359 --> 01:00:02,400
approach here. So this

1405
01:00:02,400 --> 01:00:05,359
this is not the end. Ultimately, 2 days

1406
01:00:05,359 --> 01:00:07,599
out, there was very little I could do.

1407
01:00:07,599 --> 01:00:10,079
But one of the benefits of hindsight is

1408
01:00:10,079 --> 01:00:12,319
that now I know what to do next time,

1409
01:00:12,319 --> 01:00:14,000
even if it's stuff my software teacher

1410
01:00:14,000 --> 01:00:16,296
told me to do the first time around.

1411
01:00:16,296 --> 01:00:17,760
[laughter]

1412
01:00:17,760 --> 01:00:21,359
Number one, plan your data. A data

1413
01:00:21,359 --> 01:00:22,960
dictionary is one of the things that my

1414
01:00:22,960 --> 01:00:24,960
teacher told me to do before the

1415
01:00:24,960 --> 01:00:26,559
assessment, which I ultimately did

1416
01:00:26,559 --> 01:00:29,520
afterwards. But if I had done it, I

1417
01:00:29,520 --> 01:00:32,000
would have known that in SQL, the way

1418
01:00:32,000 --> 01:00:34,160
they format dates is much different to

1419
01:00:34,160 --> 01:00:37,040
the way Python does. Number two,

1420
01:00:37,040 --> 01:00:39,760
security always comes first. It might be

1421
01:00:39,760 --> 01:00:41,280
easier just to send all your data

1422
01:00:41,280 --> 01:00:43,680
through the URL, but that's not really

1423
01:00:43,680 --> 01:00:46,480
good for your client. So, how about you

1424
01:00:46,480 --> 01:00:50,960
use post with CSRF tokens, escaping, and

1425
01:00:50,960 --> 01:00:53,520
validation, which I did in the rest of

1426
01:00:53,520 --> 01:00:55,280
my project. I just didn't want to do it

1427
01:00:55,280 --> 01:00:58,160
here. And perhaps third and the most

1428
01:00:58,160 --> 01:01:00,000
important is please listen to your

1429
01:01:00,000 --> 01:01:03,480
software teacher.

1430
01:01:03,839 --> 01:01:06,240
But um thankfully there is no reason to

1431
01:01:06,240 --> 01:01:08,160
fear because as always there will be

1432
01:01:08,160 --> 01:01:11,119
time to do it all over again. That's my

1433
01:01:11,119 --> 01:01:13,119
name. Um that's my website if you want

1434
01:01:13,119 --> 01:01:17,212
to look at it. Um thanks.

1435
01:01:17,212 --> 01:01:19,177
[cheering]

1436
01:01:19,177 --> 01:01:20,319
[applause]

1437
01:01:20,319 --> 01:01:22,319
Massive round of applause. Those are

1438
01:01:22,319 --> 01:01:26,382
some things that I wish I learned.

1439
01:01:26,382 --> 01:01:26,880
[applause]

1440
01:01:26,880 --> 01:01:30,559
actually listening to teachers. Oh, what

1441
01:01:30,559 --> 01:01:33,587
the kids think of next?

1442
01:01:33,587 --> 01:01:34,160
[laughter]

1443
01:01:34,160 --> 01:01:37,760
I see the teachers grimacing. Um,

1444
01:01:37,760 --> 01:01:40,480
next up on deck, we have the fantabulous

1445
01:01:40,480 --> 01:01:46,359
Emma. We only have two more talks left.

1446
01:01:46,880 --> 01:01:50,000
I know. Which kind of relates to Emma's

1447
01:01:50,000 --> 01:01:54,760
favorite emoji, which is

1448
01:01:55,119 --> 01:01:59,599
crying sobbing emoji. Uh Emma has been

1449
01:01:59,599 --> 01:02:03,440
programming from for six years

1450
01:02:03,440 --> 01:02:04,960
and

1451
01:02:04,960 --> 01:02:07,520
uh she told me or at least she wrote it

1452
01:02:07,520 --> 01:02:09,599
down in the form that it took a month

1453
01:02:09,599 --> 01:02:13,200
for her to figure out.

1454
01:02:13,200 --> 01:02:15,920
Hang on. [laughter]

1455
01:02:15,920 --> 01:02:17,599
Uh okay, I'm just going to read it for

1456
01:02:17,599 --> 01:02:19,920
about him. It took a month for me to

1457
01:02:19,920 --> 01:02:21,440
figure out. I really should read this

1458
01:02:21,440 --> 01:02:23,760
earlier. Uh but a monad brackets in

1459
01:02:23,760 --> 01:02:26,240
functional programming is the decoration

1460
01:02:26,240 --> 01:02:28,000
on a regular type which allows

1461
01:02:28,000 --> 01:02:30,160
computational effects like errors and

1462
01:02:30,160 --> 01:02:33,040
exceptions to be handled gracefully.

1463
01:02:33,040 --> 01:02:34,503
Almost as gracefully as I handled

1464
01:02:34,503 --> 01:02:36,468
[laughter] that.

1465
01:02:36,468 --> 01:02:38,000
[gasps] Massive round of applause for

1466
01:02:38,000 --> 01:02:39,743
Emma.

1467
01:02:39,743 --> 01:02:41,053
[cheering]

1468
01:02:41,053 --> 01:02:43,073
[applause]

1469
01:02:43,200 --> 01:02:45,920
Hey everyone. Today I'm going to discuss

1470
01:02:45,920 --> 01:02:49,440
how I made this socket app using Python

1471
01:02:49,440 --> 01:02:51,920
that I secured using a bunch of really

1472
01:02:51,920 --> 01:02:54,480
cool Python cryptography libraries as

1473
01:02:54,480 --> 01:02:57,599
well as a P2P gossip protocol over the

1474
01:02:57,599 --> 01:03:01,680
LAN. I use Arch by the way.

1475
01:03:01,680 --> 01:03:03,839
And before I get into my app, I'm just

1476
01:03:03,839 --> 01:03:08,640
going to unpack a couple network basics.

1477
01:03:08,640 --> 01:03:12,480
I hope you guys get the joke.

1478
01:03:12,480 --> 01:03:14,720
So, some network basics. First of all,

1479
01:03:14,720 --> 01:03:17,359
what is a LAN? The LAN stands for the

1480
01:03:17,359 --> 01:03:19,839
local area network, which is basically

1481
01:03:19,839 --> 01:03:22,000
your internet. That's where all of the

1482
01:03:22,000 --> 01:03:24,640
computers in a local area can talk to

1483
01:03:24,640 --> 01:03:26,480
each other. This is not the full

1484
01:03:26,480 --> 01:03:28,640
internet. The full internet is where you

1485
01:03:28,640 --> 01:03:31,039
can talk to like a computer in Europe,

1486
01:03:31,039 --> 01:03:33,760
but you're too far away from Europe. So,

1487
01:03:33,760 --> 01:03:37,039
that is the W, the WAN, the wide area

1488
01:03:37,039 --> 01:03:39,440
network. However, for now, we are just

1489
01:03:39,440 --> 01:03:41,440
sticking to LAN because one, it's

1490
01:03:41,440 --> 01:03:43,440
easier. Two, I don't need to deal with

1491
01:03:43,440 --> 01:03:45,359
NAT and the firewall and all the ports

1492
01:03:45,359 --> 01:03:49,640
shenanigans. And what

1493
01:03:50,484 --> 01:03:52,504
[laughter]

1494
01:03:53,039 --> 01:03:54,799
Sorry, that was Microsoft. That was

1495
01:03:54,799 --> 01:03:59,400
Microsoft. We can all blame Microsoft.

1496
01:03:59,654 --> 01:04:01,674
[cheering]

1497
01:04:01,920 --> 01:04:04,559
Anyways, back to the LAN. So the LAN is

1498
01:04:04,559 --> 01:04:07,359
also much faster because uh everything's

1499
01:04:07,359 --> 01:04:09,200
closer together which is why I decided

1500
01:04:09,200 --> 01:04:13,200
to use it for my project.

1501
01:04:13,200 --> 01:04:15,839
And this is the OSI model which you

1502
01:04:15,839 --> 01:04:18,480
might recognize. Uh for my project I'm

1503
01:04:18,480 --> 01:04:20,160
going to be using the Python socket

1504
01:04:20,160 --> 01:04:22,559
library which handles everything here.

1505
01:04:22,559 --> 01:04:24,319
So I don't have to worry about all of

1506
01:04:24,319 --> 01:04:26,480
that. I only have to worry about the

1507
01:04:26,480 --> 01:04:28,960
application level which I'm focusing on

1508
01:04:28,960 --> 01:04:31,119
the session presentation and application

1509
01:04:31,119 --> 01:04:33,280
levels.

