1
00:00:00,000 --> 00:00:04,280
All right. Thank you.

2
00:00:01,120 --> 00:00:04,280
It is.

3
00:00:05,860 --> 00:00:10,939
[Music]

4
00:00:15,679 --> 00:00:20,240
Who's ready for some structure and

5
00:00:17,279 --> 00:00:21,760
order?

6
00:00:20,240 --> 00:00:26,519
Great. Well, you've come to the wrong

7
00:00:21,760 --> 00:00:26,519
place. It's lightning talks time.

8
00:00:27,840 --> 00:00:31,760
There we go.

9
00:00:29,679 --> 00:00:33,440
Okay. Uh for those of you who've been

10
00:00:31,760 --> 00:00:34,960
here before, uh you don't need to listen

11
00:00:33,440 --> 00:00:36,640
to this. For those of you who haven't,

12
00:00:34,960 --> 00:00:39,520
these are five minute talks that people

13
00:00:36,640 --> 00:00:41,600
have signed up for today.

14
00:00:39,520 --> 00:00:43,600
Uh they will be presenting with or

15
00:00:41,600 --> 00:00:45,440
without slides on whatever topic they

16
00:00:43,600 --> 00:00:47,200
decided was interesting and we agreed

17
00:00:45,440 --> 00:00:49,840
was interesting.

18
00:00:47,200 --> 00:00:52,160
Uh and our job is to make these go fast

19
00:00:49,840 --> 00:00:53,440
without awkwardness in between.

20
00:00:52,160 --> 00:00:55,520
Well, we we'll do our best on the

21
00:00:53,440 --> 00:00:56,320
awkwardness thing, but you know,

22
00:00:55,520 --> 00:00:59,440
lightning talks

23
00:00:56,320 --> 00:01:01,840
audible. Really, I am. They can hear me.

24
00:00:59,440 --> 00:01:02,960
Lightning talks are chaotic for myself

25
00:01:01,840 --> 00:01:04,400
and Chris. They're chaotic for our

26
00:01:02,960 --> 00:01:07,040
presenters. They're chaotic for our AV

27
00:01:04,400 --> 00:01:08,880
team, as you've already seen, but they

28
00:01:07,040 --> 00:01:13,040
are also loads and loads of fun.

29
00:01:08,880 --> 00:01:14,880
Okay? And so, our part of this is

30
00:01:13,040 --> 00:01:16,640
audience participation. We need to make

31
00:01:14,880 --> 00:01:18,640
sure that our speakers who signed up for

32
00:01:16,640 --> 00:01:21,520
five minute talks do not get any more

33
00:01:18,640 --> 00:01:23,200
time than what we let them have. Uh our

34
00:01:21,520 --> 00:01:26,479
speakers can see a timer on the front of

35
00:01:23,200 --> 00:01:28,799
the screen and Beno here uh is managing

36
00:01:26,479 --> 00:01:31,040
the timer and when we get to about 10

37
00:01:28,799 --> 00:01:33,520
seconds left he's going to start finger

38
00:01:31,040 --> 00:01:36,079
clapping silently relatively quietly so

39
00:01:33,520 --> 00:01:38,560
it doesn't drown everyone out until he

40
00:01:36,079 --> 00:01:40,880
starts applauding a bit more. And then

41
00:01:38,560 --> 00:01:43,520
when the clock fits 5:00 we are going to

42
00:01:40,880 --> 00:01:47,520
applause uporiously and cheer really

43
00:01:43,520 --> 00:01:49,280
really really really loudly

44
00:01:47,520 --> 00:01:51,119
until we say stop so we don't waste too

45
00:01:49,280 --> 00:01:54,079
much time.

46
00:01:51,119 --> 00:01:59,350
Okay. Uh, so on deck we have uh David

47
00:01:54,079 --> 00:02:04,159
Vo, but first uh Amanda Hogan.

48
00:01:59,350 --> 00:02:07,520
[Applause]

49
00:02:04,159 --> 00:02:10,239
Okay, I need screen.

50
00:02:07,520 --> 00:02:11,920
Yay. Okay, so I have in front of me I

51
00:02:10,239 --> 00:02:15,120
have three pipe cleaners. I cheated

52
00:02:11,920 --> 00:02:17,840
appallingly by cutting uh one of the cut

53
00:02:15,120 --> 00:02:20,239
pipe cleaners into two um in with two

54
00:02:17,840 --> 00:02:21,920
cuts into three pieces so that I didn't

55
00:02:20,239 --> 00:02:23,840
take a whole minute which I've just used

56
00:02:21,920 --> 00:02:25,760
to explain to you. Uh so what I'm going

57
00:02:23,840 --> 00:02:27,920
to be doing is talking about how you can

58
00:02:25,760 --> 00:02:30,720
make earrings out of everything. While

59
00:02:27,920 --> 00:02:33,360
I'm talking I'm also going to be making

60
00:02:30,720 --> 00:02:35,680
earrings out of pipe cleaners. Uh so

61
00:02:33,360 --> 00:02:37,599
when I was very young I used to I don't

62
00:02:35,680 --> 00:02:40,400
know how many are like me. I used to

63
00:02:37,599 --> 00:02:42,160
hang cherries in pears over my ears at

64
00:02:40,400 --> 00:02:44,480
Christmas lunch. Has anyone done that?

65
00:02:42,160 --> 00:02:47,200
Yeah. Okay. So, this is my first foray

66
00:02:44,480 --> 00:02:50,319
into earring making. I I believe that it

67
00:02:47,200 --> 00:02:52,480
was a craft. Um and so after that, uh my

68
00:02:50,319 --> 00:02:56,160
mom refused to give me pierced ears. So,

69
00:02:52,480 --> 00:02:59,040
I um then dangled Christmas Bbles and

70
00:02:56,160 --> 00:03:00,239
their long string over my ear. Uh I was

71
00:02:59,040 --> 00:03:02,640
told that I was not allowed to get

72
00:03:00,239 --> 00:03:04,480
pierced ears until I was 16 when it was

73
00:03:02,640 --> 00:03:07,519
uh I was allowed to go without parental

74
00:03:04,480 --> 00:03:10,720
consent. Very nice of her. Uh, and then

75
00:03:07,519 --> 00:03:14,159
when I was in year 11, my I did get

76
00:03:10,720 --> 00:03:18,080
pierced. My school um added a Coke

77
00:03:14,159 --> 00:03:20,560
machine for fundraising purposes, which

78
00:03:18,080 --> 00:03:22,800
is a terrible idea. I don't know. Uh,

79
00:03:20,560 --> 00:03:25,680
and so do you know that if you remove

80
00:03:22,800 --> 00:03:28,159
the tab from a Coke uh can uh very

81
00:03:25,680 --> 00:03:30,720
careful way, it has a hole that um can

82
00:03:28,159 --> 00:03:32,560
be threaded through a stud earring or a

83
00:03:30,720 --> 00:03:33,680
sleeper, sleepers will come back, so you

84
00:03:32,560 --> 00:03:36,799
have to know what they look like. That's

85
00:03:33,680 --> 00:03:38,159
a sleeper. Um and uh and then you can

86
00:03:36,799 --> 00:03:40,640
really annoy some private school

87
00:03:38,159 --> 00:03:43,280
teachers. Uh

88
00:03:40,640 --> 00:03:45,519
and so that was a um then I discovered

89
00:03:43,280 --> 00:03:47,840
that uh Christmas earrings that I really

90
00:03:45,519 --> 00:03:50,640
loved were things that people That's

91
00:03:47,840 --> 00:03:53,440
one. We have a butterfly everybody.

92
00:03:50,640 --> 00:03:55,360
Uh Christmas earrings were a thing that

93
00:03:53,440 --> 00:03:57,599
people paid money for and I didn't have

94
00:03:55,360 --> 00:04:00,319
any of that. Uh and it seemed like

95
00:03:57,599 --> 00:04:03,040
wasteful to me to spend money when I had

96
00:04:00,319 --> 00:04:05,200
Bbles in a box in my house. So, the

97
00:04:03,040 --> 00:04:06,879
bbles came back again. Uh,

98
00:04:05,200 --> 00:04:08,480
here.

99
00:04:06,879 --> 00:04:10,959
Thank you.

100
00:04:08,480 --> 00:04:13,599
Oh, okay. The Bbles came back again. Uh,

101
00:04:10,959 --> 00:04:15,280
so instead of being threaded over my ear

102
00:04:13,599 --> 00:04:16,880
with the strings, they were now threaded

103
00:04:15,280 --> 00:04:18,720
through with the sleeper through the

104
00:04:16,880 --> 00:04:22,000
center. Uh, it's an excellent way to

105
00:04:18,720 --> 00:04:25,360
make an earring if you are uh rushed at

106
00:04:22,000 --> 00:04:28,720
Christmas. Um, and so then I also

107
00:04:25,360 --> 00:04:30,880
discovered that you can add pom poms to

108
00:04:28,720 --> 00:04:34,639
uh sleepers. Sleepers are a very big

109
00:04:30,880 --> 00:04:36,639
star in this story. Um, and the good uh

110
00:04:34,639 --> 00:04:38,720
pom poms that are made normal size are a

111
00:04:36,639 --> 00:04:40,479
bit big, but if you use a fork, you can

112
00:04:38,720 --> 00:04:42,479
make a pom pom that is perfect earring

113
00:04:40,479 --> 00:04:44,080
size. I'm giving you all the tips today.

114
00:04:42,479 --> 00:04:47,280
You're going to go home and uh get all

115
00:04:44,080 --> 00:04:49,440
your forks out. Um, or you can have a

116
00:04:47,280 --> 00:04:52,720
very talented daughter. Uh, I have one

117
00:04:49,440 --> 00:04:55,919
of those. Um, and so she can do use um

118
00:04:52,720 --> 00:05:00,479
oven baked clay and made ramen bowls.

119
00:04:55,919 --> 00:05:02,240
Honestly, ramen bowls with um little um

120
00:05:00,479 --> 00:05:04,080
what are they called? eyelet screws that

121
00:05:02,240 --> 00:05:06,160
you can get from Bunnings. Uh so you

122
00:05:04,080 --> 00:05:10,000
could attach an earring to the center of

123
00:05:06,160 --> 00:05:12,479
that. Um and so you are now going to go

124
00:05:10,000 --> 00:05:14,400
out into the world and look at things

125
00:05:12,479 --> 00:05:16,560
differently. If you are a TAZ teacher

126
00:05:14,400 --> 00:05:18,960
like me or if you just have a lot of

127
00:05:16,560 --> 00:05:23,199
junk in your drawers, which many of you

128
00:05:18,960 --> 00:05:26,479
do nerds, um

129
00:05:23,199 --> 00:05:28,639
you can use an eyelet screw to attach to

130
00:05:26,479 --> 00:05:32,880
anything that is a kind of malleable

131
00:05:28,639 --> 00:05:35,600
plastic uh or wood and you can make

132
00:05:32,880 --> 00:05:37,440
uh like bugs. No, earrings do not have

133
00:05:35,600 --> 00:05:40,720
to match. This is a rule that uh has

134
00:05:37,440 --> 00:05:43,280
been like applied to us for too long.

135
00:05:40,720 --> 00:05:46,639
Uh, so,

136
00:05:43,280 --> 00:05:50,000
uh, if you have access to a drill, you

137
00:05:46,639 --> 00:05:52,960
can turn and a paper clip and some

138
00:05:50,000 --> 00:05:54,960
pliers, you can turn some Lego bits into

139
00:05:52,960 --> 00:05:57,199
this is a very bad angle. Uh, this was

140
00:05:54,960 --> 00:05:59,680
practiced on a flat surface. Uh, they

141
00:05:57,199 --> 00:06:01,919
they do they do exist like they and

142
00:05:59,680 --> 00:06:05,280
they're threaded. Uh, or if you're like

143
00:06:01,919 --> 00:06:08,240
me and have laser and cutters and 3D

144
00:06:05,280 --> 00:06:11,520
printers down the hall uh, and an SVG,

145
00:06:08,240 --> 00:06:14,160
you can make these. This is the this is

146
00:06:11,520 --> 00:06:17,280
the logo of GPN everybody. Uh we have

147
00:06:14,160 --> 00:06:18,960
the CEO of GPN here right today and uh

148
00:06:17,280 --> 00:06:22,160
it's a very good organization. I have

149
00:06:18,960 --> 00:06:27,160
the shirt. Uh and then

150
00:06:22,160 --> 00:06:27,160
we have two butterflies.

151
00:06:28,479 --> 00:06:31,199
They're not the most beautiful

152
00:06:29,360 --> 00:06:34,880
butterflies I've ever made. Then if you

153
00:06:31,199 --> 00:06:37,280
thread sleepers through them,

154
00:06:34,880 --> 00:06:39,280
even while your hands are shaking, you

155
00:06:37,280 --> 00:06:40,840
have earrings that I will be wearing for

156
00:06:39,280 --> 00:06:51,440
the rest of the day.

157
00:06:40,840 --> 00:06:54,479
[Applause]

158
00:06:51,440 --> 00:06:57,479
Thank you so much.

159
00:06:54,479 --> 00:06:57,479
Hello.

160
00:06:57,600 --> 00:07:00,720
Hold up the earrings.

161
00:06:59,599 --> 00:07:02,800
One. one fell.

162
00:07:00,720 --> 00:07:06,720
There we go.

163
00:07:02,800 --> 00:07:08,319
So, we've all made mistakes

164
00:07:06,720 --> 00:07:11,280
which will be relevant to our next talk.

165
00:07:08,319 --> 00:07:13,520
Uh, for the time being, Sean, if you're

166
00:07:11,280 --> 00:07:17,240
ready to come up next on this side.

167
00:07:13,520 --> 00:07:17,240
Thank you so much, Amanda.

168
00:07:19,120 --> 00:07:23,759
And we've all made mistakes. Uh, the

169
00:07:21,759 --> 00:07:25,919
AV's been a bit screwy already. I forgot

170
00:07:23,759 --> 00:07:26,880
which order this introduction goes in.

171
00:07:25,919 --> 00:07:28,240
But, we're all apparently making

172
00:07:26,880 --> 00:07:31,039
mistakes with typins. So, David's going

173
00:07:28,240 --> 00:07:31,039
to tell us about them.

174
00:07:32,240 --> 00:07:36,639
So yeah, you're you're probably not

175
00:07:34,000 --> 00:07:38,080
using typins, right? But um what what

176
00:07:36,639 --> 00:07:40,000
are typins? You might be asking. So

177
00:07:38,080 --> 00:07:41,919
there's there these things, right? These

178
00:07:40,000 --> 00:07:44,160
these are great for telling humans and

179
00:07:41,919 --> 00:07:46,400
tools what the expected types of

180
00:07:44,160 --> 00:07:48,800
everything is.

181
00:07:46,400 --> 00:07:50,800
So the most common mistake I see is just

182
00:07:48,800 --> 00:07:53,520
not using typins. Typins are great for

183
00:07:50,800 --> 00:07:56,560
documentation. You should use them.

184
00:07:53,520 --> 00:07:59,199
Sometimes people partially type in their

185
00:07:56,560 --> 00:08:01,919
functions. So what happens here is that

186
00:07:59,199 --> 00:08:06,000
the tools will assume that these can

187
00:08:01,919 --> 00:08:07,440
just take or return any type. You should

188
00:08:06,000 --> 00:08:09,919
probably explicit if this is

189
00:08:07,440 --> 00:08:12,080
intentional.

190
00:08:09,919 --> 00:08:15,680
But what's even worse than that are

191
00:08:12,080 --> 00:08:18,319
typins that are just straight up wrong.

192
00:08:15,680 --> 00:08:21,199
There's about five different actively

193
00:08:18,319 --> 00:08:22,800
developed type checkers these days. Make

194
00:08:21,199 --> 00:08:25,360
sure you're running at least one of them

195
00:08:22,800 --> 00:08:27,840
in CI.

196
00:08:25,360 --> 00:08:30,879
So, Python has containers like lists,

197
00:08:27,840 --> 00:08:35,680
right? But what's this list that I'm

198
00:08:30,879 --> 00:08:37,839
taking here of so collections like lists

199
00:08:35,680 --> 00:08:40,800
let you specify the types of values that

200
00:08:37,839 --> 00:08:42,959
they should contain. So, yeah, this is

201
00:08:40,800 --> 00:08:47,600
great. But what if I want to allow

202
00:08:42,959 --> 00:08:49,200
passing a set of integers for example?

