WEBVTT

1
00:00:03.705 --> 00:00:04.045
<00:00:03.705>All <00:00:04.011>right.

2
00:00:04.465 --> 00:00:06.205
<00:00:04.465>Ah, <00:00:04.698>it's <00:00:04.931>so <00:00:05.165>good <00:00:05.398>to <00:00:05.631>be <00:00:05.865>here.

3
00:00:06.205 --> 00:00:07.485
<00:00:06.205>It's <00:00:06.333>so <00:00:06.461>good <00:00:06.589>to <00:00:06.717>be <00:00:06.845>here <00:00:06.973>with <00:00:07.101>all <00:00:07.229>of <00:00:07.357>you.

4
00:00:07.485 --> 00:00:09.765
<00:00:07.485>Thank <00:00:07.640>you <00:00:07.796>very <00:00:07.951>much <00:00:08.107>for <00:00:08.262>joining <00:00:08.418>me, <00:00:08.573>uh, <00:00:08.729>this <00:00:08.885>afternoon.

5
00:00:10.345 --> 00:00:12.085
<00:00:10.345>Uh, <00:00:10.545>and <00:00:10.745>thank <00:00:10.945>you, <00:00:11.145>JP <00:00:11.345>for <00:00:11.545>the <00:00:11.745>introduction.

6
00:00:12.105 --> 00:00:13.645
<00:00:12.105>Uh, <00:00:12.303>again, <00:00:12.501>my <00:00:12.699>name's <00:00:12.897>Van. <00:00:13.095>I'm <00:00:13.293>a <00:00:13.491>developer

7
00:00:13.645 --> 00:00:15.005
<00:00:13.645>relations <00:00:14.053>engineer <00:00:14.461>at <00:00:14.869>DataStax.

8
00:00:15.385 --> 00:00:16.765
<00:00:15.385>Uh, <00:00:15.633>I <00:00:15.881>was <00:00:16.130>previously <00:00:16.378>at <00:00:16.627>Sona.

9
00:00:16.765 --> 00:00:18.485
<00:00:16.765>They're <00:00:16.965>all, <00:00:17.165>they're <00:00:17.365>wonderful <00:00:17.565>people. <00:00:17.765>Go <00:00:17.965>see <00:00:18.165>them.

10
00:00:18.485 --> 00:00:20.085
<00:00:18.485>They've <00:00:18.660>got <00:00:18.835>a <00:00:19.009>little <00:00:19.185>booth <00:00:19.360>over <00:00:19.535>the <00:00:19.710>back <00:00:19.884>there.

11
00:00:20.625 --> 00:00:22.205
<00:00:20.625>Um, <00:00:20.858>I <00:00:21.091>work <00:00:21.325>at <00:00:21.558>Datas <00:00:21.791>stacks, <00:00:22.025>uh,

12
00:00:22.205 --> 00:00:25.725
<00:00:22.205>where <00:00:22.345>we <00:00:22.485>are <00:00:22.625>building <00:00:22.764>a <00:00:22.904>platform <00:00:23.044>for, <00:00:23.185>uh, <00:00:23.325>generative <00:00:23.464>ai, <00:00:23.604>uh,

13
00:00:25.785 --> 00:00:28.605
<00:00:25.785>so <00:00:25.912>that <00:00:26.039>you, <00:00:26.166>uh, <00:00:26.294>to <00:00:26.421>help <00:00:26.548>you <00:00:26.675>build <00:00:26.803>on <00:00:26.930>generative <00:00:27.057>AI <00:00:27.185>better.

14
00:00:28.605 --> 00:00:31.125
<00:00:28.605>We <00:00:28.745>have <00:00:28.885>Astro <00:00:29.025>db, <00:00:29.165>which <00:00:29.305>is <00:00:29.445>a <00:00:29.585>great <00:00:29.725>vector <00:00:29.865>store, <00:00:30.005>uh,

15
00:00:31.125 --> 00:00:33.565
<00:00:31.125>and <00:00:31.265>we <00:00:31.405>have <00:00:31.545>Lang <00:00:31.685>Flow, <00:00:31.825>uh, <00:00:31.965>which <00:00:32.104>is <00:00:32.244>a <00:00:32.384>great <00:00:32.525>drag

16
00:00:33.565 --> 00:00:36.205
<00:00:33.565>and <00:00:33.765>drop <00:00:33.964>builder, <00:00:34.165>uh, <00:00:34.364>for, <00:00:34.565>um, <00:00:34.765>generative <00:00:34.964>ai.

17
00:00:36.375 --> 00:00:37.485
<00:00:36.375>Check <00:00:36.708>those <00:00:37.041>things <00:00:37.374>out,

18
00:00:37.705 --> 00:00:39.965
<00:00:37.705>or <00:00:37.860>come <00:00:38.016>see <00:00:38.171>us <00:00:38.327>at <00:00:38.482>the <00:00:38.638>Data <00:00:38.793>Stacks <00:00:38.949>booth, <00:00:39.105>uh,

19
00:00:39.965 --> 00:00:41.365
<00:00:39.965>where <00:00:40.280>in <00:00:40.595>the <00:00:40.910>other <00:00:41.225>hall.