1510
01:04:33,280 --> 01:04:36,160
Next up is TCP sockets. TCP is much

1511
01:04:36,160 --> 01:04:38,559
better than UDP if you care about data

1512
01:04:38,559 --> 01:04:40,720
reliability. That's because we have a

1513
01:04:40,720 --> 01:04:43,359
bunch of handshakes. When you want to

1514
01:04:43,359 --> 01:04:46,799
connect to another socket using TCP, you

1515
01:04:46,799 --> 01:04:49,359
have to go through a three uh three-way

1516
01:04:49,359 --> 01:04:51,680
handshake, unlike with UDP where you

1517
01:04:51,680 --> 01:04:53,440
just throw the data out and hope they

1518
01:04:53,440 --> 01:04:55,440
get it, which usually doesn't happen

1519
01:04:55,440 --> 01:04:57,119
unless you're torrenting or something

1520
01:04:57,119 --> 01:05:00,624
cool like that. [laughter]

1521
01:05:02,480 --> 01:05:06,079
Anyways, before I get into my own app, I

1522
01:05:06,079 --> 01:05:07,760
just want to acknowledge a bunch of

1523
01:05:07,760 --> 01:05:10,079
inspirations to me. The first one is

1524
01:05:10,079 --> 01:05:12,880
Reticulum, also known as RNS. It's a

1525
01:05:12,880 --> 01:05:15,440
type of decentralized network and it's

1526
01:05:15,440 --> 01:05:18,079
super super cool. I can't emphasize how

1527
01:05:18,079 --> 01:05:20,319
cool this is. It is hardware agnostic.

1528
01:05:20,319 --> 01:05:22,480
So, you can use it over Laura radio,

1529
01:05:22,480 --> 01:05:25,839
I2P, TCP, so many different hardwares

1530
01:05:25,839 --> 01:05:28,000
you can choose from. Uh you can see all

1531
01:05:28,000 --> 01:05:30,240
of the active reticulum nodes on rw

1532
01:05:30,240 --> 01:05:32,640
world.map. There's a I2P mirror at

1533
01:05:32,640 --> 01:05:34,400
reticulum.2p

1534
01:05:34,400 --> 01:05:37,280
and you can install it using uh pip.

1535
01:05:37,280 --> 01:05:39,680
It's a python app which is great for

1536
01:05:39,680 --> 01:05:42,720
pyon I guess. And uh you might have to

1537
01:05:42,720 --> 01:05:44,240
break your system packages but that's

1538
01:05:44,240 --> 01:05:48,839
okay because it won't break your system.

1539
01:05:49,039 --> 01:05:51,520
Second of all I want to acknowledge I2P.

1540
01:05:51,520 --> 01:05:56,079
Uh I2P is a niche darknet uh which I

1541
01:05:56,079 --> 01:05:59,280
believe to be uh better than tour in

1542
01:05:59,280 --> 01:06:01,599
many ways. Sorry if there are any tour

1543
01:06:01,599 --> 01:06:03,839
operators here. You're awesome too. Um

1544
01:06:03,839 --> 01:06:07,760
but ITP is decentralized. Um unlike tour

1545
01:06:07,760 --> 01:06:10,000
which has a central or directory

1546
01:06:10,000 --> 01:06:12,079
authority which manages stuff like let's

1547
01:06:12,079 --> 01:06:15,200
hope they don't go corrupt but I2P

1548
01:06:15,200 --> 01:06:17,839
relies on all of the users. Uh if you

1549
01:06:17,839 --> 01:06:20,079
are a peer in I2P, you are actively

1550
01:06:20,079 --> 01:06:21,599
contributing to the network and you're

1551
01:06:21,599 --> 01:06:24,799
helping make I2P go faster. I have a

1552
01:06:24,799 --> 01:06:27,520
computer science network uh a blog

1553
01:06:27,520 --> 01:06:29,760
really on ITP. It's called Aiden

1554
01:06:29,760 --> 01:06:31,680
CSblog.itp

1555
01:06:31,680 --> 01:06:34,160
can go check that out. I'm uh adding to

1556
01:06:34,160 --> 01:06:36,559
it. Uh I have a lot of content on it

1557
01:06:36,559 --> 01:06:38,960
with about Linux and networking. So you

1558
01:06:38,960 --> 01:06:40,960
can check that out if you like. uh you

1559
01:06:40,960 --> 01:06:43,599
can just install it using pseudo apt and

1560
01:06:43,599 --> 01:06:46,000
there is a web console for you to look

1561
01:06:46,000 --> 01:06:51,160
at at localhost uh port 7070.

1562
01:06:51,599 --> 01:06:54,240
And finally I want to acknowledge

1563
01:06:54,240 --> 01:06:57,119
Miuwire for being a heavy inspiration on

1564
01:06:57,119 --> 01:06:59,839
this project. Muire is a app built on

1565
01:06:59,839 --> 01:07:03,680
top of I2P uh and it is a decentralized

1566
01:07:03,680 --> 01:07:06,000
file sharing network that heavily

1567
01:07:06,000 --> 01:07:08,000
influenced the way I implemented the

1568
01:07:08,000 --> 01:07:10,559
file sharing network in my app. It uses

1569
01:07:10,559 --> 01:07:13,599
the gossip protocol to share resources

1570
01:07:13,599 --> 01:07:17,119
and it's a direct descendant of GNUA uh

1571
01:07:17,119 --> 01:07:19,359
which is also a descendant of Napster.

1572
01:07:19,359 --> 01:07:23,359
So P2P networks all love each other.

1573
01:07:23,359 --> 01:07:26,640
Okay, next up the way I encrypted my

1574
01:07:26,640 --> 01:07:29,200
sockets. This is a shark uh image.

1575
01:07:29,200 --> 01:07:32,160
That's because wire shark um to sniff

1576
01:07:32,160 --> 01:07:36,319
the so uh sniff the packets. Um,

1577
01:07:36,319 --> 01:07:39,200
and the first thing I need to do is I

1578
01:07:39,200 --> 01:07:42,079
need to encrypt all of the messages

1579
01:07:42,079 --> 01:07:43,920
going through sockets because sockets

1580
01:07:43,920 --> 01:07:45,599
just sends it in plain text. That's not

1581
01:07:45,599 --> 01:07:47,440
good if you're going to use wireshot to

1582
01:07:47,440 --> 01:07:49,039
sniff the packets and then get all my

1583
01:07:49,039 --> 01:07:52,240
sensitive info. Uh, so I'm using diffy

1584
01:07:52,240 --> 01:07:55,520
helman to exchange keys over a public

1585
01:07:55,520 --> 01:07:58,799
insecure channel. So then uh two peers

1586
01:07:58,799 --> 01:08:02,079
can communicate securely using the uh

1587
01:08:02,079 --> 01:08:04,720
agreed upon key. Uh, Diffy Helman relies

1588
01:08:04,720 --> 01:08:06,640
on the commutative property of

1589
01:08:06,640 --> 01:08:09,200
multiplication to allow these two peers

1590
01:08:09,200 --> 01:08:12,240
to get towards a single uh, agreed on

1591
01:08:12,240 --> 01:08:14,720
key without like talking to each other

1592
01:08:14,720 --> 01:08:17,199
and whispering. You can just do it over

1593
01:08:17,199 --> 01:08:20,319
a public channel. And uh, another really

1594
01:08:20,319 --> 01:08:22,400
big threat for security is the

1595
01:08:22,400 --> 01:08:24,400
man-in-the-middle attack. That's where

1596
01:08:24,400 --> 01:08:27,040
in oh I think that I'm talking directly

1597
01:08:27,040 --> 01:08:29,199
to Bob but actually my message is being

1598
01:08:29,199 --> 01:08:31,359
intercepted by Eve and she's changing

1599
01:08:31,359 --> 01:08:32,960
some stuff and she's pretending to be

1600
01:08:32,960 --> 01:08:35,199
Bob and then she can see all of our

1601
01:08:35,199 --> 01:08:38,080
communications whereas I think I am

1602
01:08:38,080 --> 01:08:41,520
talking to Bob but uh I'm not. That's

1603
01:08:41,520 --> 01:08:44,080
why certificates are very important to

1604
01:08:44,080 --> 01:08:46,239
help us solve this. Unfortunately, I

1605
01:08:46,239 --> 01:08:48,000
don't want to go through the hassle of

1606
01:08:48,000 --> 01:08:50,560
uh let's encrypt and the CA. So, I need

1607
01:08:50,560 --> 01:08:53,759
to self-sign my certificate which um

1608
01:08:53,759 --> 01:08:57,440
failed a lot because uh the SSL um

1609
01:08:57,440 --> 01:08:59,839
library does not like that. So, instead

1610
01:08:59,839 --> 01:09:02,000
I do authentication using challenges.