203
00:08:47,600 --> 00:08:51,040
So the standard library does define a

204
00:08:49,200 --> 00:08:53,920
bunch of abstract collections which let

205
00:08:51,040 --> 00:08:57,040
you say that these things can take for

206
00:08:53,920 --> 00:09:00,160
example sequences which you know so

207
00:08:57,040 --> 00:09:03,120
types like lists or tpples or in this

208
00:09:00,160 --> 00:09:06,880
example uh I'd just say that uh this can

209
00:09:03,120 --> 00:09:08,880
take anything that might contain ins.

210
00:09:06,880 --> 00:09:13,440
Um, yeah. So, publishing packages to

211
00:09:08,880 --> 00:09:15,360
Pippi. Um, if you're if you have a keen

212
00:09:13,440 --> 00:09:17,760
eye, you you might notice that this

213
00:09:15,360 --> 00:09:19,920
package that I have on screen is missing

214
00:09:17,760 --> 00:09:23,279
something. This does have type hints,

215
00:09:19,920 --> 00:09:25,920
but it's missing this file. So, if you

216
00:09:23,279 --> 00:09:29,040
publish a type hinted library to Pippi,

217
00:09:25,920 --> 00:09:31,920
you have to include this marker file for

218
00:09:29,040 --> 00:09:34,080
tools to actually consume your typeins.

219
00:09:31,920 --> 00:09:38,640
And you only need one at the top level.

220
00:09:34,080 --> 00:09:40,560
So yeah, don't worry about nesting. So

221
00:09:38,640 --> 00:09:44,240
here's some common typos that you might

222
00:09:40,560 --> 00:09:46,800
be making. So here someone wants to

223
00:09:44,240 --> 00:09:49,600
return a tpple from this uh this

224
00:09:46,800 --> 00:09:52,480
function. Um you need to actually use

225
00:09:49,600 --> 00:09:54,480
the tpple type.

226
00:09:52,480 --> 00:09:58,720
Yeah.

227
00:09:54,480 --> 00:10:01,040
Um let's play a little game.

228
00:09:58,720 --> 00:10:03,200
Function or type. So here someone's

229
00:10:01,040 --> 00:10:04,480
actually tried to use the any function

230
00:10:03,200 --> 00:10:06,640
as a type.

231
00:10:04,480 --> 00:10:08,480
Oh

232
00:10:06,640 --> 00:10:11,040
yeah, the the there's a separate any

233
00:10:08,480 --> 00:10:13,040
type which you have to import and use.

234
00:10:11,040 --> 00:10:16,800
Yeah, the

235
00:10:13,040 --> 00:10:19,120
um the same same deal here. Maybe you've

236
00:10:16,800 --> 00:10:22,399
written a decorator, right? Um yeah, so

237
00:10:19,120 --> 00:10:24,720
this here is the callable function, not

238
00:10:22,399 --> 00:10:28,399
the callable type.

239
00:10:24,720 --> 00:10:33,519
Um so yeah, this is better. But um yeah,

240
00:10:28,399 --> 00:10:35,680
you can also say what uh arguments uh a

241
00:10:33,519 --> 00:10:38,320
callable can take and what will return.

242
00:10:35,680 --> 00:10:41,519
Um but yeah,

243
00:10:38,320 --> 00:10:44,079
so here someone's tried to type into

244
00:10:41,519 --> 00:10:46,000
numpy array.

245
00:10:44,079 --> 00:10:48,640
But

246
00:10:46,000 --> 00:10:52,560
if you've used numpy, you might know

247
00:10:48,640 --> 00:10:54,240
that numpy.array is a function, not a

248
00:10:52,560 --> 00:10:57,519
type.

249
00:10:54,240 --> 00:10:59,279
The type is actually called numpy. Nd

250
00:10:57,519 --> 00:11:01,760
array,

251
00:10:59,279 --> 00:11:05,040
but you probably actually want to use

252
00:11:01,760 --> 00:11:06,800
the the nd array type from numpy.typing.

253
00:11:05,040 --> 00:11:10,880
Um,

254
00:11:06,800 --> 00:11:13,920
for many reasons that uh I probably

255
00:11:10,880 --> 00:11:19,600
can't fit in 5 minutes.

256
00:11:13,920 --> 00:11:23,839
Um similarly uh PyTorch has a tensor

257
00:11:19,600 --> 00:11:27,040
type but here this function is using the

258
00:11:23,839 --> 00:11:29,360
tensor function.

259
00:11:27,040 --> 00:11:33,519
Yeah.

260
00:11:29,360 --> 00:11:37,640
So thanks everyone. Um this here is a QR

261
00:11:33,519 --> 00:11:37,640
code to the slides if you need it.

262
00:11:42,399 --> 00:11:45,360
typing is a lot better than it used to

263
00:11:44,079 --> 00:11:47,440
be, but there's still a lot of things

264
00:11:45,360 --> 00:11:48,959
that can go wrong.

265
00:11:47,440 --> 00:11:51,440
Um,

266
00:11:48,959 --> 00:11:53,680
Sam, you're up on deck on this side, but

267
00:11:51,440 --> 00:11:55,600
first, uh, we're going to hear about

268
00:11:53,680 --> 00:11:57,870
game development from Shan.

269
00:11:55,600 --> 00:12:03,980
Thank you.

270
00:11:57,870 --> 00:12:03,980
[Applause]

271
00:12:04,880 --> 00:12:09,279
So, hello everyone. Uh, I'm Sean. I'm a

272
00:12:06,959 --> 00:12:10,720
junior software engineer and I got my

273
00:12:09,279 --> 00:12:14,240
start in programming with game

274
00:12:10,720 --> 00:12:15,519
development and for me I felt that my

275
00:12:14,240 --> 00:12:17,040
journey through game development has

276
00:12:15,519 --> 00:12:18,959
laid a really strong foundation for my

277
00:12:17,040 --> 00:12:22,240
programming skills which eventually led

278
00:12:18,959 --> 00:12:24,000
me to Wolf Lander position. Um I also

279
00:12:22,240 --> 00:12:26,240
saw things go boom on screen that looked

280
00:12:24,000 --> 00:12:28,880
really cool and I wanted to uh learn how

281
00:12:26,240 --> 00:12:30,639
to make things go boom as well. Um so

282
00:12:28,880 --> 00:12:32,320
today I'm going to talk about why game

283
00:12:30,639 --> 00:12:34,320
development is an amazing way to learn

284
00:12:32,320 --> 00:12:38,560
programming. Um and yes my slides are a

285
00:12:34,320 --> 00:12:40,560
game. Um so three core reasons uh visual

286
00:12:38,560 --> 00:12:42,720
feedback touching almost every area of

287
00:12:40,560 --> 00:12:46,000
programming and solving fantastical

288
00:12:42,720 --> 00:12:47,680
problems. So number one visual feedback.

289
00:12:46,000 --> 00:12:50,639
I believe that visual feedback is

290
00:12:47,680 --> 00:12:52,800
integral to the learning process. Um

291
00:12:50,639 --> 00:12:54,320
it's a it's an indicator of like what is

292
00:12:52,800 --> 00:12:56,000
going right and what is going wrong but

293
00:12:54,320 --> 00:12:57,839
it's also something of a motivating

294
00:12:56,000 --> 00:13:00,160
force. I'm sure we've had those like

295
00:12:57,839 --> 00:13:02,000
uplifting moments where we do something

296
00:13:00,160 --> 00:13:04,000
and it works. And I'm willing to bet

297
00:13:02,000 --> 00:13:07,360
that those which had a strong visual

298
00:13:04,000 --> 00:13:10,480
reaction were the most like impactful.

299
00:13:07,360 --> 00:13:13,120
And games are very connected to this.

300
00:13:10,480 --> 00:13:15,839
They are usually well most often tech

301
00:13:13,120 --> 00:13:18,240
visual mediums. And so almost everything

302
00:13:15,839 --> 00:13:19,839
you do, anything you kind of implement

303
00:13:18,240 --> 00:13:21,920
will have some kind of visual feedback

304
00:13:19,839 --> 00:13:24,480
attached to it. Let's take something

305
00:13:21,920 --> 00:13:26,800
simple as like making the play a jump.

306
00:13:24,480 --> 00:13:28,560
Already what you're trying to do is

307
00:13:26,800 --> 00:13:31,120
reacting on the screen and we have a

308
00:13:28,560 --> 00:13:32,639
visual feedback attached to it. Then you

309
00:13:31,120 --> 00:13:34,000
can make it better, maybe vary the jump

310
00:13:32,639 --> 00:13:37,680
a little bit so it feels a little bit

311
00:13:34,000 --> 00:13:39,760
better by varying the uh gravity and uh

312
00:13:37,680 --> 00:13:41,279
jump acceleration. And you take it one

313
00:13:39,760 --> 00:13:44,720
step further by adding some animations

314
00:13:41,279 --> 00:13:47,279
and stuff. And throughout this process,

315
00:13:44,720 --> 00:13:48,320
visual feedback is always there. And so

316
00:13:47,279 --> 00:13:51,360
no matter what you do in game

317
00:13:48,320 --> 00:13:52,880
development, you will feel good about

318
00:13:51,360 --> 00:13:55,360
it, which will then help keep you

319
00:13:52,880 --> 00:13:58,000
motivated and keep you moving forward in

320
00:13:55,360 --> 00:13:59,920
the things you want to learn.

321
00:13:58,000 --> 00:14:01,680
Next is touching almost every area of

322
00:13:59,920 --> 00:14:03,120
programming. Depending on the game you

323
00:14:01,680 --> 00:14:04,480
kind of make, uh you can touch different

324
00:14:03,120 --> 00:14:05,440
areas such as databases, graphics,

325
00:14:04,480 --> 00:14:06,959
rendering, networking, hardware

326
00:14:05,440 --> 00:14:10,240
integration, low level GPU optimization,

327
00:14:06,959 --> 00:14:11,600
just to name a few. Um

328
00:14:10,240 --> 00:14:14,800
games at the end of the day are just

329
00:14:11,600 --> 00:14:18,240
software. Um and they're also not bound

330
00:14:14,800 --> 00:14:21,120
by any particular framework or tool. In

331
00:14:18,240 --> 00:14:23,760
fact, technically anything can be a game

332
00:14:21,120 --> 00:14:25,279
as long as it's well a game, an

333
00:14:23,760 --> 00:14:26,959
interactive system of rules and goals

334
00:14:25,279 --> 00:14:28,720
that engages the player in a playful

335
00:14:26,959 --> 00:14:30,800
manner.

336
00:14:28,720 --> 00:14:32,720
The game, the idea of making a game just

337
00:14:30,800 --> 00:14:34,399
gives you the kind of structure you need

338
00:14:32,720 --> 00:14:36,320
to keep yourself motivated and to give

339
00:14:34,399 --> 00:14:38,000
you direction.

340
00:14:36,320 --> 00:14:41,360
For example, you can say, I want to

341
00:14:38,000 --> 00:14:43,040
learn SQL. Um, which

342
00:14:41,360 --> 00:14:44,240
is a little bit daunting. Um, it's you

343
00:14:43,040 --> 00:14:46,959
don't really know what you want to do

344
00:14:44,240 --> 00:14:48,480
with SQL and it's very vague. You can

345
00:14:46,959 --> 00:14:50,480
instead say, "I want to learn SQL to

346
00:14:48,480 --> 00:14:52,720
manage my game's inventory system and

347
00:14:50,480 --> 00:14:54,160
database." You now have a way to kind of

348
00:14:52,720 --> 00:14:56,160
apply what you want to learn and

349
00:14:54,160 --> 00:14:57,920
direction and hopefully make something

350
00:14:56,160 --> 00:14:59,440
cool and visual out of it in the end as

351
00:14:57,920 --> 00:15:00,720
well.

352
00:14:59,440 --> 00:15:02,800
So, if you always had that kind of one

353
00:15:00,720 --> 00:15:04,079
thing you wanted to learn, maybe make a

354
00:15:02,800 --> 00:15:06,079
game out of it or incorporate in the

355
00:15:04,079 --> 00:15:08,240
game in the game in some way as a

356
00:15:06,079 --> 00:15:09,680
motivating factor.

357
00:15:08,240 --> 00:15:12,320
And now, finally, my favorite part,

358
00:15:09,680 --> 00:15:15,760
solving fantastical problems. Games are

359
00:15:12,320 --> 00:15:18,639
not bound by the chains of reality.

360
00:15:15,760 --> 00:15:20,800
You can basically do whatever you want.

361
00:15:18,639 --> 00:15:22,800
Uh, I mean, how often do you get to

362
00:15:20,800 --> 00:15:25,279
solve problems like, "How do I flood a

363
00:15:22,800 --> 00:15:28,000
city with ducks?

364
00:15:25,279 --> 00:15:32,480
How do I build skyscrapers out of cake?"

365
00:15:28,000 --> 00:15:34,480
"How do I make a pyon talk into a game?"

366
00:15:32,480 --> 00:15:38,399
Games can range from being wacky to

367
00:15:34,480 --> 00:15:39,760
realistic. And by that nature, you will

368
00:15:38,399 --> 00:15:44,440
exercise your problem, your problem

369
00:15:39,760 --> 00:15:44,440
solving skills quite extensively.

370
00:15:44,560 --> 00:15:48,720
So maybe my ramblings have kind of

371
00:15:47,040 --> 00:15:50,959
inspired you to maybe also start making

372
00:15:48,720 --> 00:15:53,519
games yourself. But where do you get

373
00:15:50,959 --> 00:15:55,440
started? Well, as mentioned before, you

374
00:15:53,519 --> 00:15:57,199
can use what you already have. Whatever

375
00:15:55,440 --> 00:15:59,279
framework you want to learn, whatever

376
00:15:57,199 --> 00:16:01,839
kind of thing you want to incorporate,

377
00:15:59,279 --> 00:16:03,440
conceptualize it in a game, and you can

378
00:16:01,839 --> 00:16:05,120
kind of go from there. You also don't

379
00:16:03,440 --> 00:16:07,839
need to worry about graphics rendering

380
00:16:05,120 --> 00:16:09,839
as much. Uh some simple asy art or some

381
00:16:07,839 --> 00:16:12,399
formatted text on a CLI can go a long

382
00:16:09,839 --> 00:16:14,320
way. For more Python related things,

383
00:16:12,399 --> 00:16:16,240
Pygame is an excellent choice. It will

384
00:16:14,320 --> 00:16:18,000
handle your graphics rendering, uh,

385
00:16:16,240 --> 00:16:20,399
audio handling and input handling as

386
00:16:18,000 --> 00:16:22,720
well. Um, and otherwise you want to go

387
00:16:20,399 --> 00:16:23,920
one step further, you have game engines,

388
00:16:22,720 --> 00:16:25,440
um, which will also handle stuff like

389
00:16:23,920 --> 00:16:27,839
physics, collisions, particles, and

390
00:16:25,440 --> 00:16:29,680
other things that might be useful in the

391
00:16:27,839 --> 00:16:31,920
game. Um, I highly recommend the GDO

392
00:16:29,680 --> 00:16:34,000
game engine because it's, uh, scripting

393
00:16:31,920 --> 00:16:36,560
language is almost onetoone with Python,

394
00:16:34,000 --> 00:16:38,560
and it's also open source.

395
00:16:36,560 --> 00:16:39,920
So games are an amazing way to learn

396
00:16:38,560 --> 00:16:41,199
programming because they provide visual

397
00:16:39,920 --> 00:16:43,279
feedback within the development process

398
00:16:41,199 --> 00:16:44,560
itself. You touch almost every area of

399
00:16:43,279 --> 00:16:47,680
programming and you get to solve

400
00:16:44,560 --> 00:16:50,839
fantastical problems. With that said, go

401
00:16:47,680 --> 00:16:50,839
make games.

402
00:16:57,120 --> 00:17:00,920
Thank you so much, Sean.

403
00:17:00,959 --> 00:17:05,919
Some of the best lightning talks have

404
00:17:04,000 --> 00:17:07,280
information as well as a fun little fun

405
00:17:05,919 --> 00:17:09,600
little trick up their sleeve. And as

406
00:17:07,280 --> 00:17:11,039
soon as we saw

407
00:17:09,600 --> 00:17:12,400
I'm going to tell talk about gamedev,

408
00:17:11,039 --> 00:17:14,720
but the slides are a game. We had to

409
00:17:12,400 --> 00:17:17,600
accept it. Up on this side next, we have

410
00:17:14,720 --> 00:17:19,280
Ada. But first,

411
00:17:17,600 --> 00:17:22,520
Sam Bishop is going to take us to a land

412
00:17:19,280 --> 00:17:22,520
down thunder.