20
00:00:41.705 --> 00:00:42.845
<00:00:41.705>Um, <00:00:41.961>come <00:00:42.218>say <00:00:42.474>hi <00:00:42.731>there.

21
00:00:43.385 --> 00:00:45.045
<00:00:43.385>Uh, <00:00:43.585>if <00:00:43.785>you <00:00:43.985>wanna <00:00:44.184>find <00:00:44.385>me <00:00:44.585>online <00:00:44.785>anywhere,

22
00:00:45.265 --> 00:00:47.765
<00:00:45.265>I'm <00:00:45.440>effectively <00:00:45.615>Phil <00:00:45.790>Nash, <00:00:45.965>uh, <00:00:46.140>all <00:00:46.315>over <00:00:46.490>the <00:00:46.665>internet.

23
00:00:48.025 --> 00:00:51.565
<00:00:48.025>Um, <00:00:48.165>or <00:00:48.305>even <00:00:48.445>like, <00:00:48.585>it's <00:00:48.725>my <00:00:48.864>favorite <00:00:49.004>URLI <00:00:49.144>got, <00:00:49.285>you <00:00:49.425>know,

24
00:00:51.565 --> 00:00:52.765
<00:00:51.565>Phil, <00:00:51.835>Phil <00:00:52.105>Nash, <00:00:52.375>that <00:00:52.645>works.

25
00:00:53.145 --> 00:00:54.285
<00:00:53.145>Domain <00:00:53.401>hacking's <00:00:53.658>the <00:00:53.914>best <00:00:54.171>thing.

26
00:00:54.595 --> 00:00:57.405
<00:00:54.595>It's <00:00:54.875>always <00:00:55.155>been <00:00:55.435>my <00:00:55.714>favorite. <00:00:55.995>Alright.

27
00:00:57.405 --> 00:00:58.885
<00:00:57.405>All <00:00:57.849>right, <00:00:58.293>so <00:00:58.736>going

28
00:00:58.885 --> 00:01:01.845
<00:00:58.885>to <00:00:59.085>look <00:00:59.285>into <00:00:59.485>building <00:00:59.684>Rag <00:00:59.885>from <00:01:00.085>Scratch <00:01:00.285>today,

29
00:01:02.065 --> 00:01:04.085
<00:01:02.065>and <00:01:02.345>I'm <00:01:02.625>glad <00:01:02.905>to <00:01:03.184>hear <00:01:03.464>that

30
00:01:04.145 --> 00:01:05.245
<00:01:04.144>or <00:01:04.343>see <00:01:04.541>that, <00:01:04.739>uh, <00:01:04.937>a <00:01:05.135>bunch

31
00:01:05.245 --> 00:01:07.605
<00:01:05.245>of <00:01:05.445>you <00:01:05.645>actually <00:01:05.845>already <00:01:06.045>know, <00:01:06.245>uh, <00:01:06.445>about <00:01:06.645>Rag.

32
00:01:07.945 --> 00:01:10.685
<00:01:07.945>Uh, <00:01:08.052>but <00:01:08.160>I <00:01:08.268>wanted <00:01:08.375>to <00:01:08.483>kind <00:01:08.591>of <00:01:08.698>dig <00:01:08.806>into <00:01:08.914>this <00:01:09.021>to <00:01:09.129>break <00:01:09.237>it <00:01:09.345>down

33
00:01:10.685 --> 00:01:13.645
<00:01:10.685>to <00:01:10.918>kind <00:01:11.151>of <00:01:11.385>better <00:01:11.618>understand <00:01:11.851>why <00:01:12.085>RAG

34
00:01:13.705 --> 00:01:16.085
<00:01:13.705>and <00:01:13.985>why <00:01:14.265>Similarity <00:01:14.545>Search <00:01:14.825>actually <00:01:15.105>works.

35
00:01:16.585 --> 00:01:18.365
<00:01:16.585>Um, <00:01:16.760>we're <00:01:16.935>gonna <00:01:17.110>do <00:01:17.285>that <00:01:17.460>through <00:01:17.635>some <00:01:17.810>live <00:01:17.985>code

36
00:01:18.945 --> 00:01:21.885
<00:01:18.945>and, <00:01:19.178>um, <00:01:19.411>hopefully <00:01:19.645>it <00:01:19.878>just <00:01:20.111>allows <00:01:20.345>it

37
00:01:21.885 --> 00:01:23.485
<00:01:21.885>to <00:01:22.085>sit <00:01:22.285>in <00:01:22.485>your <00:01:22.685>minds <00:01:22.885>better <00:01:23.085>as <00:01:23.285>to

38
00:01:23.485 --> 00:01:25.165
<00:01:23.485>how <00:01:23.718>it <00:01:23.951>all <00:01:24.185>a <00:01:24.418>actually <00:01:24.651>hangs <00:01:24.885>together.