1611
01:09:02,000 --> 01:09:04,319
So, I set up a public and private key on

1612
01:09:04,319 --> 01:09:07,839
all of the nodes and then I uh ask Bob

1613
01:09:07,839 --> 01:09:10,560
to sign it and sign a random challenge

1614
01:09:10,560 --> 01:09:13,120
and if he signs it, it's sure to be Bob,

1615
01:09:13,120 --> 01:09:15,040
not to be Eve.

1616
01:09:15,040 --> 01:09:17,279
Next up, network architectures. When I

1617
01:09:17,279 --> 01:09:20,239
first submitted the Pyon uh my proposal

1618
01:09:20,239 --> 01:09:22,560
for Pyon, it was initially a server

1619
01:09:22,560 --> 01:09:25,839
client architecture um which was not

1620
01:09:25,839 --> 01:09:28,239
good. It's it's bad. It's bad because if

1621
01:09:28,239 --> 01:09:30,159
the server goes offline, which it

1622
01:09:30,159 --> 01:09:33,040
probably would um because I run out of

1623
01:09:33,040 --> 01:09:35,440
battery very quickly, uh the whole thing

1624
01:09:35,440 --> 01:09:38,319
is going to crash. So instead, I'm

1625
01:09:38,319 --> 01:09:40,239
opting for a peer-to-peer uh

1626
01:09:40,239 --> 01:09:43,359
peer-to-peer network architecture where

1627
01:09:43,359 --> 01:09:45,279
there's a birectional connection between

1628
01:09:45,279 --> 01:09:48,000
every peer. So uh as you can see on this

1629
01:09:48,000 --> 01:09:50,159
diagram, each peer has a receiving

1630
01:09:50,159 --> 01:09:52,159
socket and a sending socket. So it

1631
01:09:52,159 --> 01:09:54,239
receives information from other peers

1632
01:09:54,239 --> 01:09:56,159
using the receiving socket and it can

1633
01:09:56,159 --> 01:09:57,760
send out information to other peers

1634
01:09:57,760 --> 01:10:00,719
using a sending socket.

1635
01:10:00,719 --> 01:10:03,120
Oh, additionally, all the peers have uh

1636
01:10:03,120 --> 01:10:06,480
their own public and private key.

1637
01:10:06,480 --> 01:10:09,360
Okay, so this is the peer-to-peer uh LAN

1638
01:10:09,360 --> 01:10:13,360
app with two users. Uh I have a Debian

1639
01:10:13,360 --> 01:10:16,719
desktop here and it's sending some

1640
01:10:16,719 --> 01:10:20,000
messages to my MacBook here and you can

1641
01:10:20,000 --> 01:10:23,920
see that they're all uh synced up.

1642
01:10:23,920 --> 01:10:26,080
Next is verify messages because you

1643
01:10:26,080 --> 01:10:28,239
never know if messages can be tampered

1644
01:10:28,239 --> 01:10:29,920
with. The way you do this is with

1645
01:10:29,920 --> 01:10:32,159
signatures. So you use a private key to

1646
01:10:32,159 --> 01:10:35,040
sign your message so uh other people can

1647
01:10:35,040 --> 01:10:37,520
be sure that you wrote it and it's not

1648
01:10:37,520 --> 01:10:40,000
being tampered with.

1649
01:10:40,000 --> 01:10:42,159
Uh I added a lot more features to this

1650
01:10:42,159 --> 01:10:44,320
app. So it's not just a chat app. I also

1651
01:10:44,320 --> 01:10:46,080
added a connections manager. So you can

1652
01:10:46,080 --> 01:10:48,719
have trusted and untrusted connections.

1653
01:10:48,719 --> 01:10:51,040
Uh you can add people to your trusted

1654
01:10:51,040 --> 01:10:53,600
list and everyone else they show up in

1655
01:10:53,600 --> 01:10:56,159
the untrusted category here. And you can

1656
01:10:56,159 --> 01:10:57,760
manage your connections so you can close

1657
01:10:57,760 --> 01:11:00,800
and restart them. All right. Now for the

1658
01:11:00,800 --> 01:11:02,960
fun stuff is file sharing. This is the

1659
01:11:02,960 --> 01:11:05,600
gossip protocol which gutella uses. If

1660
01:11:05,600 --> 01:11:07,120
you want to send out a query to a

1661
01:11:07,120 --> 01:11:09,040
decentralized network, there's no single

1662
01:11:09,040 --> 01:11:11,040
central server you have to query. So

1663
01:11:11,040 --> 01:11:12,560
instead, you send it out to all your

1664
01:11:12,560 --> 01:11:14,800
peers who send it out to their peers who

1665
01:11:14,800 --> 01:11:16,800
send it out to their peers and so on.

1666
01:11:16,800 --> 01:11:19,360
Usually uh the peers will pick uh some

1667
01:11:19,360 --> 01:11:20,960
random connections to send it to, but

1668
01:11:20,960 --> 01:11:22,480
because I'm just doing this in the land

1669
01:11:22,480 --> 01:11:24,320
and there's not many people anyways, I'm

1670
01:11:24,320 --> 01:11:26,159
just sending it to all peers instead of

1671
01:11:26,159 --> 01:11:29,120
a random selection.

1672
01:11:29,120 --> 01:11:31,199
Uh I also want to mention a concept

1673
01:11:31,199 --> 01:11:34,159
known as the web of trust which I think

1674
01:11:34,159 --> 01:11:37,360
is better than the CA. Um because

1675
01:11:37,360 --> 01:11:40,719
instead of having to register uh with a

1676
01:11:40,719 --> 01:11:44,000
single central authority, you can uh

1677
01:11:44,000 --> 01:11:46,640
just chat up with your friends and add

1678
01:11:46,640 --> 01:11:49,199
each other to your own trusted lists.

1679
01:11:49,199 --> 01:11:53,280
And this way uh if you trust Bob and Bob

1680
01:11:53,280 --> 01:11:56,480
trusts Charlie then you indirectly trust

1681
01:11:56,480 --> 01:11:59,440
Charlie which is what I use for resource

1682
01:11:59,440 --> 01:12:02,480
queries. So I added a file sharing uh

1683
01:12:02,480 --> 01:12:05,760
feature to my chat app. Uh so you can

1684
01:12:05,760 --> 01:12:09,040
query for resources and your queries get

1685
01:12:09,040 --> 01:12:12,719
bounced to different peers. Um but uh in

1686
01:12:12,719 --> 01:12:14,800
the in the initial versions I didn't

1687
01:12:14,800 --> 01:12:16,480
really have a way for you to know

1688
01:12:16,480 --> 01:12:19,679
whether uh the queries you receive the

1689
01:12:19,679 --> 01:12:21,199
resources you receive whether they're

1690
01:12:21,199 --> 01:12:23,360
trusted or not. So I had to add a trust

1691
01:12:23,360 --> 01:12:25,760
chain to implement the web of trust

1692
01:12:25,760 --> 01:12:28,159
protocol. Uh so you can see how many

1693
01:12:28,159 --> 01:12:31,040
hops the query took and whether it's

1694
01:12:31,040 --> 01:12:33,199
along a chain of trust or whether it's

1695
01:12:33,199 --> 01:12:35,600
not. Uh because remember you can

1696
01:12:35,600 --> 01:12:37,840
actually have connections with untrusted

1697
01:12:37,840 --> 01:12:41,280
peers. uh which is something that Miwire

1698
01:12:41,280 --> 01:12:42,960
and Kutella I don't think they support

1699
01:12:42,960 --> 01:12:46,640
yet. Uh unfortunately Miwire stopped uh

1700
01:12:46,640 --> 01:12:48,719
stopped maintenance which is really sad.

1701
01:12:48,719 --> 01:12:51,760
I hope someone picks that up. Um all

1702
01:12:51,760 --> 01:12:54,880
right and next for file sharing uh you

1703
01:12:54,880 --> 01:12:57,920
can share files which is really cool. Um

1704
01:12:57,920 --> 01:13:00,159
however if the file gets tampered with

1705
01:13:00,159 --> 01:13:02,400
there's a way you can know which is by

1706
01:13:02,400 --> 01:13:07,199
using hashes. So I add a hash uh a hash

1707
01:13:07,199 --> 01:13:09,520
entry to my database so you can ensure

1708
01:13:09,520 --> 01:13:12,000
that the file you receive on your end

1709
01:13:12,000 --> 01:13:14,239
has not been corrupted.