413
00:17:22,880 --> 00:17:27,600
It wouldn't be a conference without a

414
00:17:24,559 --> 00:17:29,120
pardon. So yes uh for anyone who hasn't

415
00:17:27,600 --> 00:17:30,320
worked this out at some point when

416
00:17:29,120 --> 00:17:31,919
Python programmers are talking about

417
00:17:30,320 --> 00:17:34,559
dunder we're talking about these double

418
00:17:31,919 --> 00:17:36,559
underscore things and uh you might have

419
00:17:34,559 --> 00:17:37,679
come across some of them uh but I'm

420
00:17:36,559 --> 00:17:39,200
going to cover these ones first. These

421
00:17:37,679 --> 00:17:40,799
are not dunder methods. Do not get them

422
00:17:39,200 --> 00:17:43,120
confused or ask me why I didn't include

423
00:17:40,799 --> 00:17:44,640
them. Name main and a few others. They

424
00:17:43,120 --> 00:17:46,000
are not them. I'm not going to bother.

425
00:17:44,640 --> 00:17:48,720
I'm moving on cuz I've got a hundred of

426
00:17:46,000 --> 00:17:49,919
these to cover such as in it which is

427
00:17:48,720 --> 00:17:51,679
something most people have covered

428
00:17:49,919 --> 00:17:53,120
pretty early on in learning Python. It's

429
00:17:51,679 --> 00:17:54,880
how you make classes. I'm not going to

430
00:17:53,120 --> 00:17:56,240
keep explaining. Then there's repper

431
00:17:54,880 --> 00:17:58,080
where you start having to go, "Hey,

432
00:17:56,240 --> 00:17:59,360
what's this object thing? I'd like

433
00:17:58,080 --> 00:18:00,799
something more useful and you make a

434
00:17:59,360 --> 00:18:02,640
custom version to make that easier for

435
00:18:00,799 --> 00:18:04,640
yourself." Then you get to the ones that

436
00:18:02,640 --> 00:18:06,160
people use a lot for type casting, like

437
00:18:04,640 --> 00:18:08,799
when you need to have a custom string,

438
00:18:06,160 --> 00:18:11,679
int, float, or lesser often complex or

439
00:18:08,799 --> 00:18:13,760
boolean output for what your class does.

440
00:18:11,679 --> 00:18:15,840
And then there's bytes and uh the much

441
00:18:13,760 --> 00:18:17,200
much more infrequently used index, which

442
00:18:15,840 --> 00:18:20,559
is actually called when you're trying to

443
00:18:17,200 --> 00:18:22,400
use it as a reference for indexing. Then

444
00:18:20,559 --> 00:18:23,919
there's the math ones like comparison.

445
00:18:22,400 --> 00:18:25,760
So we've got our greater than, our less

446
00:18:23,919 --> 00:18:27,440
than, our greater than or equal to, our

447
00:18:25,760 --> 00:18:29,039
less than or equal to, are equal to, and

448
00:18:27,440 --> 00:18:30,720
our not equal to. All of these are

449
00:18:29,039 --> 00:18:32,559
hidden behind dunder methods. If you

450
00:18:30,720 --> 00:18:35,520
want to mess with how Python treats

451
00:18:32,559 --> 00:18:36,559
them, this is how you do it. And then

452
00:18:35,520 --> 00:18:38,960
there's the same thing for all the

453
00:18:36,559 --> 00:18:41,600
arithmetic operators. Now, that's a big

454
00:18:38,960 --> 00:18:43,760
old list of 8 by3 and they are all just

455
00:18:41,600 --> 00:18:46,160
variations on the same thing. We have

456
00:18:43,760 --> 00:18:48,160
our add, our right ad, and our in place

457
00:18:46,160 --> 00:18:49,679
ad, which are is my thing on the left,

458
00:18:48,160 --> 00:18:51,679
my thing on the right, or am I doing it

459
00:18:49,679 --> 00:18:53,280
with the weird in place operators? And

460
00:18:51,679 --> 00:18:55,039
that's just our addition, subtraction,

461
00:18:53,280 --> 00:18:56,559
multiplication, division, modulo, into

462
00:18:55,039 --> 00:18:58,400
division, exponentiation, and matrix

463
00:18:56,559 --> 00:19:00,080
multiplication. And then we have our

464
00:18:58,400 --> 00:19:02,000
boolean arithmetic operators. So we've

465
00:19:00,080 --> 00:19:05,039
got our left shift, right shift, and or

466
00:19:02,000 --> 00:19:06,640
exor. Same deal, same kind of pattern.

467
00:19:05,039 --> 00:19:08,400
We've got our value conversions.

468
00:19:06,640 --> 00:19:10,880
Absolute value, negative value, positive

469
00:19:08,400 --> 00:19:13,120
value, and invert along with our

470
00:19:10,880 --> 00:19:15,200
truncations. Truncate, floor, ceiling,

471
00:19:13,120 --> 00:19:16,960
and round. Now that I've finally gotten

472
00:19:15,200 --> 00:19:18,480
rid of all of the math except one, we've

473
00:19:16,960 --> 00:19:20,480
got the weird little side call that is

474
00:19:18,480 --> 00:19:22,240
divod, which most people never use.

475
00:19:20,480 --> 00:19:24,000
they'll just kind of do it twice. But

476
00:19:22,240 --> 00:19:26,960
that gives you your output in terms of

477
00:19:24,000 --> 00:19:28,240
your integer division and the overflow.

478
00:19:26,960 --> 00:19:30,080
Then we've got the fun stuff starting

479
00:19:28,240 --> 00:19:33,280
now. We've got our context managers. So

480
00:19:30,080 --> 00:19:35,120
enter and exit as was a enter and a exit

481
00:19:33,280 --> 00:19:37,039
are how the context managers work both

482
00:19:35,120 --> 00:19:38,640
in synchronous and asynchronous style.

483
00:19:37,039 --> 00:19:40,480
If you customize these, you can do your

484
00:19:38,640 --> 00:19:42,640
own fancy setup logic like spin up a

485
00:19:40,480 --> 00:19:44,320
database for a test or tear it down when

486
00:19:42,640 --> 00:19:45,679
you're done. I've written async managers

487
00:19:44,320 --> 00:19:47,840
that spin up whole Docker environments

488
00:19:45,679 --> 00:19:51,240
in order to run unit tests. These things

489
00:19:47,840 --> 00:19:51,240
are pretty powerful.

490
00:19:53,039 --> 00:19:57,200
They're very very powerful.

491
00:19:55,600 --> 00:19:58,720
You've got the container type methods

492
00:19:57,200 --> 00:20:00,320
which most people again don't tend to

493
00:19:58,720 --> 00:20:01,919
play with too much except for when

494
00:20:00,320 --> 00:20:03,600
they're playing with fake dictionaries.

495
00:20:01,919 --> 00:20:05,440
So you've got your get item, set item,

496
00:20:03,600 --> 00:20:06,880
delete item. You've got the reversed if

497
00:20:05,440 --> 00:20:08,640
you want to customize reversing the

498
00:20:06,880 --> 00:20:10,880
list. Maybe it's a bad idea to reverse

499
00:20:08,640 --> 00:20:12,480
the list in memory for say 10 million

500
00:20:10,880 --> 00:20:14,000
items or something. And you've got

501
00:20:12,480 --> 00:20:16,080
contains which for the same reason maybe

502
00:20:14,000 --> 00:20:17,919
you don't want to check that whole list.

503
00:20:16,080 --> 00:20:19,280
You've got length and length hint, which

504
00:20:17,919 --> 00:20:20,640
again to use the same theme here, maybe

505
00:20:19,280 --> 00:20:22,160
if you've got 10 million items in your

506
00:20:20,640 --> 00:20:23,840
list, you don't actually want to count

507
00:20:22,160 --> 00:20:25,919
them. You might want to do a smart

508
00:20:23,840 --> 00:20:27,520
algorithm kind of guess. And length hint

509
00:20:25,919 --> 00:20:29,280
is how you can do that quickly. You've

510
00:20:27,520 --> 00:20:30,960
got missing, and that's used when it's

511
00:20:29,280 --> 00:20:32,640
like, hey, I didn't find it here. What

512
00:20:30,960 --> 00:20:34,000
do I do now? And then you've got it,

513
00:20:32,640 --> 00:20:36,240
which walks through them all one at a

514
00:20:34,000 --> 00:20:37,919
time. And the asynchronous aitter.

515
00:20:36,240 --> 00:20:39,039
You've also got next and a next, which

516
00:20:37,919 --> 00:20:41,280
speaking of iterators are how the

517
00:20:39,039 --> 00:20:43,440
iterator protocol works. They basically

518
00:20:41,280 --> 00:20:45,200
just spit out things one by one. You've

519
00:20:43,440 --> 00:20:47,360
got get attribute, get adder, set adder,

520
00:20:45,200 --> 00:20:49,120
and dell adder, which are how you do the

521
00:20:47,360 --> 00:20:51,360
attributes on a class. If you want to

522
00:20:49,120 --> 00:20:53,039
delete a thing on a class, the dot

523
00:20:51,360 --> 00:20:55,360
something, that's how you do it. As well

524
00:20:53,039 --> 00:20:57,440
as setting new ones, getting ones, and

525
00:20:55,360 --> 00:20:58,880
the get attribute is called all the time

526
00:20:57,440 --> 00:21:00,400
regardless. So, you can get very

527
00:20:58,880 --> 00:21:02,640
powerfully into messing with other

528
00:21:00,400 --> 00:21:04,720
classes and subasses that way. You've

529
00:21:02,640 --> 00:21:05,919
got new and get new arags, which are

530
00:21:04,720 --> 00:21:08,559
actually controlling how you create

531
00:21:05,919 --> 00:21:10,080
classes. Then you've got instance check,

532
00:21:08,559 --> 00:21:11,440
subclass check, and class item get

533
00:21:10,080 --> 00:21:14,159
check, which are actually able to mess

534
00:21:11,440 --> 00:21:16,480
with the multi uh metaroming. Again,

535
00:21:14,159 --> 00:21:18,080
still same thing for set name and init

536
00:21:16,480 --> 00:21:19,520
subclass because why wouldn't you want

537
00:21:18,080 --> 00:21:21,440
to mess with the subasses someone else

538
00:21:19,520 --> 00:21:24,159
makes from your code? Then you've got

539
00:21:21,440 --> 00:21:25,600
prepare and init subclass in here, which

540
00:21:24,159 --> 00:21:27,840
again more messing with your children

541
00:21:25,600 --> 00:21:30,640
and the code that way. This stuff

542
00:21:27,840 --> 00:21:32,240
reduces only normally used in making uh

543
00:21:30,640 --> 00:21:33,760
pickles because you don't normally sort

544
00:21:32,240 --> 00:21:35,679
of don't know what you're passing in,

545
00:21:33,760 --> 00:21:37,679
but it's there for that. Format is how

546
00:21:35,679 --> 00:21:40,240
you change what Python does when you run

547
00:21:37,679 --> 00:21:42,080
string format on things. Uh, again, not

548
00:21:40,240 --> 00:21:43,679
usually the way you do things. Hash is

549
00:21:42,080 --> 00:21:45,919
how you make things hashable for use in

550
00:21:43,679 --> 00:21:47,280
sets and dictionary lookups. Size of is

551
00:21:45,919 --> 00:21:49,200
how you pretend that you're smaller than

552
00:21:47,280 --> 00:21:51,200
you really are. A wait is how you

553
00:21:49,200 --> 00:21:52,880
actually say, "No, I swear I'm going to

554
00:21:51,200 --> 00:21:55,039
get back to you later in an async

555
00:21:52,880 --> 00:21:56,799
context. Call is how you pretend to be a

556
00:21:55,039 --> 00:21:58,880
function even though you're a class."

557
00:21:56,799 --> 00:22:00,480
And finally, dur is actually how you

558
00:21:58,880 --> 00:22:02,400
list off all the attributes that are

559
00:22:00,480 --> 00:22:04,559
available inside your object or just

560
00:22:02,400 --> 00:22:06,720
lie.

561
00:22:04,559 --> 00:22:08,510
And that is the hundred of them that I

562
00:22:06,720 --> 00:22:11,760
was trying to get through.

563
00:22:08,510 --> 00:22:15,000
[Applause]

564
00:22:11,760 --> 00:22:15,000
Thank you.

565
00:22:15,440 --> 00:22:20,640
Thank you, Sam. Uh that talk has will

566
00:22:18,720 --> 00:22:22,559
certainly help me and many of my other

567
00:22:20,640 --> 00:22:25,880
colleagues in this room be better men at

568
00:22:22,559 --> 00:22:25,880
work in Python.

569
00:22:26,880 --> 00:22:31,919
Hey uh Jack, you're up on uh on the

570
00:22:29,360 --> 00:22:33,440
other lect turn. Uh but first uh I'm

571
00:22:31,919 --> 00:22:36,320
sure there's a bunch of undergrads in

572
00:22:33,440 --> 00:22:38,559
this room. Um if you want the rest of

573
00:22:36,320 --> 00:22:40,720
your year at uni to be more like

574
00:22:38,559 --> 00:22:43,120
lightning talks. Uh Ada is going to help

575
00:22:40,720 --> 00:22:45,440
you.

576
00:22:43,120 --> 00:22:46,880
Cool. Hello. My name is Adah and today

577
00:22:45,440 --> 00:22:48,559
I'm going to tell you the story of how

578
00:22:46,880 --> 00:22:49,840
my friend and I use Python to get

579
00:22:48,559 --> 00:22:51,200
everything we needed to know from our

580
00:22:49,840 --> 00:22:53,440
university lectures without ever

581
00:22:51,200 --> 00:22:55,919
watching them. So here's my

582
00:22:53,440 --> 00:22:57,760
co-conspirator Genie. Together we made

583
00:22:55,919 --> 00:22:59,760
Compendio, a tool for students to make

584
00:22:57,760 --> 00:23:01,919
the most out of their lecture videos. So

585
00:22:59,760 --> 00:23:04,000
you give it a link and instantly get

586
00:23:01,919 --> 00:23:06,480
structured summaries, ask questions

587
00:23:04,000 --> 00:23:08,640
live, like generate revision qu

588
00:23:06,480 --> 00:23:10,480
questions and dig deeper into the parts

589
00:23:08,640 --> 00:23:12,000
you really want to know about. No more

590
00:23:10,480 --> 00:23:14,960
passive learning on two times speed

591
00:23:12,000 --> 00:23:17,679
again. So how did this happen? Now let's

592
00:23:14,960 --> 00:23:19,200
go back to 2021. We were two university

593
00:23:17,679 --> 00:23:21,360
students learning how to software

594
00:23:19,200 --> 00:23:23,919
engineer and as a right of passage we

595
00:23:21,360 --> 00:23:25,440
discovered this adage. Never spend six

596
00:23:23,919 --> 00:23:27,919
minutes doing something by hand that you

597
00:23:25,440 --> 00:23:29,440
can spend six hours failing to automate.

598
00:23:27,919 --> 00:23:31,039
So,

599
00:23:29,440 --> 00:23:32,799
setting the scene, picture this.

600
00:23:31,039 --> 00:23:34,400
University is fully online. We had

601
00:23:32,799 --> 00:23:36,240
several hours of lecture recordings to

602
00:23:34,400 --> 00:23:38,000
catch up on, tutorials and labs, which

603
00:23:36,240 --> 00:23:40,400
is where the real learning happens, a

604
00:23:38,000 --> 00:23:42,400
group assignment due that weekend, and a

605
00:23:40,400 --> 00:23:46,480
global pandemic. So, we're faced with a

606
00:23:42,400 --> 00:23:48,320
choice. Do I work on my assignment, or

607
00:23:46,480 --> 00:23:51,039
sign up for the 24-hour university

608
00:23:48,320 --> 00:23:52,799
hackathon that same weekend? And then we

609
00:23:51,039 --> 00:23:54,960
have, can the summarize my lecture was

610
00:23:52,799 --> 00:23:57,039
born? So the hackathon theme was

611
00:23:54,960 --> 00:23:59,039
education and we wanted to summarize

612
00:23:57,039 --> 00:24:01,120
insights you can't get just by skimming

613
00:23:59,039 --> 00:24:03,280
through lecture slides. So we decided to

614
00:24:01,120 --> 00:24:06,400
focus on the interaction behind between

615
00:24:03,280 --> 00:24:08,720
lecturers and students. So we focused on

616
00:24:06,400 --> 00:24:11,039
a specific problem. Could we match

617
00:24:08,720 --> 00:24:13,279
questions asked by students in the live

618
00:24:11,039 --> 00:24:15,520
chats with real-time answers by the

619
00:24:13,279 --> 00:24:17,919
lecturer from the recording? So we used

620
00:24:15,520 --> 00:24:19,360
a heristic. We noticed that lecturers

621
00:24:17,919 --> 00:24:21,120
would usually mention that they had a

622
00:24:19,360 --> 00:24:23,279
look at a chat or so and so had a good

623
00:24:21,120 --> 00:24:25,279
question. repeat the question from chat

624
00:24:23,279 --> 00:24:26,640
and answer it immediately after. And

625
00:24:25,279 --> 00:24:28,799
that's how we built our first

626
00:24:26,640 --> 00:24:31,120
summarizer. We matched on a trigger word

627
00:24:28,799 --> 00:24:33,039
like question. We used fuzzy wuzzy.

628
00:24:31,120 --> 00:24:34,480
Yeah, that's the name of the library to

629
00:24:33,039 --> 00:24:36,400
string match the next sentence from the

630
00:24:34,480 --> 00:24:38,480
lecturer like a boss. That's in the

631
00:24:36,400 --> 00:24:40,240
official docs um against chat messages

632
00:24:38,480 --> 00:24:42,320
from students. And then we just asserted

633
00:24:40,240 --> 00:24:45,039
that whatever came next is the answer.

634
00:24:42,320 --> 00:24:46,640
It worked pretty well. So then during

635
00:24:45,039 --> 00:24:49,200
our exam period, we were faced with

636
00:24:46,640 --> 00:24:52,240
another choice. Do we study for exams

637
00:24:49,200 --> 00:24:55,120
with our new tool or write a 5,000word

638
00:24:52,240 --> 00:24:56,640
blog post for fun? Um, and there we have

639
00:24:55,120 --> 00:24:58,159
summarize my hackathon. So, you can

640
00:24:56,640 --> 00:25:00,960
check out my website for the full

641
00:24:58,159 --> 00:25:02,960
hackathon story. But now we must address

642
00:25:00,960 --> 00:25:06,080
the elephant in the room. You fast

643
00:25:02,960 --> 00:25:08,640
forward 2 years, 2023, and you guessed

644
00:25:06,080 --> 00:25:10,880
it, LLMs came out seemingly out of

645
00:25:08,640 --> 00:25:13,200
nowhere. So, OpenAI released its first

646
00:25:10,880 --> 00:25:15,039
API for developers. And at this point, I

647
00:25:13,200 --> 00:25:16,960
wasn't even in university, but I still

648
00:25:15,039 --> 00:25:19,120
loved learning. And also, I kept getting

649
00:25:16,960 --> 00:25:20,799
these weird videos from my friends, like

650
00:25:19,120 --> 00:25:22,240
the dark side of home building, what

651
00:25:20,799 --> 00:25:23,679
they don't want you to know. I kind of

652
00:25:22,240 --> 00:25:27,600
wanted to know it, but not enough to

653
00:25:23,679 --> 00:25:29,600
watch a 32-minute video on it. So,

654
00:25:27,600 --> 00:25:31,600
so what do we do instead? Oh, yeah.

655
00:25:29,600 --> 00:25:33,600
We'll bring back summarize my lecture.

656
00:25:31,600 --> 00:25:35,760
So, Genie and I decided to extend the

657
00:25:33,600 --> 00:25:37,520
product. Uh, we use fast API, lang

658
00:25:35,760 --> 00:25:40,080
chain, and versel and learned a bunch of

659
00:25:37,520 --> 00:25:42,480
new concepts like vector databases, rag,

660
00:25:40,080 --> 00:25:44,799
really cool convers, and conversations.

661
00:25:42,480 --> 00:25:46,880
And we made compendio. So, put in that

662
00:25:44,799 --> 00:25:48,880
long YouTube video and get a list of all

663
00:25:46,880 --> 00:25:51,360
the non-compliance issues mentioned in

664
00:25:48,880 --> 00:25:53,039
the other video about home renovations.

665
00:25:51,360 --> 00:25:54,720
Um, and we also tried it with like

666
00:25:53,039 --> 00:25:56,480
two-hour public lectures like

667
00:25:54,720 --> 00:25:57,840
introduction to supply and demand. You

668
00:25:56,480 --> 00:26:00,880
know, kind of just want to learn for

669
00:25:57,840 --> 00:26:02,240
fun. Um, so the other thing we did was

670
00:26:00,880 --> 00:26:03,760
well this is actually pretty cool. So,

671
00:26:02,240 --> 00:26:06,000
we decided to put it in front of users

672
00:26:03,760 --> 00:26:07,520
and deploy the site. Cost money but get

673
00:26:06,000 --> 00:26:09,679
people to try it out. And then we

674
00:26:07,520 --> 00:26:11,840
watched as people used the website and

675
00:26:09,679 --> 00:26:13,440
clicked around. Um, so it was really

676
00:26:11,840 --> 00:26:15,279
cool because we got an iterative process

677
00:26:13,440 --> 00:26:17,760
of improving the tool. So we noticed

678
00:26:15,279 --> 00:26:19,279
that people wanted to like go to that

679
00:26:17,760 --> 00:26:21,520
part of the video when they get answers.

680
00:26:19,279 --> 00:26:23,679
So we added uh structured summaries with

681
00:26:21,520 --> 00:26:25,120
timestamps. Um, the other thing that I

682
00:26:23,679 --> 00:26:27,279
found really fun was like trying to make

683
00:26:25,120 --> 00:26:29,360
it faster. So LMS weren't that fast at

684
00:26:27,279 --> 00:26:31,520
the time. We started with cing. So

685
00:26:29,360 --> 00:26:32,960
videos are stable. Um, if you're

686
00:26:31,520 --> 00:26:34,880
watching lectures, there's like bunch of

687
00:26:32,960 --> 00:26:37,360
people watching the same video. So we c

688
00:26:34,880 --> 00:26:40,640
the summaries and we c like transcribing

689
00:26:37,360 --> 00:26:43,360
some the uh transcript. Uh we also

690
00:26:40,640 --> 00:26:45,039
stream chat messages um like instead of

691
00:26:43,360 --> 00:26:47,039
waiting for the response just like give

692
00:26:45,039 --> 00:26:48,159
it as as it comes. And then we also

693
00:26:47,039 --> 00:26:50,640
realized that you could anchor

694
00:26:48,159 --> 00:26:52,880
expectations. So I replaced the loading

695
00:26:50,640 --> 00:26:54,960
symbol with watching video and two times

696
00:26:52,880 --> 00:26:56,559
speed and suddenly 20 seconds to

697
00:26:54,960 --> 00:26:58,960
summarize a video is really fast

698
00:26:56,559 --> 00:27:01,120
compared to like two times speed 2our

699
00:26:58,960 --> 00:27:02,880
video. Um I also learned about mirrors

700
00:27:01,120 --> 00:27:04,880
and elevators. So when people complained

701
00:27:02,880 --> 00:27:06,400
about the speed of elevators, designers

702
00:27:04,880 --> 00:27:08,080
realized you could just put mirrors in

703
00:27:06,400 --> 00:27:10,720
there. you check your hair and bam,

704
00:27:08,080 --> 00:27:12,480
you're on level 13. Um, so we added some

705
00:27:10,720 --> 00:27:14,880
distractions like what questions can you

706
00:27:12,480 --> 00:27:16,960
ask beyond summaries and had other ideas

707
00:27:14,880 --> 00:27:19,279
like showing uh other videos that were

708
00:27:16,960 --> 00:27:21,039
summarized. So it's now been 2 years

709
00:27:19,279 --> 00:27:22,720
since we built Compendio. Um, it's now

710
00:27:21,039 --> 00:27:24,480
in archive mode and I learned a lot from

711
00:27:22,720 --> 00:27:26,720
building it more than watching the

712
00:27:24,480 --> 00:27:31,000
equivalent of lectures um than the time

713
00:27:26,720 --> 00:27:31,000
spent. Uh, thank you very much.

714
00:27:36,240 --> 00:27:39,840
Thank you so much, Anna, for carrying on

715
00:27:37,919 --> 00:27:42,559
the tradition of avoiding a little bit

716
00:27:39,840 --> 00:27:43,919
of work by doing way more.

717
00:27:42,559 --> 00:27:46,720
Uh, up next on this side, we have

718
00:27:43,919 --> 00:27:48,559
Justin. Uh, but for the time being, uh,

719
00:27:46,720 --> 00:27:51,200
last year, Jack Vichelt managed to

720
00:27:48,559 --> 00:27:53,919
squeeze 18 animal facts into 5 minutes.

721
00:27:51,200 --> 00:27:57,240
Uh, this year, he's set himself a more

722
00:27:53,919 --> 00:27:57,240
modest goal.

723
00:28:00,000 --> 00:28:03,840
All right. Uh, so I'm just going to talk

724
00:28:01,760 --> 00:28:05,919
about just the one cursed animal fact.

725
00:28:03,840 --> 00:28:09,120
Uh, yeah, as said last year, I had more

726
00:28:05,919 --> 00:28:11,840
facts. Um, I have a favorite fish. This

727
00:28:09,120 --> 00:28:13,600
is not the cursed fact. Uh, there are

728
00:28:11,840 --> 00:28:15,200
lots of good fish. There are plenty of

729
00:28:13,600 --> 00:28:17,039
good options for what your favorite fish

730
00:28:15,200 --> 00:28:19,679
might be or what my favorite fish might

731
00:28:17,039 --> 00:28:21,520
be. Is it the hammerhead shark? Uh, no.

732
00:28:19,679 --> 00:28:23,440
Very cool. Not my favorite. How about

733
00:28:21,520 --> 00:28:24,960
the bareyed fish? Those green things,

734
00:28:23,440 --> 00:28:27,679
those are its eyes. They look up through

735
00:28:24,960 --> 00:28:29,679
the transparent head. Uh, very cool. Not

736
00:28:27,679 --> 00:28:31,840
my favorite fish. Maybe it's one of

737
00:28:29,679 --> 00:28:34,880
these fish. They're good fish, but

738
00:28:31,840 --> 00:28:37,520
they're not the best. My favorite fish

739
00:28:34,880 --> 00:28:40,640
is really smart. It has interesting

740
00:28:37,520 --> 00:28:43,760
social habits. It displays evidence of

741
00:28:40,640 --> 00:28:47,520
use. I like it enough that I even have

742
00:28:43,760 --> 00:28:50,399
one at home. I like it so much that a

743
00:28:47,520 --> 00:28:51,679
year ago, I paid a professional about a

744
00:28:50,399 --> 00:28:54,559
year ago, I paid a professional

745
00:28:51,679 --> 00:28:57,039
photographer to take a bunch of photos

746
00:28:54,559 --> 00:29:00,240
of me and my favorite fish. Here is me

747
00:28:57,039 --> 00:29:00,240
and my fish.

748
00:29:01,360 --> 00:29:08,640
What am I a fish? Um, yes.

749
00:29:05,360 --> 00:29:11,640
Sorry. Uh, let's learn about phlogenetic

750
00:29:08,640 --> 00:29:11,640
bracketing.

751
00:29:12,559 --> 00:29:17,039
Stop. It's V species A belongs to a

752
00:29:14,880 --> 00:29:19,919
group. Uh, phoggenetic bracketing is a

753
00:29:17,039 --> 00:29:22,480
tool used in bioinformatics and and a

754
00:29:19,919 --> 00:29:24,000
bunch of other fields. And basically, it

755
00:29:22,480 --> 00:29:25,679
can be used to identify a bunch of

756
00:29:24,000 --> 00:29:28,320
different traits or categorize different

757
00:29:25,679 --> 00:29:31,520
things. If species A belongs to a group

758
00:29:28,320 --> 00:29:33,760
and species B belongs to the same group,

759
00:29:31,520 --> 00:29:36,240
then their earliest common ancestor is

760
00:29:33,760 --> 00:29:39,600
also a part of that group. You cannot

761
00:29:36,240 --> 00:29:42,159
evolve out of a group and so all their

762
00:29:39,600 --> 00:29:44,960
descendants of that common ancestor are

763
00:29:42,159 --> 00:29:48,000
also in that group. This means that if

764
00:29:44,960 --> 00:29:51,039
the hammerhead shark is a fish and the

765
00:29:48,000 --> 00:29:52,720
barrel fish is a fish, we can look at

766
00:29:51,039 --> 00:29:55,039
this tree of life and see the blue line

767
00:29:52,720 --> 00:29:57,600
for the hammerhead shark, the red line

768
00:29:55,039 --> 00:30:00,080
for the barrel fish. We can see where

769
00:29:57,600 --> 00:30:02,880
they meet and the humans are on the

770
00:30:00,080 --> 00:30:05,840
green line, which is along those lines.

771
00:30:02,880 --> 00:30:07,679
So, humans are a fish. Uh, humans are a

772
00:30:05,840 --> 00:30:10,480
fish. There are lots of other fish you

773
00:30:07,679 --> 00:30:14,000
probably didn't know are fish, you know,

774
00:30:10,480 --> 00:30:16,080
snakes, cows, frogs, dinosaurs.

775
00:30:14,000 --> 00:30:17,919
fish.

776
00:30:16,080 --> 00:30:19,919
So, what is the actual physical evidence

777
00:30:17,919 --> 00:30:23,760
of this apart from just me showing a

778
00:30:19,919 --> 00:30:26,240
graph? Um, your eyes are fish eyes. Uh,

779
00:30:23,760 --> 00:30:29,120
what I mean by this is the the lens in

780
00:30:26,240 --> 00:30:31,360
your eye, the refractive index closely

781
00:30:29,120 --> 00:30:34,080
matches water. They are evolved for

782
00:30:31,360 --> 00:30:35,279
seeing in water. Now, you might have

783
00:30:34,080 --> 00:30:38,240
opened your eyes under water and not

784
00:30:35,279 --> 00:30:40,000
seen very well in water. We have adapted

785
00:30:38,240 --> 00:30:41,760
extra things to make them work out of

786
00:30:40,000 --> 00:30:43,600
water where we're not supposed to have

787
00:30:41,760 --> 00:30:44,720
them. uh we've attached a whole bunch of

788
00:30:43,600 --> 00:30:46,000
little muscles around them and we're

789
00:30:44,720 --> 00:30:49,039
constantly pulling on the lens and

790
00:30:46,000 --> 00:30:51,440
distorting it and such. Uh this makes

791
00:30:49,039 --> 00:30:54,159
them work in air. Now um it also means

792
00:30:51,440 --> 00:30:55,679
that your eyes get tired. Fish fish in

793
00:30:54,159 --> 00:30:57,840
the water, their eyes don't get tired.

794
00:30:55,679 --> 00:31:01,520
Fish on the land, their eyes get tired.

795
00:30:57,840 --> 00:31:03,679
Um your eyes are fish eyes.

796
00:31:01,520 --> 00:31:06,000
In the opposite direction, a fish's swim

797
00:31:03,679 --> 00:31:08,880
bladder is just a lung. Uh the lung

798
00:31:06,000 --> 00:31:10,880
evolved first, not the swim bladder. Um,

799
00:31:08,880 --> 00:31:13,520
the lung is a very fancy and weird part

800
00:31:10,880 --> 00:31:16,159
of the digestive tract. That's why it

801
00:31:13,520 --> 00:31:18,799
shares the tube and you can choke. Um,

802
00:31:16,159 --> 00:31:22,080
don't uh

803
00:31:18,799 --> 00:31:24,000
but that's why um there's a if you've

804
00:31:22,080 --> 00:31:26,080
got if you got an area that can absorb

805
00:31:24,000 --> 00:31:27,919
nutrients, it can also absorb oxygen.

806
00:31:26,080 --> 00:31:30,559
Put that off in a bubble somewhere. Seal

807
00:31:27,919 --> 00:31:34,960
that. Hey, now you've got a swim bladder

808
00:31:30,559 --> 00:31:37,600
or a lung. And lastly, you have a jaw.