1710
01:13:14,239 --> 01:13:17,120
Uh and this is a demo where I sent the

1711
01:13:17,120 --> 01:13:21,360
Otus uh PDF to another computer. Otus is

1712
01:13:21,360 --> 01:13:23,840
a great Math Olympia textbook. Highly

1713
01:13:23,840 --> 01:13:25,920
recommend.

1714
01:13:25,920 --> 01:13:28,159
Anyways, commenting. Now you can also

1715
01:13:28,159 --> 01:13:30,800
comment on resources which uh is an

1716
01:13:30,800 --> 01:13:33,360
extra feature that um UI didn't have.

1717
01:13:33,360 --> 01:13:35,280
But I wanted to addit it uh because I

1718
01:13:35,280 --> 01:13:37,199
think it's really cool. You basically

1719
01:13:37,199 --> 01:13:38,880
create these the same way as you do with

1720
01:13:38,880 --> 01:13:41,280
other resources. Um but you just have to

1721
01:13:41,280 --> 01:13:44,239
paste the hash of the resource as the

1722
01:13:44,239 --> 01:13:47,199
label. So that way your comment is uh

1723
01:13:47,199 --> 01:13:50,000
tied towards the resource. This way is

1724
01:13:50,000 --> 01:13:53,120
if you select a resource and you press

1725
01:13:53,120 --> 01:13:55,520
uh query comments then you can find all

1726
01:13:55,520 --> 01:13:57,520
the comments with the hash of the

1727
01:13:57,520 --> 01:14:00,239
resource as its label.

1728
01:14:00,239 --> 01:14:02,000
Now for the stuff I learned. First of

1729
01:14:02,000 --> 01:14:04,560
all, decentralized network is really

1730
01:14:04,560 --> 01:14:07,440
really cool and I don't need to let the

1731
01:14:07,440 --> 01:14:10,159
CIA uh the CA tell you that I'm a human

1732
01:14:10,159 --> 01:14:11,679
being.

1733
01:14:11,679 --> 01:14:15,440
Um and secondly, uh there's so many

1734
01:14:15,440 --> 01:14:17,679
really cool decentralized B2B networks

1735
01:14:17,679 --> 01:14:19,440
and I highly encourage you to check them

1736
01:14:19,440 --> 01:14:22,800
out, especially Reticular. Third of all,

1737
01:14:22,800 --> 01:14:24,880
um IPv6

1738
01:14:24,880 --> 01:14:28,320
is great, NAT is terrible, IPv4 is

1739
01:14:28,320 --> 01:14:31,280
terrible. I cannot stress how many times

1740
01:14:31,280 --> 01:14:34,960
CGNAT caused to have so many bugs in my

1741
01:14:34,960 --> 01:14:38,400
code and my I2B setup. So CGNAT is a

1742
01:14:38,400 --> 01:14:40,400
really no really big no no. Please turn

1743
01:14:40,400 --> 01:14:45,546
it off when you get home. Um [laughter]

1744
01:14:46,239 --> 01:14:48,159
uh fourth of all uh please don't laugh

1745
01:14:48,159 --> 01:14:50,000
at other people because they spent 2,000

1746
01:14:50,000 --> 01:14:51,760
lines of code for a simple chatup. It's

1747
01:14:51,760 --> 01:14:54,717
actually pretty hard. [laughter]

1748
01:14:57,281 --> 01:14:58,640
[applause]

1749
01:14:58,640 --> 01:15:01,120
Uh, number five, check your firewall

1750
01:15:01,120 --> 01:15:05,120
before you give up debugging. I um had a

1751
01:15:05,120 --> 01:15:08,080
UFW and it blocked all my ports, which

1752
01:15:08,080 --> 01:15:09,520
was really annoying. I didn't know what

1753
01:15:09,520 --> 01:15:12,159
went wrong, but you just enable the port

1754
01:15:12,159 --> 01:15:16,000
and it works. [laughter] Uh, and here is

1755
01:15:16,000 --> 01:15:18,640
my GitHub repo. You can scan the QR

1756
01:15:18,640 --> 01:15:21,040
code. It's not that great. 2,000 lines

1757
01:15:21,040 --> 01:15:23,840
of code in one file, but I'm making a

1758
01:15:23,840 --> 01:15:25,840
dev branch and I'm refactoring all my

1759
01:15:25,840 --> 01:15:27,679
code and then modularizing it and

1760
01:15:27,679 --> 01:15:30,000
splitting it up like a professional dev.

1761
01:15:30,000 --> 01:15:32,000
So hopefully

1762
01:15:32,000 --> 01:15:33,679
hopefully that would be better and

1763
01:15:33,679 --> 01:15:35,679
easier to debug. If you find any issues,

1764
01:15:35,679 --> 01:15:39,280
please submit a PR. Um, [laughter]

1765
01:15:39,280 --> 01:15:41,600
uh, some steps for the future. First of

1766
01:15:41,600 --> 01:15:43,600
all, peer discovery. I think I'm already

1767
01:15:43,600 --> 01:15:44,960
doing a good job at that because I'm

1768
01:15:44,960 --> 01:15:46,800
allowing untrusted connections although

1769
01:15:46,800 --> 01:15:50,239
I flag them but I also want to have u a

1770
01:15:50,239 --> 01:15:52,719
specialized feature for discovering new

1771
01:15:52,719 --> 01:15:55,600
peers. Second of all uh user creation of

1772
01:15:55,600 --> 01:15:57,920
channels. My current chat app is similar

1773
01:15:57,920 --> 01:16:00,239
to IRC in terms of aesthetic even though

1774
01:16:00,239 --> 01:16:01,920
it's not actually a server client

1775
01:16:01,920 --> 01:16:04,000
architecture. It's P2P but it does have

1776
01:16:04,000 --> 01:16:06,400
channels but they're all hardcoded. So I

1777
01:16:06,400 --> 01:16:08,400
think that uh user creation of channels

1778
01:16:08,400 --> 01:16:11,120
would be a good step to add. Um, third

1779
01:16:11,120 --> 01:16:13,600
of all, uh, I want to expand this to the

1780
01:16:13,600 --> 01:16:17,840
WAN, potentially I2P. Um, since like,

1781
01:16:17,840 --> 01:16:19,280
you know, if you're in a land, you can

1782
01:16:19,280 --> 01:16:21,120
just talk to people. Like, do you really

1783
01:16:21,120 --> 01:16:22,560
need to? [laughter]

1784
01:16:22,560 --> 01:16:25,199
Uh, and yeah, now I need to clean up the

1785
01:16:25,199 --> 01:16:26,960
code base. Hopefully, it will look

1786
01:16:26,960 --> 01:16:28,719
really good. Add more comments. It will

1787
01:16:28,719 --> 01:16:31,920
all be great. And that's it. Thank you

1788
01:16:31,920 --> 01:16:34,222
for listening.

1789
01:16:34,222 --> 01:16:36,242
[applause and cheering]

1790
01:16:37,199 --> 01:16:40,773
Massive round of applause for Emma.

1791
01:16:40,773 --> 01:16:42,793
[applause and cheering]

1792
01:16:43,600 --> 01:16:48,640
Yes. H I mean I just Yes. Yes. Yes. Yes.

1793
01:16:48,640 --> 01:16:50,239
Okay.

1794
01:16:50,239 --> 01:16:54,440
It's the last talk.

1795
01:16:56,080 --> 01:16:59,520
We might have time at the end of this

1796
01:16:59,520 --> 01:17:02,880
talk for a couple of questions. I'm not

1797
01:17:02,880 --> 01:17:05,040
going to promise. I don't know when the

1798
01:17:05,040 --> 01:17:07,679
venue I'm avoiding eye contact will kick

1799
01:17:07,679 --> 01:17:09,120
us out.

1800
01:17:09,120 --> 01:17:10,640
Um,

1801
01:17:10,640 --> 01:17:13,440
but what we will be doing is getting all

1802
01:17:13,440 --> 01:17:17,440
of the links that the students have uh

1803
01:17:17,440 --> 01:17:20,000
put on the put on slides I want to share

1804
01:17:20,000 --> 01:17:21,520
and I will be putting them all in the

1805
01:17:21,520 --> 01:17:23,199
discord unless the students have already

1806
01:17:23,199 --> 01:17:25,920
put them in the discord. Email them to

1807
01:17:25,920 --> 01:17:28,159
me right now otherwise. Um, but it's

1808
01:17:28,159 --> 01:17:30,800
okay. There's one last talk. We can do

1809
01:17:30,800 --> 01:17:32,518
it.