809
00:31:34,960 --> 00:31:39,600
Everything with a jaw is a fish. If we

810
00:31:37,600 --> 00:31:41,440
go back here, you might have noticed

811
00:31:39,600 --> 00:31:43,360
that those three lines all converge on

812
00:31:41,440 --> 00:31:46,159
the jawed vertebrates. Congratulations,

813
00:31:43,360 --> 00:31:48,080
you have a jaw. You are a fish. Um,

814
00:31:46,159 --> 00:31:50,640
everything with a jaw is a has

815
00:31:48,080 --> 00:31:54,000
everything with a jaw is a fish. Not

816
00:31:50,640 --> 00:31:57,440
every fish has a jaw. Uh, there are two

817
00:31:54,000 --> 00:31:59,679
families of fish that don't. Um, the the

818
00:31:57,440 --> 00:32:00,880
hagfish and the lamp rays. If you've

819
00:31:59,679 --> 00:32:02,559
ever seen a picture of one of them, you

820
00:32:00,880 --> 00:32:06,399
might go, "That is a weird looking

821
00:32:02,559 --> 00:32:08,399
fish." That is definitely true. And so

822
00:32:06,399 --> 00:32:11,519
the one cursed fact that I have for you

823
00:32:08,399 --> 00:32:15,880
is every terrestrial vertebrate is a

824
00:32:11,519 --> 00:32:15,880
fish. Thank you.

825
00:32:18,740 --> 00:32:23,519
[Applause]

826
00:32:20,960 --> 00:32:26,240
Um I happen to be sitting in relatively

827
00:32:23,519 --> 00:32:29,440
close proximity to a trained geneticist

828
00:32:26,240 --> 00:32:33,039
and I have not seen someone wincing that

829
00:32:29,440 --> 00:32:36,080
hard in quite some time.

830
00:32:33,039 --> 00:32:40,399
Uh Jordy, you are up on deck on this

831
00:32:36,080 --> 00:32:44,320
side. Uh Jay is going to uh talk to us

832
00:32:40,399 --> 00:32:45,760
about um radios.

833
00:32:44,320 --> 00:32:47,440
Thank you very much. Good afternoon. My

834
00:32:45,760 --> 00:32:48,880
name is Justin Smwood, but as you are

835
00:32:47,440 --> 00:32:50,880
all my friends, you may call me Jay.

836
00:32:48,880 --> 00:32:52,240
This is my preferred name. Uh and today

837
00:32:50,880 --> 00:32:54,240
I'm going to tell you about how we're

838
00:32:52,240 --> 00:32:56,399
going to save radio astronomy using some

839
00:32:54,240 --> 00:32:58,240
Python. First of all, a little bit about

840
00:32:56,399 --> 00:33:00,480
myself. I'm currently a PhD student at

841
00:32:58,240 --> 00:33:02,080
Swinburn University. Before that, I had

842
00:33:00,480 --> 00:33:03,840
a seven-year career in finance. These

843
00:33:02,080 --> 00:33:06,159
are some of the nicer names I was called

844
00:33:03,840 --> 00:33:07,360
during that time. Uh, and before that, I

845
00:33:06,159 --> 00:33:10,000
have degrees in economics, mathematics

846
00:33:07,360 --> 00:33:12,399
from the University of Melbourne. So,

847
00:33:10,000 --> 00:33:14,799
why does radio astronomy need saving?

848
00:33:12,399 --> 00:33:17,600
Well, first of all, here is a pretty

849
00:33:14,799 --> 00:33:20,320
typical radioastronomy observation. On

850
00:33:17,600 --> 00:33:22,559
the y-axis, we have the frequency. On

851
00:33:20,320 --> 00:33:25,039
the x-axis, we have time. And the pixel

852
00:33:22,559 --> 00:33:26,640
intensity is how strong the signal is in

853
00:33:25,039 --> 00:33:28,559
that particular frequency at that

854
00:33:26,640 --> 00:33:30,559
particular time. As you can see, we have

855
00:33:28,559 --> 00:33:31,919
a bit of a beam going down the middle.

856
00:33:30,559 --> 00:33:33,519
This could be from a black hole. This

857
00:33:31,919 --> 00:33:36,880
could be from a pulsar. This could be

858
00:33:33,519 --> 00:33:38,799
from many uh curious cosmic things. But

859
00:33:36,880 --> 00:33:40,799
I'm more interested in the bits that

860
00:33:38,799 --> 00:33:42,399
we've cut out.

861
00:33:40,799 --> 00:33:43,840
These are bits where the data has been

862
00:33:42,399 --> 00:33:47,120
corrupted by what we call radio

863
00:33:43,840 --> 00:33:49,679
frequency interference or RFI.

864
00:33:47,120 --> 00:33:52,480
And the problem is is that if we do

865
00:33:49,679 --> 00:33:54,720
nothing to stop these entering our

866
00:33:52,480 --> 00:33:57,360
particular things, things are going to

867
00:33:54,720 --> 00:33:58,720
go badly. And this is why. So here I've

868
00:33:57,360 --> 00:34:00,799
got a little image and we've got our

869
00:33:58,720 --> 00:34:02,799
little telescope in the middle and what

870
00:34:00,799 --> 00:34:05,200
we really want to observe is this blue

871
00:34:02,799 --> 00:34:07,919
signal. But there are things which are

872
00:34:05,200 --> 00:34:09,599
getting in the way. For instance, this

873
00:34:07,919 --> 00:34:11,760
red one in the sky might be a satellite

874
00:34:09,599 --> 00:34:14,000
or a plane or this one on the ground

875
00:34:11,760 --> 00:34:16,639
might be a cell tower or a microwave

876
00:34:14,000 --> 00:34:18,399
oven. And these things are also going on

877
00:34:16,639 --> 00:34:21,119
our telescope and they're getting into

878
00:34:18,399 --> 00:34:23,280
our data sets. And the problem is is

879
00:34:21,119 --> 00:34:24,720
that if we do nothing, these white

880
00:34:23,280 --> 00:34:27,280
regions that have been cut out are going

881
00:34:24,720 --> 00:34:29,839
to grow and grow and grow until there is

882
00:34:27,280 --> 00:34:31,760
nothing left. And I've even blocked out

883
00:34:29,839 --> 00:34:33,760
the citation at the top because if no

884
00:34:31,760 --> 00:34:37,000
one's getting any data, no one is

885
00:34:33,760 --> 00:34:37,000
getting published.

886
00:34:37,040 --> 00:34:41,839
And so how are we going to solve it? So

887
00:34:40,240 --> 00:34:44,320
this is Marang. This is the parks

888
00:34:41,839 --> 00:34:46,960
telescope. If you've seen the dish, this

889
00:34:44,320 --> 00:34:48,639
is the the dish. And I'm particularly

890
00:34:46,960 --> 00:34:50,399
interested what's the focus of this

891
00:34:48,639 --> 00:34:52,079
dish? This is something called a phased

892
00:34:50,399 --> 00:34:54,800
array feed. And there's a zoom in the

893
00:34:52,079 --> 00:34:56,240
little uh left hand corner. And each one

894
00:34:54,800 --> 00:34:58,320
of those little checkerboards is a

895
00:34:56,240 --> 00:35:01,440
receiver. Instead of one receiver, we're

896
00:34:58,320 --> 00:35:04,640
now using 200 receivers. And what this

897
00:35:01,440 --> 00:35:05,599
means is we can use the uh arrival times

898
00:35:04,640 --> 00:35:07,440
between each of these different

899
00:35:05,599 --> 00:35:10,000
receivers in order to figure out which

900
00:35:07,440 --> 00:35:13,200
directions things are coming from and

901
00:35:10,000 --> 00:35:15,280
get rid of them, which is great.

902
00:35:13,200 --> 00:35:16,960
So we're going to mathematically fill

903
00:35:15,280 --> 00:35:18,480
filter out this interference using a

904
00:35:16,960 --> 00:35:21,119
branch of mathematics called spatial

905
00:35:18,480 --> 00:35:23,359
filtering. But we have a problem. The

906
00:35:21,119 --> 00:35:26,000
data velocity is huge. We're talking

907
00:35:23,359 --> 00:35:29,119
about the order of terabytes per second.

908
00:35:26,000 --> 00:35:30,480
So really this has to be done in C++

909
00:35:29,119 --> 00:35:32,320
running on graphics cards at the

910
00:35:30,480 --> 00:35:34,720
telescope.

911
00:35:32,320 --> 00:35:39,200
So how am I using Python to help me do

912
00:35:34,720 --> 00:35:41,359
this? Python use one. I'm making small

913
00:35:39,200 --> 00:35:42,800
interactive applications to simulate and

914
00:35:41,359 --> 00:35:44,960
understand the mathematical techniques

915
00:35:42,800 --> 00:35:46,240
that I need to apply. And Streamlit is

916
00:35:44,960 --> 00:35:49,119
something that's really really good for

917
00:35:46,240 --> 00:35:51,040
this. Um there's a little link to one of

918
00:35:49,119 --> 00:35:53,440
them on on my website, but essentially

919
00:35:51,040 --> 00:35:55,599
this lets me simulate different signals,

920
00:35:53,440 --> 00:35:57,200
apply the techniques and mess around

921
00:35:55,599 --> 00:35:58,160
with all the sliders and things and make

922
00:35:57,200 --> 00:36:00,160
graphs which look completely

923
00:35:58,160 --> 00:36:02,000
unintelligible. But this helps me

924
00:36:00,160 --> 00:36:04,000
understand the techniques before I then

925
00:36:02,000 --> 00:36:05,920
go and apply them in a really high

926
00:36:04,000 --> 00:36:07,520
performance environment.

927
00:36:05,920 --> 00:36:10,079
The second way that Python is helping me

928
00:36:07,520 --> 00:36:11,920
save radio astronomy is to shorten the

929
00:36:10,079 --> 00:36:16,800
develop loops when I'm actually writing

930
00:36:11,920 --> 00:36:19,280
C++ code. I'm fairly new to C++ and so

931
00:36:16,800 --> 00:36:20,800
this helps me uh figure out what I'm

932
00:36:19,280 --> 00:36:24,240
going to do before I actually have to go

933
00:36:20,800 --> 00:36:26,240
and do it in C++ and that helps me. So

934
00:36:24,240 --> 00:36:29,760
that's a really good use of Python in my

935
00:36:26,240 --> 00:36:32,960
in my uh opinion.

936
00:36:29,760 --> 00:36:35,920
So where is all this going? Well, at the

937
00:36:32,960 --> 00:36:39,680
end of my project, what we're eventually

938
00:36:35,920 --> 00:36:41,280
hoping is that we turn this

939
00:36:39,680 --> 00:36:44,280
into

940
00:36:41,280 --> 00:36:44,280
this.

941
00:36:44,310 --> 00:36:54,640
[Applause]

942
00:36:51,920 --> 00:36:57,280
Thank you so much, Justin. So, up next

943
00:36:54,640 --> 00:37:00,079
on this side, we have Toby. Love it. But

944
00:36:57,280 --> 00:37:01,599
for now, Jordi is going to tell us how

945
00:37:00,079 --> 00:37:03,680
we can shove eight graphics cards into

946
00:37:01,599 --> 00:37:07,960
one computer without starting an

947
00:37:03,680 --> 00:37:07,960
electrical fire. Take it away, Jordi.

948
00:37:09,040 --> 00:37:15,040
So, a lot of um ML and AI workloads

949
00:37:12,240 --> 00:37:18,640
depend on two things, Python and GPU

950
00:37:15,040 --> 00:37:21,680
power. But the thing is GPUs are only

951
00:37:18,640 --> 00:37:24,480
getting wider and half your slots end up

952
00:37:21,680 --> 00:37:26,880
under the damn card. So, what you can

953
00:37:24,480 --> 00:37:30,960
see there is a uh photo someone has

954
00:37:26,880 --> 00:37:33,040
leaked to me of the GTX uh 6090 um which

955
00:37:30,960 --> 00:37:34,880
my sources say will try and draw twice

956
00:37:33,040 --> 00:37:38,560
the power through a power connector half

957
00:37:34,880 --> 00:37:40,240
as thick. But um yeah, what can we do

958
00:37:38,560 --> 00:37:41,920
other than being dependent on cloud

959
00:37:40,240 --> 00:37:44,320
services? Maybe we just need some

960
00:37:41,920 --> 00:37:47,119
extension cords. But what could possibly

961
00:37:44,320 --> 00:37:49,520
go wrong? So, a little warning here. You

962
00:37:47,119 --> 00:37:51,599
are opening Pandora's box. Uh the

963
00:37:49,520 --> 00:37:54,960
smaller PCIe slots are only rated for

964
00:37:51,599 --> 00:37:57,119
2.1 amps or maybe even half an amp. Time

965
00:37:54,960 --> 00:37:59,359
16 high power cards can actually pull 6

966
00:37:57,119 --> 00:38:02,079
amps out of the slot. And all of that

967
00:37:59,359 --> 00:38:04,320
comes from two pins on your 24 pin

968
00:38:02,079 --> 00:38:05,599
motherboard power connector. So three

969
00:38:04,320 --> 00:38:08,480
cards on the board and you'll be

970
00:38:05,599 --> 00:38:10,480
breaking that limit. Also, if you stick

971
00:38:08,480 --> 00:38:13,280
a card on a riser cable, it's not

972
00:38:10,480 --> 00:38:14,960
screwed down. Um if it hits your metal

973
00:38:13,280 --> 00:38:17,119
case, you're going to shortcircuit

974
00:38:14,960 --> 00:38:19,040
things. And not that I've done that, but

975
00:38:17,119 --> 00:38:20,880
the slot does still work afterwards if

976
00:38:19,040 --> 00:38:23,280
you add a riser to it that provides

977
00:38:20,880 --> 00:38:25,920
power again.

978
00:38:23,280 --> 00:38:29,680
So there's a there's a lot of pins here,

979
00:38:25,920 --> 00:38:33,040
but which ones do we really need?

980
00:38:29,680 --> 00:38:35,200
So 12 volts, 3.3 ground, supply them at

981
00:38:33,040 --> 00:38:37,359
the remote end. Uh ref clock, most

982
00:38:35,200 --> 00:38:40,400
devices actually do need that with some

983
00:38:37,359 --> 00:38:42,480
exceptions. PCIe reset, most devices do

984
00:38:40,400 --> 00:38:44,240
need that to start up properly. And

985
00:38:42,480 --> 00:38:47,599
lastly, there's two differential pairs

986
00:38:44,240 --> 00:38:49,680
for each lane. Um, so 1 2 4 8 or 16 of

987
00:38:47,599 --> 00:38:52,320
those depending on what width of link

988
00:38:49,680 --> 00:38:55,760
you want. Trouble is, this is all

989
00:38:52,320 --> 00:38:57,680
practically RF. It's like 10 GHz in

990
00:38:55,760 --> 00:39:00,000
terms of signals. Corners might have

991
00:38:57,680 --> 00:39:01,680
been cut somewhere. Your It's not really

992
00:39:00,000 --> 00:39:03,520
just a matter of connected. You need a

993
00:39:01,680 --> 00:39:05,200
very good signal between both ends, but

994
00:39:03,520 --> 00:39:06,880
your motherboard vendor probably cut

995
00:39:05,200 --> 00:39:09,520
some corners. Your card vendor probably

996
00:39:06,880 --> 00:39:11,119
did, too. The newer PCIe modes run on

997
00:39:09,520 --> 00:39:12,560
much tighter margins. and sticking a

998
00:39:11,119 --> 00:39:14,880
bunch of wire in the middle is not

999
00:39:12,560 --> 00:39:16,640
helping. So, there's no guarantees.

1000
00:39:14,880 --> 00:39:18,320
Expect some frustration, but that just

1001
00:39:16,640 --> 00:39:21,280
makes it more satisfying when it does

1002
00:39:18,320 --> 00:39:23,599
work. So, crypto miners love large

1003
00:39:21,280 --> 00:39:25,839
clusters of GPUs. What do they use?

1004
00:39:23,599 --> 00:39:27,920
Wait, what? What's going on here? Well,

1005
00:39:25,839 --> 00:39:31,440
that's not actually USB. It's just using

1006
00:39:27,920 --> 00:39:33,520
the USB3 wires. But we want more than a

1007
00:39:31,440 --> 00:39:35,760
times one link. Machine learning and AI,

1008
00:39:33,520 --> 00:39:38,560
we put a lot of data in and get a lot of

1009
00:39:35,760 --> 00:39:40,800
data out. We're not just trying to guess

1010
00:39:38,560 --> 00:39:43,200
a number.