1810
01:17:32,518 --> 01:17:33,280
[cheering]

1811
01:17:33,280 --> 01:17:35,920
This is Jasmine. [applause]

1812
01:17:35,920 --> 01:17:38,000
Jasmine is going to be talking about

1813
01:17:38,000 --> 01:17:40,880
friendship.

1814
01:17:40,880 --> 01:17:42,880
Uh Jasmine has been programming since

1815
01:17:42,880 --> 01:17:44,960
primary school.

1816
01:17:44,960 --> 01:17:48,080
Um oh gosh,

1817
01:17:48,080 --> 01:17:50,480
really?

1818
01:17:50,480 --> 01:17:53,478
Okay, there are two of them. Um

1819
01:17:53,478 --> 01:17:55,199
[laughter]

1820
01:17:55,199 --> 01:17:59,920
I'm an adult. Uh okay. Um

1821
01:17:59,920 --> 01:18:03,760
cat. Cat. Squinty cat. Angry cat.

1822
01:18:03,760 --> 01:18:07,199
Oh, squinty cat. Angry cat. I don't

1823
01:18:07,199 --> 01:18:08,880
I'm going to say that's Yes, it's

1824
01:18:08,880 --> 01:18:11,120
smirking cat.

1825
01:18:11,120 --> 01:18:13,760
Yeah, I that's and I don't even know how

1826
01:18:13,760 --> 01:18:16,080
to uh

1827
01:18:16,080 --> 01:18:19,120
uh quick

1828
01:18:19,120 --> 01:18:22,400
I need four more green shirts or three

1829
01:18:22,400 --> 01:18:24,640
more green shirts. Come on. Come on.

1830
01:18:24,640 --> 01:18:28,480
This is totally worth it. Yeah. Okay. Um

1831
01:18:28,480 --> 01:18:31,520
you stand there. You No. No. Alex,

1832
01:18:31,520 --> 01:18:34,239
you're not needed. You stand. Kneel down

1833
01:18:34,239 --> 01:18:38,019
here. And now be more plant-based.

1834
01:18:38,019 --> 01:18:40,039
[laughter]

1835
01:18:41,679 --> 01:18:47,960
Yes. A fourleaf clover. Yeah.

1836
01:18:49,809 --> 01:18:52,320
[applause] Thank you. Amazing GPN

1837
01:18:52,320 --> 01:18:53,739
tutors. Always willing to step

1838
01:18:53,739 --> 01:18:57,040
[laughter] in. Um

1839
01:18:57,040 --> 01:18:59,199
we're normal adults. Um, Jasmine

1840
01:18:59,199 --> 01:19:02,239
recently learned about the new reaction

1841
01:19:02,239 --> 01:19:04,640
video setting when you're screen

1842
01:19:04,640 --> 01:19:06,560
recording on your phone and she said

1843
01:19:06,560 --> 01:19:09,040
that her mom has loved this discovery

1844
01:19:09,040 --> 01:19:11,920
and that she now gets silly reaction

1845
01:19:11,920 --> 01:19:14,320
videos to music videos. I love this. I

1846
01:19:14,320 --> 01:19:17,600
love this so much. So, one more massive

1847
01:19:17,600 --> 01:19:19,939
round of applause to welcome Jasmine.

1848
01:19:19,939 --> 01:19:21,905
[cheering]

1849
01:19:21,905 --> 01:19:23,679
[applause]

1850
01:19:23,679 --> 01:19:27,679
Hello everyone. Hi. Um, so I'm Jasmine

1851
01:19:27,679 --> 01:19:30,239
Terron and this is my talk, Friendship

1852
01:19:30,239 --> 01:19:33,679
is Magic. Um, and it's a data

1853
01:19:33,679 --> 01:19:35,040
investigation that uses machine

1854
01:19:35,040 --> 01:19:38,800
learning. So, a little bit about me. I'm

1855
01:19:38,800 --> 01:19:41,360
a year 12 student and I study at Bergman

1856
01:19:41,360 --> 01:19:44,719
Anglican School in CRA. And as you may

1857
01:19:44,719 --> 01:19:48,400
guess from the topic, I really like My

1858
01:19:48,400 --> 01:19:51,199
Little Pony.

1859
01:19:51,199 --> 01:19:56,159
Like, I really like My Little Pony.

1860
01:19:56,159 --> 01:20:00,480
Look how cool I am. That's pretty cool.

1861
01:20:00,480 --> 01:20:02,239
So, what did I want to do for my

1862
01:20:02,239 --> 01:20:04,719
project? I wanted to find out what makes

1863
01:20:04,719 --> 01:20:07,920
their friendship so special to me or

1864
01:20:07,920 --> 01:20:10,640
find the magic behind their friendship

1865
01:20:10,640 --> 01:20:13,280
and learn how to replicate that.

1866
01:20:13,280 --> 01:20:16,000
So, what is my project?

1867
01:20:16,000 --> 01:20:18,159
It analyzes the 30 characters with the

1868
01:20:18,159 --> 01:20:20,159
most lines in Friendship is Magic. And

1869
01:20:20,159 --> 01:20:21,920
it starts by looking at all the

1870
01:20:21,920 --> 01:20:23,360
characters and what makes them different

1871
01:20:23,360 --> 01:20:25,920
and interesting. Then we do sentiment

1872
01:20:25,920 --> 01:20:28,960
analysis and investigation.

1873
01:20:28,960 --> 01:20:30,640
We do sentiment analysis with two

1874
01:20:30,640 --> 01:20:32,400
different methods. We do Vader and

1875
01:20:32,400 --> 01:20:34,560
Roberta. And once the better option is

1876
01:20:34,560 --> 01:20:36,400
found, investigation was continued with

1877
01:20:36,400 --> 01:20:38,000
that.

1878
01:20:38,000 --> 01:20:40,960
So what is sentiment? Sentiment is

1879
01:20:40,960 --> 01:20:43,280
basically the overall emotional tone of

1880
01:20:43,280 --> 01:20:46,560
a piece of text. So if numerical values

1881
01:20:46,560 --> 01:20:48,560
come back higher in positive, it would

1882
01:20:48,560 --> 01:20:51,360
evoke positive feelings in the viewer.

1883
01:20:51,360 --> 01:20:54,640
My hypothesis was that characters in the

1884
01:20:54,640 --> 01:20:56,320
show will speak with an overall positive

1885
01:20:56,320 --> 01:20:58,560
or neutral sentiment on average in order

1886
01:20:58,560 --> 01:21:00,159
to appeal to the show's young target

1887
01:21:00,159 --> 01:21:02,000
audience.

1888
01:21:02,000 --> 01:21:04,640
This is the data set that I used. It's

1889
01:21:04,640 --> 01:21:06,480
um seasons 1 through eight of Friendship

1890
01:21:06,480 --> 01:21:08,560
is Magic as well as two of the movies.

1891
01:21:08,560 --> 01:21:11,120
Um basically in My Little Pony, each

1892
01:21:11,120 --> 01:21:13,440
season contains 26 episodes except for

1893
01:21:13,440 --> 01:21:15,840
season 3 which was created as a possible

1894
01:21:15,840 --> 01:21:17,679
season finale before the show got

1895
01:21:17,679 --> 01:21:20,560
renewed for season 4. Um, and the data

1896
01:21:20,560 --> 01:21:22,880
is broken up into four smaller subsets,

1897
01:21:22,880 --> 01:21:25,360
which pretty much just made it easier to

1898
01:21:25,360 --> 01:21:28,320
access the data that I needed. Um, yeah,

1899
01:21:28,320 --> 01:21:30,400
onto visualizations.

1900
01:21:30,400 --> 01:21:32,480
So, here we see the top 30 characters

1901
01:21:32,480 --> 01:21:34,320
with the most lines sorted from highest

1902
01:21:34,320 --> 01:21:36,719
to lowest. Um, these will basically just

1903
01:21:36,719 --> 01:21:38,640
be used as the order for further graphs

1904
01:21:38,640 --> 01:21:41,199
in the future. As we can see, Twilight

1905
01:21:41,199 --> 01:21:44,159
speaks a lot. She is the main character.

1906
01:21:44,159 --> 01:21:46,880
Nice. Here, instead of the most lines

1907
01:21:46,880 --> 01:21:48,480
said, it's the most words said within

1908
01:21:48,480 --> 01:21:50,480
the script. This follows a similar

1909
01:21:50,480 --> 01:21:52,639
trend. However, some characters say more

1910
01:21:52,639 --> 01:21:55,360
words in their sentences. For example,

1911
01:21:55,360 --> 01:21:57,280
Rarity says lots of words within her

1912
01:21:57,280 --> 01:21:58,800
lines, which makes sense within her

1913
01:21:58,800 --> 01:22:00,800
character. The other graph just shows

1914
01:22:00,800 --> 01:22:02,560
the average number of words per line

1915
01:22:02,560 --> 01:22:05,760
with the overall being around 11. Now,

1916
01:22:05,760 --> 01:22:07,120
we'll move on to the start of our

1917
01:22:07,120 --> 01:22:09,840
sentiment analysis.

1918
01:22:09,840 --> 01:22:12,560
The Vader method shows majority neutral

1919
01:22:12,560 --> 01:22:15,520
sentiment at around 70%. With little

1920
01:22:15,520 --> 01:22:17,840
differences between characters, all

1921
01:22:17,840 --> 01:22:19,520
characters yielding similar results

1922
01:22:19,520 --> 01:22:22,080
highlights how the VETA model is likely

1923
01:22:22,080 --> 01:22:24,159
inaccurate. Vader is basically a

1924
01:22:24,159 --> 01:22:25,920
rule-based tool that uses a dictionary

1925
01:22:25,920 --> 01:22:28,080
of words with assigned sentiment values

1926
01:22:28,080 --> 01:22:30,159
where the sum of the words present gives

1927
01:22:30,159 --> 01:22:32,080
the overall sentiment as a numerical

1928
01:22:32,080 --> 01:22:34,800
value. Basically, you've got a bunch of

1929
01:22:34,800 --> 01:22:36,960
words. Each word has a number. You add

1930
01:22:36,960 --> 01:22:39,840
them together. That's your sentiment.

1931
01:22:39,840 --> 01:22:41,600
The key limitation with this tool is

1932
01:22:41,600 --> 01:22:43,280
that the dictionary is static and it

1933
01:22:43,280 --> 01:22:44,880
works from predefined values for the

1934
01:22:44,880 --> 01:22:46,639
words. Therefore, it lacks the

1935
01:22:46,639 --> 01:22:48,480
contextual understanding to understand

1936
01:22:48,480 --> 01:22:50,800
things like sarcasm.

1937
01:22:50,800 --> 01:22:53,360
In the Robera model, it shows characters

1938
01:22:53,360 --> 01:22:55,440
having an average sentiment of around

1939
01:22:55,440 --> 01:22:58,239
50% neutral and 25% positive and

1940
01:22:58,239 --> 01:23:00,880
negative. The Roberto model is a

1941
01:23:00,880 --> 01:23:02,880
transformersbased machine learning model

1942
01:23:02,880 --> 01:23:04,400
that is used in order to analyze

1943
01:23:04,400 --> 01:23:06,159
sentiment in terms of the context words

1944
01:23:06,159 --> 01:23:08,000
are coming from instead of the words

1945
01:23:08,000 --> 01:23:10,080
themselves.

1946
01:23:10,080 --> 01:23:11,920
That makes it the preferred method for

1947
01:23:11,920 --> 01:23:15,679
heavily sarcastic or ironic pieces. The

1948
01:23:15,679 --> 01:23:17,280
limitation with the Roberto model is

1949
01:23:17,280 --> 01:23:20,560
that it has a token limit of 512. Um, so

1950
01:23:20,560 --> 01:23:22,159
it means that not all the text within a

1951
01:23:22,159 --> 01:23:23,920
section can be processed within the same

1952
01:23:23,920 --> 01:23:26,719
group. And just so you know, 512 tokens

1953
01:23:26,719 --> 01:23:30,239
is typically 400 to 500 words. So that

1954
01:23:30,239 --> 01:23:32,560
means while Roberta was the obvious

1955
01:23:32,560 --> 01:23:34,960
better pick, the graph here only

1956
01:23:34,960 --> 01:23:36,880
corresponds to the data in the first

1957
01:23:36,880 --> 01:23:40,239
season that a character's in.

1958
01:23:40,239 --> 01:23:42,960
So I decided to gather sentiment data

1959
01:23:42,960 --> 01:23:44,560
for every season that was available

1960
01:23:44,560 --> 01:23:46,719
within the data set in order to analyze

1961
01:23:46,719 --> 01:23:48,800
how sentiment changed over the show and

1962
01:23:48,800 --> 01:23:51,120
how characters grew. The differences

1963
01:23:51,120 --> 01:23:52,639
within these graphs could be due to the

1964
01:23:52,639 --> 01:23:54,880
character story arcs including shifts in

1965
01:23:54,880 --> 01:23:57,199
character motivations, relationships,

1966
01:23:57,199 --> 01:23:59,280
personal growth, and all of this could

1967
01:23:59,280 --> 01:24:02,400
potentially create differences.

1968
01:24:02,400 --> 01:24:03,920
Interestingly,

1969
01:24:03,920 --> 01:24:05,600
Friendship is Magic also had many

1970
01:24:05,600 --> 01:24:07,760
changes to showrunners and directors

1971
01:24:07,760 --> 01:24:09,679
during its time, which may have also

1972
01:24:09,679 --> 01:24:11,840
influenced its direction.

1973
01:24:11,840 --> 01:24:13,760
The top section shows the seasons with

1974
01:24:13,760 --> 01:24:15,920
the original developer and showrunner,

1975
01:24:15,920 --> 01:24:17,600
and the next section is the next

1976
01:24:17,600 --> 01:24:20,320
showrunner. In the first seasons, the

1977
01:24:20,320 --> 01:24:21,920
show was still finding its footing and

1978
01:24:21,920 --> 01:24:24,719
as a result in this section in season 2,

1979
01:24:24,719 --> 01:24:26,960
a higher negativity within the main six

1980
01:24:26,960 --> 01:24:29,679
can be seen which does not occur in any

1981
01:24:29,679 --> 01:24:31,440
other season.

1982
01:24:31,440 --> 01:24:33,440
Additionally, all these seasons were d

1983
01:24:33,440 --> 01:24:36,639
created under this director. Now, watch

1984
01:24:36,639 --> 01:24:38,320
what happens when we change director in

1985
01:24:38,320 --> 01:24:40,239
season 6.

1986
01:24:40,239 --> 01:24:42,400
Not only are there many new characters,

1987
01:24:42,400 --> 01:24:44,159
but some episodes become more highly

1988
01:24:44,159 --> 01:24:47,040
positive with less negative or emotional

1989
01:24:47,040 --> 01:24:50,080
episodes. From reviews, some fans prefer

1990
01:24:50,080 --> 01:24:51,920
the original portrayal, while others

1991
01:24:51,920 --> 01:24:53,760
appreciate the new directions taken by

1992
01:24:53,760 --> 01:24:56,320
subsequent writers. I believe that this

1993
01:24:56,320 --> 01:24:58,400
speaks to every version of a character

1994
01:24:58,400 --> 01:25:00,880
being important.

1995
01:25:00,880 --> 01:25:02,639
Interestingly,

1996
01:25:02,639 --> 01:25:04,400
the first season that this new director

1997
01:25:04,400 --> 01:25:06,719
directs looks highly similar to the data

1998
01:25:06,719 --> 01:25:09,840
from season 3, which was the big hit

1999
01:25:09,840 --> 01:25:11,440
season that ended up getting the show

2000
01:25:11,440 --> 01:25:13,199
renewed. And I just thought this was

2001
01:25:13,199 --> 01:25:14,880
interesting.

2002
01:25:14,880 --> 01:25:16,239
Looking back to the initial

2003
01:25:16,239 --> 01:25:18,000
visualization, I wanted to look at the

2004
01:25:18,000 --> 01:25:20,880
outliers in a further depth. These are

2005
01:25:20,880 --> 01:25:23,360
our positive outliers. Up the top, we'll

2006
01:25:23,360 --> 01:25:25,040
start with Starlight Glimmer. You can

2007
01:25:25,040 --> 01:25:27,360
see a huge level of positivity within

2008
01:25:27,360 --> 01:25:31,120
her first season, which is interesting

2009
01:25:31,120 --> 01:25:32,800
because she's actually a villain to the

2010
01:25:32,800 --> 01:25:34,560
main six in this season. So, she

2011
01:25:34,560 --> 01:25:37,280
shouldn't be positive. Um, the reason

2012
01:25:37,280 --> 01:25:39,120
for this is that while the things she

2013
01:25:39,120 --> 01:25:41,440
says are positive, the overall mood of

2014
01:25:41,440 --> 01:25:43,920
the episodes and the real meaning of

2015
01:25:43,920 --> 01:25:46,320
what she's saying is negative, which the

2016
01:25:46,320 --> 01:25:48,080
model couldn't pick up from just the

2017
01:25:48,080 --> 01:25:50,639
text that it was given. You can then see

2018
01:25:50,639 --> 01:25:54,239
that in season 6, seven, and eight,

2019
01:25:54,239 --> 01:25:56,239
slow increases in sentiment as she

2020
01:25:56,239 --> 01:25:58,719
finally begins to be genuine and becomes

2021
01:25:58,719 --> 01:26:01,920
more friendly. Cheerly is highly

2022
01:26:01,920 --> 01:26:04,719
positive as she is a teacher and she has

2023
01:26:04,719 --> 01:26:08,009
to do that for her job.