1011
00:39:40,800 --> 00:39:46,079
So, speaking of USB, uh we got

1012
00:39:43,200 --> 00:39:49,599
Thunderbolt or USB 4. Good option for

1013
00:39:46,079 --> 00:39:52,560
getting cards on a laptop, but you can

1014
00:39:49,599 --> 00:39:54,960
only put one of those in your PC, and

1015
00:39:52,560 --> 00:39:57,040
it's still only times four. At the other

1016
00:39:54,960 --> 00:40:00,079
end, your choices are a little sketchy

1017
00:39:57,040 --> 00:40:01,760
or quite extremely expensive, and I've

1018
00:40:00,079 --> 00:40:04,800
heard Intel doesn't really like these

1019
00:40:01,760 --> 00:40:07,440
things existing anyway. So, you go on

1020
00:40:04,800 --> 00:40:09,200
eBay, you search for the right cables,

1021
00:40:07,440 --> 00:40:11,599
ribbon rises. They're generally the most

1022
00:40:09,200 --> 00:40:14,160
common option. No external power. The

1023
00:40:11,599 --> 00:40:16,320
good ones are even shielded.

1024
00:40:14,160 --> 00:40:17,839
Um, I did not have very good luck with

1025
00:40:16,320 --> 00:40:20,720
these, especially when they were longer

1026
00:40:17,839 --> 00:40:22,160
than 10 cm. So, that one, yeah, good

1027
00:40:20,720 --> 00:40:25,119
luck with that. It's probably going to

1028
00:40:22,160 --> 00:40:27,599
drop you to PCIe 1.0. But, can we get

1029
00:40:25,119 --> 00:40:29,520
more than one device on the slot? So,

1030
00:40:27,599 --> 00:40:32,079
bifocation, it's not actually a feature

1031
00:40:29,520 --> 00:40:33,599
of PCIe. Your motherboard's chipset or

1032
00:40:32,079 --> 00:40:35,440
your CPU is programmed to say which

1033
00:40:33,599 --> 00:40:37,599
lanes go to which slot. We just change

1034
00:40:35,440 --> 00:40:39,839
that. Pretend there's multiple slots.

1035
00:40:37,599 --> 00:40:42,079
get a card like this to alter reality to

1036
00:40:39,839 --> 00:40:45,520
match the software. Um, and then that's

1037
00:40:42,079 --> 00:40:48,640
got two downstream um connectors to go

1038
00:40:45,520 --> 00:40:50,400
to 2 * 8 links and uh yeah, that

1039
00:40:48,640 --> 00:40:52,160
actually borrows some variant of a SAS

1040
00:40:50,400 --> 00:40:54,160
connector because like previous one,

1041
00:40:52,160 --> 00:40:56,079
it's not SAS, but we're just using the

1042
00:40:54,160 --> 00:40:58,079
connector. So that usually lets you

1043
00:40:56,079 --> 00:41:00,240
split it two ways into 8 and 8 or four

1044
00:40:58,079 --> 00:41:03,119
ways into 444. You can't really go below

1045
00:41:00,240 --> 00:41:05,040
that. Your BIOS might let you down by

1046
00:41:03,119 --> 00:41:06,640
not having the menu option visible to

1047
00:41:05,040 --> 00:41:08,240
actually select that on the slot. So you

1048
00:41:06,640 --> 00:41:11,839
just decompile it and poke the right

1049
00:41:08,240 --> 00:41:14,240
value into the EFI variables with RU.

1050
00:41:11,839 --> 00:41:17,680
Simple simple, but surely there's a

1051
00:41:14,240 --> 00:41:19,920
better way. Um, PCIe is packet-based

1052
00:41:17,680 --> 00:41:21,280
switches. And there's even some cheap

1053
00:41:19,920 --> 00:41:23,920
options. You don't just get two

1054
00:41:21,280 --> 00:41:25,760
downstream times 8 links, you get four.

1055
00:41:23,920 --> 00:41:27,200
Uh, with the right software, you can

1056
00:41:25,760 --> 00:41:30,480
even reconfigure this thing to get it

1057
00:41:27,200 --> 00:41:31,920
into times one uh or go much further. So

1058
00:41:30,480 --> 00:41:34,160
then how do you get that to a slot?

1059
00:41:31,920 --> 00:41:35,839
Breakouts. Note the SATA connector. Only

1060
00:41:34,160 --> 00:41:37,440
two amps. Also, we're supposed to have

1061
00:41:35,839 --> 00:41:39,040
some pins connected up there. It

1062
00:41:37,440 --> 00:41:41,359
doesn't. No one actually cares. It

1063
00:41:39,040 --> 00:41:43,359
works. This one, bit of a problem. You

1064
00:41:41,359 --> 00:41:46,560
can see on the right. Yep, they mirrored

1065
00:41:43,359 --> 00:41:48,880
the pin out. Um, so what we have to do

1066
00:41:46,560 --> 00:41:52,079
is move those things up there by doing

1067
00:41:48,880 --> 00:41:54,400
surgery on the cable and then it works.

1068
00:41:52,079 --> 00:41:56,640
Uh, fortunately, um, only had

1069
00:41:54,400 --> 00:41:59,359
fortunately ref clocks only 100 MHz

1070
00:41:56,640 --> 00:42:00,960
practically DC. This one has the lanes

1071
00:41:59,359 --> 00:42:04,480
backwards, but they got the rest of it

1072
00:42:00,960 --> 00:42:07,720
correct. And

1073
00:42:04,480 --> 00:42:07,720
it works.

1074
00:42:10,319 --> 00:42:18,000
Thanks, Jordy. I I I think

1075
00:42:15,359 --> 00:42:20,400
um

1076
00:42:18,000 --> 00:42:23,400
Alex, you are up on deck on this side.

1077
00:42:20,400 --> 00:42:23,400
Um

1078
00:42:24,960 --> 00:42:30,680
uh Toby is going to talk to us about

1079
00:42:27,359 --> 00:42:30,680
data structures.

1080
00:42:32,079 --> 00:42:37,359
I will I will quickly preface this uh

1081
00:42:35,040 --> 00:42:39,520
with if you happen to be in the audience

1082
00:42:37,359 --> 00:42:42,400
and you're my father, you're not allowed

1083
00:42:39,520 --> 00:42:44,480
to comment on this.

1084
00:42:42,400 --> 00:42:47,040
Seems unlikely.

1085
00:42:44,480 --> 00:42:49,920
What are the chances?

1086
00:42:47,040 --> 00:42:53,280
Last year, I had a friend over who had

1087
00:42:49,920 --> 00:42:58,960
some strong opinions on the cleanliness

1088
00:42:53,280 --> 00:43:01,440
or perhaps lack thereof of my room.

1089
00:42:58,960 --> 00:43:03,200
You know, fair enough. I just moved. It

1090
00:43:01,440 --> 00:43:06,000
was covered in clothes and shoes and

1091
00:43:03,200 --> 00:43:09,240
boxes and push bike and plastic and

1092
00:43:06,000 --> 00:43:09,240
other things.

1093
00:43:09,440 --> 00:43:12,880
But that's not what I want to talk

1094
00:43:10,880 --> 00:43:16,079
about.

1095
00:43:12,880 --> 00:43:21,000
Last FM is a music stats website for

1096
00:43:16,079 --> 00:43:21,000
nerds. Any nerds in the audience?

1097
00:43:22,000 --> 00:43:27,280
It keeps track of every song that you've

1098
00:43:24,400 --> 00:43:30,960
listened to across platforms, across

1099
00:43:27,280 --> 00:43:32,640
everything. And it lets you see

1100
00:43:30,960 --> 00:43:34,640
collections of this. It lets you see

1101
00:43:32,640 --> 00:43:38,920
exactly what you listen to and when. And

1102
00:43:34,640 --> 00:43:38,920
it lets you see this for your friends.

1103
00:43:39,200 --> 00:43:44,800
Despite have despite last FM having a

1104
00:43:42,079 --> 00:43:46,319
recommendation system already in it, I

1105
00:43:44,800 --> 00:43:49,040
decided

1106
00:43:46,319 --> 00:43:51,760
I want to do it myself. It's very

1107
00:43:49,040 --> 00:43:53,599
programmercoded, I think, to say, "This

1108
00:43:51,760 --> 00:43:57,520
thing already exists, but I want to do

1109
00:43:53,599 --> 00:43:59,680
it worse, but for more effort."

1110
00:43:57,520 --> 00:44:02,400
Now, in the process of downloading all

1111
00:43:59,680 --> 00:44:04,640
of this user data, I'd made a mistake

1112
00:44:02,400 --> 00:44:07,760
with the API, and I wouldn't realize

1113
00:44:04,640 --> 00:44:10,079
this for a couple more weeks, but it led

1114
00:44:07,760 --> 00:44:13,040
me to an interesting moment that I think

1115
00:44:10,079 --> 00:44:15,119
is helpful for teaching about some data

1116
00:44:13,040 --> 00:44:17,359
structures.

1117
00:44:15,119 --> 00:44:20,319
Because of the mistake I made, I was

1118
00:44:17,359 --> 00:44:22,720
getting some overlaps in when I was

1119
00:44:20,319 --> 00:44:24,640
pageionating these downloads and I

1120
00:44:22,720 --> 00:44:28,319
couldn't really see where the overlaps

1121
00:44:24,640 --> 00:44:30,400
were starting and finishing. And so I

1122
00:44:28,319 --> 00:44:32,720
just had to check here's the new songs

1123
00:44:30,400 --> 00:44:35,440
that I've got and here's every other

1124
00:44:32,720 --> 00:44:37,520
song I've ever downloaded,

1125
00:44:35,440 --> 00:44:41,359
which was regularly hundreds of

1126
00:44:37,520 --> 00:44:43,839
thousands of songs per user.

1127
00:44:41,359 --> 00:44:46,079
This is what a list looks like.

1128
00:44:43,839 --> 00:44:48,319
Basically, in CPython, we've got one

1129
00:44:46,079 --> 00:44:50,800
light from here. And it's very easy to

1130
00:44:48,319 --> 00:44:52,480
see if there's any overlap. If I'm

1131
00:44:50,800 --> 00:44:54,720
looking for a song, I can see if it's

1132
00:44:52,480 --> 00:44:57,520
right there. But if I've got lots and

1133
00:44:54,720 --> 00:45:00,000
lots and lots of songs, it becomes much

1134
00:44:57,520 --> 00:45:02,079
harder to go through every single one

1135
00:45:00,000 --> 00:45:05,440
and check is this song in here, and it's

1136
00:45:02,079 --> 00:45:07,760
this song, is this song. Lists get very

1137
00:45:05,440 --> 00:45:10,640
time consuming for checking. And I was

1138
00:45:07,760 --> 00:45:12,800
checking every page against every item

1139
00:45:10,640 --> 00:45:14,640
in this list for every user and it was

1140
00:45:12,800 --> 00:45:16,720
adding minutes and minutes and minutes

1141
00:45:14,640 --> 00:45:20,200
to my downloads.

1142
00:45:16,720 --> 00:45:20,200
Didn't want that.

1143
00:45:20,480 --> 00:45:25,760
By the end of that, I had kind of come

1144
00:45:24,079 --> 00:45:28,079
up with the idea that I needed to do

1145
00:45:25,760 --> 00:45:30,319
something differently. And so I said

1146
00:45:28,079 --> 00:45:32,160
sets

1147
00:45:30,319 --> 00:45:35,680
sets are a much better structure for

1148
00:45:32,160 --> 00:45:37,839
what I'm doing here because

1149
00:45:35,680 --> 00:45:40,720
and and somebody much smarter than me

1150
00:45:37,839 --> 00:45:42,560
can correct if I've misrepresented

1151
00:45:40,720 --> 00:45:46,319
something. But what we're doing is

1152
00:45:42,560 --> 00:45:49,760
instead of doing a list of items, we

1153
00:45:46,319 --> 00:45:52,720
just put everything there. And we can

1154
00:45:49,760 --> 00:45:55,119
see that each of these spaces will

1155
00:45:52,720 --> 00:45:58,000
represent a specific song, a specific

1156
00:45:55,119 --> 00:45:59,760
ID, a specific something. And we can

1157
00:45:58,000 --> 00:46:01,760
populate these and very quickly just

1158
00:45:59,760 --> 00:46:06,240
look at it, look at it all at once and

1159
00:46:01,760 --> 00:46:08,560
see if it is there.

1160
00:46:06,240 --> 00:46:11,920
So when I'm trying to download my songs,

1161
00:46:08,560 --> 00:46:15,359
it becomes almost instantaneous. It went

1162
00:46:11,920 --> 00:46:19,079
from a few seconds per page to I didn't

1163
00:46:15,359 --> 00:46:19,079
even notice it running.

1164
00:46:19,280 --> 00:46:23,599
So,

1165
00:46:21,359 --> 00:46:26,240
this is good. And I actually ended up

1166
00:46:23,599 --> 00:46:28,160
just putting everything into a set

1167
00:46:26,240 --> 00:46:30,079
because sets are unique. You can see

1168
00:46:28,160 --> 00:46:32,880
here that there's no two spots per

1169
00:46:30,079 --> 00:46:35,040
thing. We can just store unique items

1170
00:46:32,880 --> 00:46:37,280
and everything into a set. Sort it at

1171
00:46:35,040 --> 00:46:39,440
the end.

1172
00:46:37,280 --> 00:46:43,200
Now, my final aside is that now that

1173
00:46:39,440 --> 00:46:46,000
I've got my recommener working, you can

1174
00:46:43,200 --> 00:46:48,000
DM me. Here's my link. I don't have a QR

1175
00:46:46,000 --> 00:46:50,480
code.

1176
00:46:48,000 --> 00:46:53,920
It's not very good at recommending, but

1177
00:46:50,480 --> 00:46:56,560
you're welcome to ask me for it.

1178
00:46:53,920 --> 00:46:58,240
Um, you will need a Last FM account and

1179
00:46:56,560 --> 00:47:01,680
a few hundred listens to get anything

1180
00:46:58,240 --> 00:47:03,520
good from this. Now, with all that being

1181
00:47:01,680 --> 00:47:06,640
said, I think there is actually an

1182
00:47:03,520 --> 00:47:09,280
obvious takeaway from this.

1183
00:47:06,640 --> 00:47:12,240
My room isn't messy. It's highly

1184
00:47:09,280 --> 00:47:15,280
optimized and efficient for finding

1185
00:47:12,240 --> 00:47:17,760
unique items quickly.

1186
00:47:15,280 --> 00:47:20,760
And finally, does the data agree with

1187
00:47:17,760 --> 00:47:20,760
me?

1188
00:47:21,040 --> 00:47:27,669
Kate, fine. I'll clean my room.

1189
00:47:24,590 --> 00:47:27,669
[Applause]

1190
00:47:30,960 --> 00:47:35,200
Thank you so much, Toby, for giving me a

1191
00:47:33,680 --> 00:47:37,920
plausible explanation for my why my

1192
00:47:35,200 --> 00:47:40,800
apartment looks the way it does.

1193
00:47:37,920 --> 00:47:43,680
Uh, up on this side, we have Alan Rubin.

1194
00:47:40,800 --> 00:47:46,160
Uh but first uh so some of you may

1195
00:47:43,680 --> 00:47:47,599
remember in the preandemic Pythons at

1196
00:47:46,160 --> 00:47:48,880
lightning talks we had some uh

1197
00:47:47,599 --> 00:47:51,040
presentations about history and

1198
00:47:48,880 --> 00:47:56,359
politics.