2024
01:26:08,009 --> 01:26:10,029
[laughter]

2025
01:26:10,719 --> 01:26:13,520
But this data misses the nuance with her

2026
01:26:13,520 --> 01:26:16,880
facial expressions that is shown in

2027
01:26:16,880 --> 01:26:20,960
which conveys her real emotions.

2028
01:26:21,040 --> 01:26:23,199
Additionally, Mort and Spitfire are both

2029
01:26:23,199 --> 01:26:25,280
highly sarcastic characters. And while

2030
01:26:25,280 --> 01:26:26,800
Roberta is the preferred method for

2031
01:26:26,800 --> 01:26:29,520
picking up sarcasm, it also can't show

2032
01:26:29,520 --> 01:26:31,040
their facial expressions that show how

2033
01:26:31,040 --> 01:26:33,120
they truly feel and therefore portrays

2034
01:26:33,120 --> 01:26:35,679
them as more positive than they are. I

2035
01:26:35,679 --> 01:26:37,600
think interestingly, you can also see

2036
01:26:37,600 --> 01:26:40,159
character arcs within this data, and

2037
01:26:40,159 --> 01:26:41,360
these characters aren't all

2038
01:26:41,360 --> 01:26:43,920
overwhelmingly positive.

2039
01:26:43,920 --> 01:26:45,840
I also made visualizations for the main

2040
01:26:45,840 --> 01:26:48,639
six. I think most interestingly in this

2041
01:26:48,639 --> 01:26:50,880
is that Twilight has a huge negative

2042
01:26:50,880 --> 01:26:53,360
section in season 4. This is because

2043
01:26:53,360 --> 01:26:55,199
she's learning to become a princess and

2044
01:26:55,199 --> 01:26:58,960
like I would be stressed about that too.

2045
01:26:58,960 --> 01:27:00,960
And then she's positive again in season

2046
01:27:00,960 --> 01:27:02,639
6 when she's basically becoming a

2047
01:27:02,639 --> 01:27:04,960
teacher and finding her calling. I think

2048
01:27:04,960 --> 01:27:06,560
it's cool that across all these

2049
01:27:06,560 --> 01:27:08,400
characters, their key traits still stay

2050
01:27:08,400 --> 01:27:10,480
the same even when some characterize

2051
01:27:10,480 --> 01:27:12,400
them slightly differently or they're

2052
01:27:12,400 --> 01:27:14,560
going through a tough time. I think this

2053
01:27:14,560 --> 01:27:16,320
consistency is what makes it so

2054
01:27:16,320 --> 01:27:20,719
comforting for me and for many others.

2055
01:27:20,719 --> 01:27:22,880
I've attempted and am working on further

2056
01:27:22,880 --> 01:27:24,560
project which would use this data set of

2057
01:27:24,560 --> 01:27:26,880
scripts for generation of new and unique

2058
01:27:26,880 --> 01:27:29,040
episodes through machine learning.

2059
01:27:29,040 --> 01:27:31,840
However, while that is cool, I think a

2060
01:27:31,840 --> 01:27:33,199
lot more value can be gained from

2061
01:27:33,199 --> 01:27:35,679
analyzing existing work and appreciating

2062
01:27:35,679 --> 01:27:38,080
the show that was so special to little

2063
01:27:38,080 --> 01:27:40,960
me as well. And because we looked at how

2064
01:27:40,960 --> 01:27:42,880
each writer and time period shaped the

2065
01:27:42,880 --> 01:27:44,719
characters and their sentiment, this

2066
01:27:44,719 --> 01:27:46,239
will be very useful for building

2067
01:27:46,239 --> 01:27:49,600
accuracy within this project to add on.

2068
01:27:49,600 --> 01:27:53,360
So I think what I really learned wasn't

2069
01:27:53,360 --> 01:27:56,159
my struggles with using Roberta or how

2070
01:27:56,159 --> 01:27:58,400
I'd managed to store all the variables I

2071
01:27:58,400 --> 01:28:00,880
was gen generating in a spreadsheet of

2072
01:28:00,880 --> 01:28:04,880
doom. I think being authentic in all

2073
01:28:04,880 --> 01:28:06,880
your negatives and positives, in all

2074
01:28:06,880 --> 01:28:09,920
your achievements and flaws, in every

2075
01:28:09,920 --> 01:28:12,320
characterization of yourself, being

2076
01:28:12,320 --> 01:28:14,800
authentically you is the greatest magic

2077
01:28:14,800 --> 01:28:17,199
you can give yourself.

2078
01:28:17,199 --> 01:28:19,010
Thank you.

2079
01:28:19,010 --> 01:28:19,666
[cheering]

2080
01:28:19,666 --> 01:28:21,686
[applause]

2081
01:28:23,679 --> 01:28:26,239
Can we get a massive round of applause?

2082
01:28:26,239 --> 01:28:29,199
And indeed, a massive round of applause

2083
01:28:29,199 --> 01:28:33,120
for all our superbly fabulous speakers.

2084
01:28:33,120 --> 01:28:36,041
Students, come on up here.

2085
01:28:36,041 --> 01:28:37,520
[applause]

2086
01:28:37,520 --> 01:28:39,199
All of you, come on. I didn't warn them

2087
01:28:39,199 --> 01:28:42,080
about this. Come on, students. Go on. Up

2088
01:28:42,080 --> 01:28:44,960
on deck. Up on deck. [applause]

2089
01:28:44,960 --> 01:28:47,360
Yes. I think we've all learned

2090
01:28:47,360 --> 01:28:48,960
something.

2091
01:28:48,960 --> 01:28:53,726
It's that I'm crying. No. [cheering]

2092
01:28:53,726 --> 01:28:55,746
[applause]

2093
01:28:57,840 --> 01:28:59,120
Yes. [screaming]

2094
01:28:59,120 --> 01:29:01,360
Okay. All of the students are being

2095
01:29:01,360 --> 01:29:02,960
mugged.

2096
01:29:02,960 --> 01:29:08,136
Keep that applause going. Epic. Yes.

2097
01:29:08,136 --> 01:29:10,101
[cheering]

2098
01:29:10,101 --> 01:29:12,121
[applause]

2099
01:29:16,560 --> 01:29:19,927
What a journey we've been on.

2100
01:29:19,927 --> 01:29:20,400
[cheering]

2101
01:29:20,400 --> 01:29:22,400
We've got three minutes before we get

2102
01:29:22,400 --> 01:29:25,822
kicked out of the venue.

2103
01:29:25,822 --> 01:29:26,000
[cheering]

2104
01:29:26,000 --> 01:29:28,560
Who has questions?

2105
01:29:28,560 --> 01:29:32,120
Who has questions?

2106
01:29:34,239 --> 01:29:36,880
Yes, there are too many questions. It's

2107
01:29:36,880 --> 01:29:38,880
okay. We

2108
01:29:38,880 --> 01:29:41,280
there's I I know there's so so many

2109
01:29:41,280 --> 01:29:42,880
things.

2110
01:29:42,880 --> 01:29:45,520
Um I've got a 16-year-old sister who

2111
01:29:45,520 --> 01:29:47,360
hasn't even started programming. How

2112
01:29:47,360 --> 01:29:49,199
should I get her into it?

2113
01:29:49,199 --> 01:29:52,022
Oh,

2114
01:29:52,022 --> 01:29:53,600
[laughter] remember that fourleaf

2115
01:29:53,600 --> 01:29:55,440
clover?

2116
01:29:55,440 --> 01:29:57,280
Uh, would you like to direct this at any

2117
01:29:57,280 --> 01:29:59,199
particular one or should they all just

2118
01:29:59,199 --> 01:30:02,239
yell out a word at once and then I will

2119
01:30:02,239 --> 01:30:05,280
summarize that?

2120
01:30:05,280 --> 01:30:06,000
Pay her.