1199
00:47:51,040 --> 00:47:56,359
This year it's a fullblown conspiracy

1200
00:48:01,839 --> 00:48:08,240
ginger nuts are not equal. So there's

1201
00:48:04,560 --> 00:48:10,000
actually four of them. uh bought out

1202
00:48:08,240 --> 00:48:14,480
many bakeries between uh in the

1203
00:48:10,000 --> 00:48:16,319
1940s,50s and60s uh and well they kept

1204
00:48:14,480 --> 00:48:19,119
some of them because basically what

1205
00:48:16,319 --> 00:48:20,960
happened is well they tried to go to the

1206
00:48:19,119 --> 00:48:22,319
arnut skew and then people hated it

1207
00:48:20,960 --> 00:48:25,599
because why would you want the New South

1208
00:48:22,319 --> 00:48:27,920
Wales Ginger skew? Uh it's just bad. So

1209
00:48:25,599 --> 00:48:31,040
anyway, this entire situation happened

1210
00:48:27,920 --> 00:48:33,520
like uh so I was curious after this

1211
00:48:31,040 --> 00:48:37,520
after I came uh home from a road trip

1212
00:48:33,520 --> 00:48:39,359
with some friends and we arrived home uh

1213
00:48:37,520 --> 00:48:41,119
and all they had in their pantry was a

1214
00:48:39,359 --> 00:48:43,200
packet of ginger nuts. It wasn't quite

1215
00:48:41,119 --> 00:48:44,800
right. It was like it tasted weird. The

1216
00:48:43,200 --> 00:48:48,800
texture was wrong. Turns out it was a

1217
00:48:44,800 --> 00:48:52,079
South Australian skew. Uh so as as

1218
00:48:48,800 --> 00:48:53,760
everyone would want to finding out this

1219
00:48:52,079 --> 00:48:56,079
conspiracy, I would want to find out

1220
00:48:53,760 --> 00:48:58,000
where I can buy them. So obviously I had

1221
00:48:56,079 --> 00:49:01,200
to reverse engineer the Coohl's online

1222
00:48:58,000 --> 00:49:03,119
API like I I looked at Coohl's and Worth

1223
00:49:01,200 --> 00:49:08,000
and I went Coohl's is easier so I just

1224
00:49:03,119 --> 00:49:13,920
went with that. Uh so I like queried

1225
00:49:08,000 --> 00:49:15,440
their API like 10,000 times cuz I uh PQ

1226
00:49:13,920 --> 00:49:18,319
uh because I didn't know what stores

1227
00:49:15,440 --> 00:49:21,119
actually existed because like uh the API

1228
00:49:18,319 --> 00:49:25,440
uh uh required a store. So, I just tried

1229
00:49:21,119 --> 00:49:30,040
every integer between uh 0 and 9,999

1230
00:49:25,440 --> 00:49:30,040
and I found about 800 stores.

1231
00:49:30,079 --> 00:49:35,839
Uh I probably queried their API probably

1232
00:49:33,040 --> 00:49:39,119
about 110,000 times,

1233
00:49:35,839 --> 00:49:43,480
but they earned $1.1 billion in profit

1234
00:49:39,119 --> 00:49:43,480
last year, so I don't really give a

1235
00:49:46,240 --> 00:49:51,680
uh one.

1236
00:49:48,960 --> 00:49:53,280
Uh, so then I went to go, "Oh, I've got

1237
00:49:51,680 --> 00:49:55,920
all of these stores. I don't know where

1238
00:49:53,280 --> 00:49:58,319
these 800 stores are." So I looked at

1239
00:49:55,920 --> 00:50:00,800
the easy option, went, "Oh, where's the

1240
00:49:58,319 --> 00:50:04,240
click and collect location uh near me

1241
00:50:00,800 --> 00:50:06,400
API?" And then I realized uh and decided

1242
00:50:04,240 --> 00:50:09,520
to expand the amount of stores it would

1243
00:50:06,400 --> 00:50:11,280
return and the distance and it didn't

1244
00:50:09,520 --> 00:50:15,200
work. Uh, turns out they have a hard

1245
00:50:11,280 --> 00:50:17,040
limit of well distance like kilometers

1246
00:50:15,200 --> 00:50:19,839
and amount of stores returned. So that

1247
00:50:17,040 --> 00:50:22,079
wasn't feasible. So, I ended up going

1248
00:50:19,839 --> 00:50:24,880
digging and found this random really

1249
00:50:22,079 --> 00:50:26,640
looking internal looking ASPX uh web

1250
00:50:24,880 --> 00:50:28,480
page.

1251
00:50:26,640 --> 00:50:31,119
Unfortunately, it was quite outdated and

1252
00:50:28,480 --> 00:50:34,960
it uh includes lots of uh pick and pays

1253
00:50:31,119 --> 00:50:36,960
and a bo uh a former BO locations and I

1254
00:50:34,960 --> 00:50:40,319
looked up one and it was actually an IG

1255
00:50:36,960 --> 00:50:42,319
now. So, I guess they sold it off. So,

1256
00:50:40,319 --> 00:50:44,000
obviously that's not feasible. like I

1257
00:50:42,319 --> 00:50:47,359
was missing a large amount of stores

1258
00:50:44,000 --> 00:50:51,599
that I knew existed because I had uh uh

1259
00:50:47,359 --> 00:50:54,880
stock uh amounts for all of them. So,

1260
00:50:51,599 --> 00:50:57,920
I just found another API. Uh, like I

1261
00:50:54,880 --> 00:51:02,240
can't find it, but uh I went on Google

1262
00:50:57,920 --> 00:51:04,319
search uh Cooh's locations and I found a

1263
00:51:02,240 --> 00:51:08,000
a page for a particular store and turns

1264
00:51:04,319 --> 00:51:11,280
out they had a redirect on uh for most

1265
00:51:08,000 --> 00:51:13,760
store locations using on their old store

1266
00:51:11,280 --> 00:51:15,920
location page uh except for this one

1267
00:51:13,760 --> 00:51:18,559
store. So, I found their API. Turns out

1268
00:51:15,920 --> 00:51:21,119
they've since killed it. Uh so you can

1269
00:51:18,559 --> 00:51:23,599
see up the top left the uh top right

1270
00:51:21,119 --> 00:51:27,440
that's the result that I got in May 2014

1271
00:51:23,599 --> 00:51:29,200
and now it's dead. Oh well. Uh so I

1272
00:51:27,440 --> 00:51:31,040
still wasn't happy. Like I still had

1273
00:51:29,200 --> 00:51:33,440
quite a few stores that I didn't know

1274
00:51:31,040 --> 00:51:35,920
what it was. So I found another free API

1275
00:51:33,440 --> 00:51:38,800
because of course I did.

1276
00:51:35,920 --> 00:51:42,559
So I found all of this data. Now how the

1277
00:51:38,800 --> 00:51:44,400
hell would I display it? So I like all

1278
00:51:42,559 --> 00:51:45,599
of this were I started this entire

1279
00:51:44,400 --> 00:51:47,040
project because I wanted to learn

1280
00:51:45,599 --> 00:51:49,520
Python. And I had never done Python

1281
00:51:47,040 --> 00:51:52,000
before and well there's a lot of

1282
00:51:49,520 --> 00:51:54,400
everything in this thing. So I'm like

1283
00:51:52,000 --> 00:51:56,880
okay I'll have a look. And I ended up

1284
00:51:54,400 --> 00:51:59,760
finding uh geopandas and created a

1285
00:51:56,880 --> 00:52:01,920
static plot. But I went no I don't want

1286
00:51:59,760 --> 00:52:03,760
that. I want to be able to explore the

1287
00:52:01,920 --> 00:52:06,400
map. I want to zoom in because this is

1288
00:52:03,760 --> 00:52:08,880
kind of not workable. Like I can't tell

1289
00:52:06,400 --> 00:52:10,800
which uh stores in Melbourne have what

1290
00:52:08,880 --> 00:52:12,880
because they're just one pixel. It

1291
00:52:10,800 --> 00:52:15,119
doesn't make sense. It doesn't work. So

1292
00:52:12,880 --> 00:52:17,280
back to the drawing bud. Uh and then I

1293
00:52:15,119 --> 00:52:19,280
also looked at more tutorials and found

1294
00:52:17,280 --> 00:52:21,359
out that geopandis actually has another

1295
00:52:19,280 --> 00:52:24,160
thing for like uh uh I think it's called

1296
00:52:21,359 --> 00:52:27,119
explore and it just uses folium uh which

1297
00:52:24,160 --> 00:52:30,480
uh uh uses like generates like a leaflet

1298
00:52:27,119 --> 00:52:33,200
map and well you can just do anything

1299
00:52:30,480 --> 00:52:37,440
and you can export it as HTML. So I did.

1300
00:52:33,200 --> 00:52:39,920
So then I created a uh oh

1301
00:52:37,440 --> 00:52:42,160
yep there we go.

1302
00:52:39,920 --> 00:52:44,800
If I can do it right, I've now have a

1303
00:52:42,160 --> 00:52:46,000
thingy in the top corner.

1304
00:52:44,800 --> 00:52:49,800
I don't know what to do.

1305
00:52:46,000 --> 00:52:49,800
Okay, we can do that.

1306
00:52:51,280 --> 00:52:54,760
There we go.

1307
00:52:55,920 --> 00:53:01,280
Stop. Stop. So, I made a web page about

1308
00:53:00,079 --> 00:53:06,200
it. Uh, and

1309
00:53:01,280 --> 00:53:06,200
that's it. That's your five minutes.

1310
00:53:07,680 --> 00:53:17,040
Uh so some things to note that are we?

1311
00:53:12,160 --> 00:53:20,000
Yeah. So uh G, you're on that side. Um

1312
00:53:17,040 --> 00:53:22,319
so um I mentioned earlier that we had a

1313
00:53:20,000 --> 00:53:27,280
geneticist in the room. Uh Jack, I think

1314
00:53:22,319 --> 00:53:28,480
we've got a response to your talk. Um

1315
00:53:27,280 --> 00:53:28,800
that that is what you're doing, right,

1316
00:53:28,480 --> 00:53:29,680
Alan?

1317
00:53:28,800 --> 00:53:30,880
Uh sure. Now.

1318
00:53:29,680 --> 00:53:31,440
Yeah, let's go with that. Okay. Okay.

1319
00:53:30,880 --> 00:53:34,680
Maybe tomorrow.

1320
00:53:31,440 --> 00:53:34,680
Hello everyone.

1321
00:53:35,440 --> 00:53:38,160
All right. Thanks everyone. I'm going to

1322
00:53:36,640 --> 00:53:39,520
tell you about genomic data standards.

1323
00:53:38,160 --> 00:53:40,640
So first of all, here's a cartoon of a

1324
00:53:39,520 --> 00:53:42,160
DNA molecule. We're talking about

1325
00:53:40,640 --> 00:53:44,559
genomics. We're going to talk about DNA.

1326
00:53:42,160 --> 00:53:46,000
DNA stands for deoxxyribboucleic acid.

1327
00:53:44,559 --> 00:53:47,040
It's how your genome is encoded. And

1328
00:53:46,000 --> 00:53:48,319
that's the blueprint for all the cells

1329
00:53:47,040 --> 00:53:49,599
in your body. I hope many of you saw

1330
00:53:48,319 --> 00:53:51,839
Nick's talk this morning. It was really

1331
00:53:49,599 --> 00:53:53,359
great. Uh so DNA is a polymer chain made

1332
00:53:51,839 --> 00:53:56,240
up of what we call bases. And each base

1333
00:53:53,359 --> 00:53:57,839
is given a letter A, T, C, and G. And we

1334
00:53:56,240 --> 00:53:59,440
can now sequence DNA very efficiently.

1335
00:53:57,839 --> 00:54:00,800
So here's a picture of a DNA sequencer.

1336
00:53:59,440 --> 00:54:02,079
It's an instrument that fits on a

1337
00:54:00,800 --> 00:54:03,680
benchtop. I found this one on

1338
00:54:02,079 --> 00:54:05,040
AliExpress. I'm not sure if it's any

1339
00:54:03,680 --> 00:54:07,280
good because there weren't any reviews,

1340
00:54:05,040 --> 00:54:08,559
but it uses nanoport technology. So, you

1341
00:54:07,280 --> 00:54:10,240
can use these instruments to read out

1342
00:54:08,559 --> 00:54:11,599
the sequence of any DNA molecule, and we

1343
00:54:10,240 --> 00:54:14,079
usually represent those as text strings

1344
00:54:11,599 --> 00:54:15,280
of AC, C's, and G's. And so, we store

1345
00:54:14,079 --> 00:54:16,559
those text sequences and some

1346
00:54:15,280 --> 00:54:18,559
information that comes off the sequencer

1347
00:54:16,559 --> 00:54:19,760
using a file format called fast Q. But

1348
00:54:18,559 --> 00:54:21,200
usually, we aren't actually really

1349
00:54:19,760 --> 00:54:23,200
interested in the sequences themselves,

1350
00:54:21,200 --> 00:54:25,280
but rather how they are compared against

1351
00:54:23,200 --> 00:54:26,880
another genome or how they're different.

1352
00:54:25,280 --> 00:54:28,640
And so variant calling is the process

1353
00:54:26,880 --> 00:54:29,920
where we take these sequences and we

1354
00:54:28,640 --> 00:54:31,599
compare them against a reference genome

1355
00:54:29,920 --> 00:54:32,880
to identify differences. And when people

1356
00:54:31,599 --> 00:54:34,960
talk about how scientists sequenced the

1357
00:54:32,880 --> 00:54:35,920
human genome in 2003 or so, they're

1358
00:54:34,960 --> 00:54:37,920
talking about the human reference

1359
00:54:35,920 --> 00:54:39,440
genome. And that's me and that's

1360
00:54:37,920 --> 00:54:40,640
represented by this little database icon

1361
00:54:39,440 --> 00:54:42,400
here. And the human reference genome

1362
00:54:40,640 --> 00:54:44,240
isn't one specific person. It's about

1363
00:54:42,400 --> 00:54:45,839
70% of it is an African-American man

1364
00:54:44,240 --> 00:54:48,559
from Buffalo, New York. And the rest is

1365
00:54:45,839 --> 00:54:50,640
other people. Um and so this is the uh

1366
00:54:48,559 --> 00:54:53,520
genome that all all other sequences are

1367
00:54:50,640 --> 00:54:54,960
compared against. and and so um that's

1368
00:54:53,520 --> 00:54:56,480
what we're going to use as the basis for

1369
00:54:54,960 --> 00:54:57,680
the variant calling. It's complicated,

1370
00:54:56,480 --> 00:54:59,359
so we're going to hide it in this magic

1371
00:54:57,680 --> 00:55:00,880
block box. And the output of most

1372
00:54:59,359 --> 00:55:02,079
variant callers is what's called a VCF

1373
00:55:00,880 --> 00:55:03,920
file, which stands for variant call

1374
00:55:02,079 --> 00:55:05,680
format. It is the most cursed file

1375
00:55:03,920 --> 00:55:08,000
format I have ever worked with, and I am

1376
00:55:05,680 --> 00:55:09,119
not going to talk about it anymore.

1377
00:55:08,000 --> 00:55:10,559
But we want to be able to talk about

1378
00:55:09,119 --> 00:55:11,839
individual genetic variants. And so

1379
00:55:10,559 --> 00:55:13,040
that's the change in the genome with

1380
00:55:11,839 --> 00:55:14,160
respect to the reference. And so one of

1381
00:55:13,040 --> 00:55:16,720
the most common formats for doing that

1382
00:55:14,160 --> 00:55:18,559
is this HVS format, which was originally

1383
00:55:16,720 --> 00:55:21,359
defined by the human genome variation

1384
00:55:18,559 --> 00:55:22,800
society. Uh you'll find HTVs everywhere.

1385
00:55:21,359 --> 00:55:24,640
It was in Nick's talk. It's very common

1386
00:55:22,800 --> 00:55:26,240
in clinical reports from genetic tests

1387
00:55:24,640 --> 00:55:28,160
or in scientific papers. You can also

1388
00:55:26,240 --> 00:55:30,800
find it on the AliExpress listing for

1389
00:55:28,160 --> 00:55:32,800
listing page for DNA sequencers. Uh

1390
00:55:30,800 --> 00:55:34,240
here's an HVS string. It has a reference

1391
00:55:32,800 --> 00:55:35,760
accession which is the sequence we are

1392
00:55:34,240 --> 00:55:37,680
indexing into. Here this is human

1393
00:55:35,760 --> 00:55:38,800
chromosome 19. It has a prefix that

1394
00:55:37,680 --> 00:55:40,960
tells us what kind of sequence variant

1395
00:55:38,800 --> 00:55:42,319
it is. Here this is G for genomic. Uh we

1396
00:55:40,960 --> 00:55:43,599
have the position for the sequence and

1397
00:55:42,319 --> 00:55:45,280
then the substitution and the reference

1398
00:55:43,599 --> 00:55:47,760
base. So this is a C to T at whatever

1399
00:55:45,280 --> 00:55:49,040
position that is. Uh HVS can represent a

1400
00:55:47,760 --> 00:55:50,480
lot of variants. Here's the same variant

1401
00:55:49,040 --> 00:55:52,160
in the genome showing with respect to a

1402
00:55:50,480 --> 00:55:53,839
transcript that's a coding coding

1403
00:55:52,160 --> 00:55:56,000
variant and with respect to the protein

1404
00:55:53,839 --> 00:55:57,200
that's these p dot variants. So the DNA

1405
00:55:56,000 --> 00:55:58,400
format as you'll notice is really

1406
00:55:57,200 --> 00:55:59,520
different to the protein format and

1407
00:55:58,400 --> 00:56:01,839
that's by design so that you can't

1408
00:55:59,520 --> 00:56:03,680
easily confuse them. And so HVS is very

1409
00:56:01,839 --> 00:56:05,280
flexible right in fact be too flexible.