2121
01:30:06,000 --> 01:30:08,960
Pay her. [laughter] Honestly,

2122
01:30:08,960 --> 01:30:13,037
that is the best idea. [cheering]

2123
01:30:13,679 --> 01:30:16,239
Uh, there are heaps of resources. I will

2124
01:30:16,239 --> 01:30:18,239
make sure that some get posted in the

2125
01:30:18,239 --> 01:30:20,800
Discord, but paying would be a fantastic

2126
01:30:20,800 --> 01:30:22,862
idea. [laughter]

2127
01:30:23,199 --> 01:30:27,080
Is there another question?

2128
01:30:27,520 --> 01:30:29,199
I I heard someone say, "I have a

2129
01:30:29,199 --> 01:30:29,760
question."

2130
01:30:29,760 --> 01:30:30,719
Scared everybody.

2131
01:30:30,719 --> 01:30:33,040
I had questions. I know. I just lost

2132
01:30:33,040 --> 01:30:36,080
them all in the glory of all of those.

2133
01:30:36,080 --> 01:30:39,760
Uh here there's a question here. Yes.

2134
01:30:39,760 --> 01:30:41,600
Uh this is a question for the networking

2135
01:30:41,600 --> 01:30:43,760
one. Um,

2136
01:30:43,760 --> 01:30:46,320
uh, have you heard of, uh, what's it

2137
01:30:46,320 --> 01:30:50,880
called? Sisters. It's a thing, a a group

2138
01:30:50,880 --> 01:30:53,040
for women to get into network

2139
01:30:53,040 --> 01:30:54,960
engineering.

2140
01:30:54,960 --> 01:30:56,320
Oh, sorry. Can you say that again? What

2141
01:30:56,320 --> 01:30:56,719
was it?

2142
01:30:56,719 --> 01:31:00,400
It's called like I I aa sisters.

2143
01:31:00,400 --> 01:31:02,080
Okay, I I'll take a look at that.

2144
01:31:02,080 --> 01:31:02,880
Check it out.

2145
01:31:02,880 --> 01:31:04,880
Yeah, thank you.

2146
01:31:04,880 --> 01:31:07,520
Excellent. I will accept that comment. I

2147
01:31:07,520 --> 01:31:09,679
will absolutely accept that comment.

2148
01:31:09,679 --> 01:31:13,440
Okay, then I think we are Oh no, there's

2149
01:31:13,440 --> 01:31:17,360
one more. There's one more.

2150
01:31:17,360 --> 01:31:19,679
There you go.

2151
01:31:19,679 --> 01:31:21,600
Somebody's sister is going to be asking

2152
01:31:21,600 --> 01:31:22,400
question.

2153
01:31:22,400 --> 01:31:24,480
Oh,

2154
01:31:24,480 --> 01:31:26,159
I mean it could be their teacher as

2155
01:31:26,159 --> 01:31:27,600
well. So, you know,

2156
01:31:27,600 --> 01:31:29,600
definitely a speaker sister.

2157
01:31:29,600 --> 01:31:30,960
Okay, go for it.

2158
01:31:30,960 --> 01:31:32,400
Yeah. [laughter]

2159
01:31:32,400 --> 01:31:35,360
Um, okay. So, this question is directed

2160
01:31:35,360 --> 01:31:38,560
towards Jasmine. You said that they

2161
01:31:38,560 --> 01:31:40,880
couldn't see like the facial expressions

2162
01:31:40,880 --> 01:31:43,199
and body language. So, how did you

2163
01:31:43,199 --> 01:31:48,000
actually get the text from episode to

2164
01:31:48,000 --> 01:31:49,679
code?

2165
01:31:49,679 --> 01:31:50,880
Yes.

2166
01:31:50,880 --> 01:31:51,920
Hello.

2167
01:31:51,920 --> 01:31:53,360
Did I turn that off? No.

2168
01:31:53,360 --> 01:31:55,920
No, I'm here. Okay. Hi. Um, basically it

2169
01:31:55,920 --> 01:31:58,159
came up in like the data set was a big

2170
01:31:58,159 --> 01:32:00,000
thing of transcripts which is basically

2171
01:32:00,000 --> 01:32:02,080
just all the words that they say. So, it

2172
01:32:02,080 --> 01:32:03,840
doesn't actually take in any of like

2173
01:32:03,840 --> 01:32:05,679
what's visually happen happening on

2174
01:32:05,679 --> 01:32:07,760
screen. So, it only like does stuff

2175
01:32:07,760 --> 01:32:09,360
based on the words they're saying. Does

2176
01:32:09,360 --> 01:32:12,159
that make sense? Yeah. Okay, cool.

2177
01:32:12,159 --> 01:32:15,120
Fantastic question. Okay,

2178
01:32:15,120 --> 01:32:17,120
there's one more very important thing to

2179
01:32:17,120 --> 01:32:18,880
do,

2180
01:32:18,880 --> 01:32:23,000
which is give out some Lego.

2181
01:32:23,120 --> 01:32:24,800
Students, you're going to awkwardly stay

2182
01:32:24,800 --> 01:32:27,280
there [laughter]

2183
01:32:27,280 --> 01:32:29,440
because you're all going to get one more

2184
01:32:29,440 --> 01:32:31,360
massive round of applause. Absolutely.

2185
01:32:31,360 --> 01:32:33,440
Oh, no. No, no, not yet. Not yet. I'm

2186
01:32:33,440 --> 01:32:36,560
sorry. I'm sorry. I was ambiguous.

2187
01:32:36,560 --> 01:32:38,960
Um, but first,

2188
01:32:38,960 --> 01:32:41,120
uh, okay, we've got two winners. One for

2189
01:32:41,120 --> 01:32:43,679
the adult Lego drawer and one for the

2190
01:32:43,679 --> 01:32:45,679
student Lego drawer. They are both the

2191
01:32:45,679 --> 01:32:48,320
same thing. Uh, but it's in honor of the

2192
01:32:48,320 --> 01:32:49,679
student showcase, we have to have one,

2193
01:32:49,679 --> 01:32:50,800
especially for the students, cuz I'd be

2194
01:32:50,800 --> 01:32:52,239
sad if a student go didn't go home with

2195
01:32:52,239 --> 01:32:55,040
Lego. So, firstly, for the adult Lego

2196
01:32:55,040 --> 01:32:59,800
winner today, it is Nicole King.

2197
01:33:00,159 --> 01:33:01,600
She might be [applause]

2198
01:33:01,600 --> 01:33:04,400
in another room. It's mine. That's okay.

2199
01:33:04,400 --> 01:33:05,199
Wow.

2200
01:33:05,199 --> 01:33:09,120
For now. And for the student winner, and

2201
01:33:09,120 --> 01:33:10,320
I'm just going to note that some of our

2202
01:33:10,320 --> 01:33:12,000
students had to go back to Touumba on

2203
01:33:12,000 --> 01:33:14,159
the bus. So, I don't know if this

2204
01:33:14,159 --> 01:33:16,400
student is part of them, but this

2205
01:33:16,400 --> 01:33:19,440
student is Arya Bray.

2206
01:33:19,440 --> 01:33:21,840
I think she's gone back to Toumba.

2207
01:33:21,840 --> 01:33:22,719
Toua.

2208
01:33:22,719 --> 01:33:24,800
Yay. [cheering]

2209
01:33:24,800 --> 01:33:26,159
We will make sure they get that. Thank

2210
01:33:26,159 --> 01:33:28,159
you, GPN. Thank you again so much for

2211
01:33:28,159 --> 01:33:30,639
sponsoring the education track.

2212
01:33:30,639 --> 01:33:32,719
Yay.

2213
01:33:32,719 --> 01:33:34,639
for being fantastic. Thank you audience

2214
01:33:34,639 --> 01:33:36,159
for being fantastic and supportive as

2215
01:33:36,159 --> 01:33:38,000
you always are. Thank you students for

2216
01:33:38,000 --> 01:33:40,619
being amazing.

2217
01:33:40,619 --> 01:33:41,679
[cheering]

2218
01:33:41,679 --> 01:33:43,440
Can I also note that several of these

2219
01:33:43,440 --> 01:33:48,239
students are GPN students? Yes.

2220
01:33:48,239 --> 01:33:50,639
Okay. One, two, three. We're going to do

2221
01:33:50,639 --> 01:33:53,760
one more big woo. Yay. I'm going to take

2222
01:33:53,760 --> 01:33:57,080
a big breath.

2223
01:33:57,650 --> 01:33:59,670
[cheering]

2224
01:34:01,280 --> 01:34:04,760
Let's go home.