1410
00:56:03,680 --> 00:56:06,799
So here's an example where instead of a

1411
00:56:05,280 --> 00:56:08,640
version transcript or something from a

1412
00:56:06,799 --> 00:56:10,160
database we have a gene name. It's not

1413
00:56:08,640 --> 00:56:11,040
specific anymore but it's still valid

1414
00:56:10,160 --> 00:56:12,799
and it's actually much more

1415
00:56:11,040 --> 00:56:14,559
interpretable. The most popular clinical

1416
00:56:12,799 --> 00:56:15,839
reporting software creates HVAC variants

1417
00:56:14,559 --> 00:56:17,520
that look like this since it's much

1418
00:56:15,839 --> 00:56:19,200
easier for our clinicians to see the

1419
00:56:17,520 --> 00:56:20,960
gene name rather than the accession

1420
00:56:19,200 --> 00:56:23,280
number because most people don't rec

1421
00:56:20,960 --> 00:56:25,359
memorize gene accession numbers. Uh

1422
00:56:23,280 --> 00:56:27,200
here's another example of a valid HTVs

1423
00:56:25,359 --> 00:56:29,280
string for a large DNA rearrangement

1424
00:56:27,200 --> 00:56:31,119
involving an insertion and inversion of

1425
00:56:29,280 --> 00:56:32,319
a sequence with uncertain break points.

1426
00:56:31,119 --> 00:56:34,319
You do not want to have to write a

1427
00:56:32,319 --> 00:56:36,319
parser for this. It is very challenging.

1428
00:56:34,319 --> 00:56:37,920
There's an HVS package that parses these

1429
00:56:36,319 --> 00:56:39,920
strings. It's open source. It's written

1430
00:56:37,920 --> 00:56:41,440
in Python. It uses contextf free

1431
00:56:39,920 --> 00:56:43,280
grammarss and a large sequence database

1432
00:56:41,440 --> 00:56:44,880
to do this work. If you are a fan of

1433
00:56:43,280 --> 00:56:46,319
contextf free grammarss and railroad

1434
00:56:44,880 --> 00:56:48,000
diagrams, you are probably at the right

1435
00:56:46,319 --> 00:56:49,119
conference and you will also probably

1436
00:56:48,000 --> 00:56:50,400
enjoy reading their reference

1437
00:56:49,119 --> 00:56:52,160
documentation. There's a lot of

1438
00:56:50,400 --> 00:56:53,440
beautiful stuff in there. Uh to the best

1439
00:56:52,160 --> 00:56:55,119
of my knowledge, there is still no

1440
00:56:53,440 --> 00:56:56,640
software implementation that covers the

1441
00:56:55,119 --> 00:56:57,920
entire spec and I believe the first

1442
00:56:56,640 --> 00:56:59,599
version of the spec was released in

1443
00:56:57,920 --> 00:57:01,839
1996.

1444
00:56:59,599 --> 00:57:04,000
Um it is also very hard to pass this

1445
00:57:01,839 --> 00:57:05,280
around via API and compute on it. I did

1446
00:57:04,000 --> 00:57:06,880
write a regular expressionbased

1447
00:57:05,280 --> 00:57:08,720
implementation of a tiny portion of the

1448
00:57:06,880 --> 00:57:10,480
spec that we needed for our project. I

1449
00:57:08,720 --> 00:57:12,400
think it is terrible. We are still using

1450
00:57:10,480 --> 00:57:14,400
it in production. Also, because we are

1451
00:57:12,400 --> 00:57:17,040
open source software uh developers,

1452
00:57:14,400 --> 00:57:18,880
there is also a PI HTTPS on PI which has

1453
00:57:17,040 --> 00:57:21,280
not been updated I think in 7 years.

1454
00:57:18,880 --> 00:57:22,640
Don't get confused. Now, to try to solve

1455
00:57:21,280 --> 00:57:24,240
the problem to allow us to do things

1456
00:57:22,640 --> 00:57:25,520
like build database APIs that can talk

1457
00:57:24,240 --> 00:57:26,880
to each other more easily, we have a

1458
00:57:25,520 --> 00:57:28,559
different standard called verse or the

1459
00:57:26,880 --> 00:57:30,240
variation representation specification

1460
00:57:28,559 --> 00:57:31,440
which is much more recent. It's from an

1461
00:57:30,240 --> 00:57:33,200
international standards what I call

1462
00:57:31,440 --> 00:57:36,160
G4GH, the Global Alliance for Genomics

1463
00:57:33,200 --> 00:57:37,680
and Health. Um, some when Nick came to

1464
00:57:36,160 --> 00:57:39,520
one of our meetings, he said it reminded

1465
00:57:37,680 --> 00:57:41,440
him of the LE E. I don't know if that's

1466
00:57:39,520 --> 00:57:43,119
good or bad. Um, here's the verse

1467
00:57:41,440 --> 00:57:45,040
representation of that same variant. It

1468
00:57:43,119 --> 00:57:46,319
is punishingly verbose, but it also

1469
00:57:45,040 --> 00:57:48,079
makes it much easier to validate and

1470
00:57:46,319 --> 00:57:49,440
parse these objects. So, the reference

1471
00:57:48,079 --> 00:57:51,040
implementation of verse is also in

1472
00:57:49,440 --> 00:57:52,559
Python. It does not use a context

1473
00:57:51,040 --> 00:57:54,640
grammar. It does not have any railroad

1474
00:57:52,559 --> 00:57:56,559
diagrams. But really complex variants

1475
00:57:54,640 --> 00:57:58,400
can be built up and composed using

1476
00:57:56,559 --> 00:58:00,240
things like JSON schema to make these

1477
00:57:58,400 --> 00:58:01,599
complex variant objects. But you can't

1478
00:58:00,240 --> 00:58:02,960
easily paste this into a clinical

1479
00:58:01,599 --> 00:58:04,400
report. So we still need HTBS for a

1480
00:58:02,960 --> 00:58:05,920
human readable version. And so of course

1481
00:58:04,400 --> 00:58:07,520
there's an HKC for that. But I think

1482
00:58:05,920 --> 00:58:08,799
this is different because it's really an

1483
00:58:07,520 --> 00:58:10,720
example of how we do need different

1484
00:58:08,799 --> 00:58:12,160
standards for different purposes and we

1485
00:58:10,720 --> 00:58:13,680
need standards for people that people

1486
00:58:12,160 --> 00:58:14,960
can read and those should be systematic

1487
00:58:13,680 --> 00:58:16,400
but we need different standards for

1488
00:58:14,960 --> 00:58:17,760
computers. So maybe this problem needing

1489
00:58:16,400 --> 00:58:19,599
to juggle multiple standards for

1490
00:58:17,760 --> 00:58:21,280
different use cases is also true in your

1491
00:58:19,599 --> 00:58:22,720
work that is maybe not in genomics and

1492
00:58:21,280 --> 00:58:26,760
I'm interested in talking more about how

1493
00:58:22,720 --> 00:58:26,760
you manage this. So thank you.

1494
00:58:29,330 --> 00:58:34,559
[Applause]

1495
00:58:31,760 --> 00:58:35,760
Thank you so much Alan. Also as someone

1496
00:58:34,559 --> 00:58:37,440
that has been fighting with Russell

1497
00:58:35,760 --> 00:58:38,720
Keith McGee for the title of the person

1498
00:58:37,440 --> 00:58:39,920
that speaks the fastest in this

1499
00:58:38,720 --> 00:58:42,720
conference I'm worried that the two of

1500
00:58:39,920 --> 00:58:45,920
us have a third contender. Up next on

1501
00:58:42,720 --> 00:58:48,160
this side is unfortunately nobody.

1502
00:58:45,920 --> 00:58:50,480
Oh

1503
00:58:48,160 --> 00:58:52,079
un next on this side is unfortunately

1504
00:58:50,480 --> 00:58:53,520
not a lightning talk presenter. I am

1505
00:58:52,079 --> 00:58:55,520
told that our conference director is not

1506
00:58:53,520 --> 00:58:58,480
nobody because Peter will be closing the

1507
00:58:55,520 --> 00:59:00,960
event. Fortunately there is one more

1508
00:58:58,480 --> 00:59:02,319
lightning talk to go.

1509
00:59:00,960 --> 00:59:05,040
One of the really interesting things

1510
00:59:02,319 --> 00:59:07,200
about lightning talks is some of the

1511
00:59:05,040 --> 00:59:09,520
really like deep metaphors and the ways

1512
00:59:07,200 --> 00:59:11,520
that we can connect technology with the

1513
00:59:09,520 --> 00:59:13,599
rest of our lives.

1514
00:59:11,520 --> 00:59:16,900
G Kid is going to tell us how to get a

1515
00:59:13,599 --> 00:59:22,079
blowfly out of bedroom.

1516
00:59:16,900 --> 00:59:26,040
[Applause]

1517
00:59:22,079 --> 00:59:26,040
How do I follow that?

1518
00:59:26,799 --> 00:59:32,240
Quickly. Right. Thank you, Beno.

1519
00:59:29,920 --> 00:59:34,799
Right. Uh, how to get a blowfly out of a

1520
00:59:32,240 --> 00:59:37,040
bedroom. Um, I am here to tell you how

1521
00:59:34,799 --> 00:59:40,000
to get a blowfly out of a bedroom. Um, I

1522
00:59:37,040 --> 00:59:41,920
regret I was mumble old uh before I

1523
00:59:40,000 --> 00:59:44,559
finally figured this out for myself. Uh,

1524
00:59:41,920 --> 00:59:48,480
quick poll. Um, who here already knows

1525
00:59:44,559 --> 00:59:52,680
how to get a blowfly out of a bedroom?

1526
00:59:48,480 --> 00:59:52,680
Why didn't you tell anyone?

1527
00:59:55,680 --> 01:00:00,640
Right. Um, theory of operation as a

1528
00:59:57,920 --> 01:00:02,079
co-orker of mine might call it. um or uh

1529
01:00:00,640 --> 01:00:05,359
as I've started calling it whenever it's

1530
01:00:02,079 --> 01:00:08,640
a bug, a theory of failure.

1531
01:00:05,359 --> 01:00:10,640
There is a fly here.

1532
01:00:08,640 --> 01:00:13,440
Now, I drew this blue fly. This blowfly,

1533
01:00:10,640 --> 01:00:16,400
uh this blowfly needs work. Um it's a

1534
01:00:13,440 --> 01:00:19,280
bit David Thorne.

1535
01:00:16,400 --> 01:00:21,440
It's only got five legs.

1536
01:00:19,280 --> 01:00:24,559
So, I've asked people to help me with my

1537
01:00:21,440 --> 01:00:27,200
blowflies. Uh thanks, Althia.

1538
01:00:24,559 --> 01:00:30,799
Um but this question needs work. How do

1539
01:00:27,200 --> 01:00:33,040
I get the fly out of my bedroom? Um,

1540
01:00:30,799 --> 01:00:35,359
it's misleading in that it it it sort of

1541
01:00:33,040 --> 01:00:37,200
leads you off into

1542
01:00:35,359 --> 01:00:40,319
like considering methodology before you

1543
01:00:37,200 --> 01:00:42,079
understand the problem.

1544
01:00:40,319 --> 01:00:44,799
A better question is why is the fly

1545
01:00:42,079 --> 01:00:47,359
here?

1546
01:00:44,799 --> 01:00:49,520
Thank you, Lily. Um, and this might just

1547
01:00:47,359 --> 01:00:51,280
be for verbal processing types, but does

1548
01:00:49,520 --> 01:00:53,040
anyone have that thing where they just

1549
01:00:51,280 --> 01:00:54,640
keep mulling over a question with this

1550
01:00:53,040 --> 01:00:55,920
mate and that mate and the other, or if

1551
01:00:54,640 --> 01:00:57,920
they're really unlucky, the same mate

1552
01:00:55,920 --> 01:00:59,440
over and over,

1553
01:00:57,920 --> 01:01:00,799
and they keep asking the same question

1554
01:00:59,440 --> 01:01:03,839
over and over in different ways until

1555
01:01:00,799 --> 01:01:08,280
they finally find themselves expressing

1556
01:01:03,839 --> 01:01:08,280
the problem in terms of the solution?

1557
01:01:09,520 --> 01:01:14,720
This is a much better question.

1558
01:01:12,160 --> 01:01:17,119
Why doesn't the fly get out of my

1559
01:01:14,720 --> 01:01:21,480
bedroom?

1560
01:01:17,119 --> 01:01:21,480
And the answer is regrettably obvious.

1561
01:01:22,079 --> 01:01:25,599
It's dark.

1562
01:01:24,079 --> 01:01:27,599
That's why it keeps flying into the

1563
01:01:25,599 --> 01:01:30,599
walls going

1564
01:01:27,599 --> 01:01:30,599
and

1565
01:01:30,720 --> 01:01:37,040
can't see to get out. It the reason it's

1566
01:01:32,640 --> 01:01:39,760
in our bedroom is it followed us here.

1567
01:01:37,040 --> 01:01:42,000
Or better put, it followed the light

1568
01:01:39,760 --> 01:01:44,960
here.

1569
01:01:42,000 --> 01:01:47,200
And now that you understand

1570
01:01:44,960 --> 01:01:50,640
what happened, the methodology is

1571
01:01:47,200 --> 01:01:52,960
obvious, right?

1572
01:01:50,640 --> 01:01:54,720
Thanks, Maya. Sorry, I'm changing

1573
01:01:52,960 --> 01:01:56,400
blowflies repeatedly and I'm failing to

1574
01:01:54,720 --> 01:01:58,079
give credit. I'll fix that at the end.

1575
01:01:56,400 --> 01:02:00,480
Um, so what you do is you walk out of

1576
01:01:58,079 --> 01:02:02,319
your bedroom to the next room out, you

1577
01:02:00,480 --> 01:02:05,839
turn the light on, and you wait until

1578
01:02:02,319 --> 01:02:09,200
the fly joins you.

1579
01:02:05,839 --> 01:02:10,640
And then you move to the next room along

1580
01:02:09,200 --> 01:02:12,400
and you turn the light on there. And

1581
01:02:10,640 --> 01:02:15,960
then you walk back to the other room and

1582
01:02:12,400 --> 01:02:15,960
you turn the light off.

1583
01:02:16,480 --> 01:02:20,599
Lather, rinse, and repeat.

1584
01:02:24,000 --> 01:02:28,000
I ran out of blowflies at this point.

1585
01:02:26,400 --> 01:02:29,760
This is my blowfly again. You lather,

1586
01:02:28,000 --> 01:02:31,359
rinse, and repeat until you lure the

1587
01:02:29,760 --> 01:02:32,880
blowfly

1588
01:02:31,359 --> 01:02:35,200
to the most distant point of the house

1589
01:02:32,880 --> 01:02:39,240
from your bedroom.

1590
01:02:35,200 --> 01:02:39,240
And then you turn that light off

1591
01:02:39,520 --> 01:02:44,200
and tiptoe back to your bedroom.

1592
01:02:47,760 --> 01:02:54,079
Uh so yeah, too long didn't read. Lure

1593
01:02:50,960 --> 01:02:56,480
it out with light.

1594
01:02:54,079 --> 01:02:58,799
And I'd like to thank my Blowfly

1595
01:02:56,480 --> 01:03:02,599
collaborators

1596
01:02:58,799 --> 01:03:02,599
and thank you for listening.

1597
01:03:07,280 --> 01:03:08,559
You done?

1598
01:03:08,079 --> 01:03:12,839
Oh, yeah.

1599
01:03:08,559 --> 01:03:12,839
Oh, thanks G. That was great.

1600
01:03:15,930 --> 01:03:21,019
[Music]