39
00:01:25.985 --> 00:01:27.365
<00:01:25.985>Um, <00:01:26.295>that's <00:01:26.606>my <00:01:26.916>intention <00:01:27.226>here.

40
00:01:27.505 --> 00:01:30.885
<00:01:27.505>Uh, <00:01:27.645>and <00:01:27.785>we've <00:01:27.925>got <00:01:28.065>28 <00:01:28.205>minutes <00:01:28.345>to <00:01:28.485>do <00:01:28.625>so, <00:01:28.765>so <00:01:28.905>hopefully,

41
00:01:30.885 --> 00:01:33.445
<00:01:30.885>you <00:01:31.040>know, <00:01:31.196>that <00:01:31.351>sort <00:01:31.507>of <00:01:31.662>problem <00:01:31.818>with <00:01:31.973>large <00:01:32.129>language <00:01:32.285>models,

42
00:01:33.945 --> 00:01:36.765
<00:01:33.945>uh, <00:01:34.145>they <00:01:34.345>are <00:01:34.544>incredible <00:01:34.744>bits <00:01:34.945>of, <00:01:35.145>uh, <00:01:35.345>technology.

43
00:01:36.925 --> 00:01:38.805
<00:01:36.925>I, <00:01:37.158>I'm <00:01:37.391>sure <00:01:37.625>we <00:01:37.858>all <00:01:38.091>agree, <00:01:38.325>uh,

44
00:01:38.805 --> 00:01:41.125
<00:01:38.805>but <00:01:39.005>large <00:01:39.205>language <00:01:39.405>models <00:01:39.605>only <00:01:39.805>know <00:01:40.005>as <00:01:40.205>much

45
00:01:41.125 --> 00:01:42.845
<00:01:41.125>as <00:01:41.358>they've <00:01:41.591>been <00:01:41.825>told, <00:01:42.058>only <00:01:42.291>as <00:01:42.525>know

46
00:01:42.905 --> 00:01:44.125
<00:01:42.905>as <00:01:43.088>much <00:01:43.271>as <00:01:43.454>they've <00:01:43.637>been <00:01:43.820>trained <00:01:44.003>on.

47
00:01:44.465 --> 00:01:46.285
<00:01:44.465>Uh, <00:01:44.665>and <00:01:44.865>that <00:01:45.065>has <00:01:45.265>both <00:01:45.465>a <00:01:45.665>cutoff <00:01:45.865>date

48
00:01:46.585 --> 00:01:49.365
<00:01:46.585>and <00:01:46.760>a <00:01:46.935>lack <00:01:47.110>of, <00:01:47.285>um, <00:01:47.460>you <00:01:47.635>know, <00:01:47.810>private <00:01:47.985>data.

49
00:01:50.025 --> 00:01:52.365
<00:01:50.025>Uh, <00:01:50.200>these <00:01:50.375>are <00:01:50.550>our <00:01:50.725>issues <00:01:50.900>with, <00:01:51.075>uh, <00:01:51.250>language <00:01:51.425>models

50
00:01:52.825 --> 00:01:56.005
<00:01:52.825>and, <00:01:52.952>uh, <00:01:53.079>that <00:01:53.206>kind <00:01:53.334>of <00:01:53.461>can <00:01:53.588>be <00:01:53.715>shown, <00:01:53.843>uh, <00:01:53.970>with <00:01:54.097>my, <00:01:54.225>uh,

51
00:01:56.075 --> 00:01:59.125
<00:01:56.075>very <00:01:56.202>basic <00:01:56.329>chat <00:01:56.456>that <00:01:56.584>I've <00:01:56.711>built <00:01:56.838>for <00:01:56.965>a <00:01:57.093>demo, <00:01:57.220>uh, <00:01:57.347>for <00:01:57.475>this.

52
00:01:59.545 --> 00:02:01.325
<00:01:59.545>Uh, <00:01:59.720>so <00:01:59.895>this <00:02:00.070>is <00:02:00.245>an <00:02:00.420>event <00:02:00.595>chat <00:02:00.770>that's <00:02:00.945>supposed

53
00:02:01.325 --> 00:02:03.045
<00:02:01.325>to <00:02:01.675>answer <00:02:02.025>questions <00:02:02.375>about, <00:02:02.725>uh,

54
00:02:03.045 --> 00:02:04.325
<00:02:03.045>the <00:02:03.333>We <00:02:03.621>Are <00:02:03.909>Developers <00:02:04.197>conference.

55
00:02:04.705 --> 00:02:07.765
<00:02:04.705>Um, <00:02:04.860>it's <00:02:05.016>currently <00:02:05.171>hooked <00:02:05.327>up <00:02:05.482>to, <00:02:05.638>uh, <00:02:05.793>straight <00:02:05.949>into, <00:02:06.105>uh,

56
00:02:07.765 --> 00:02:11.405
<00:02:07.765>Google, <00:02:07.965>Gemini, <00:02:08.164>uh, <00:02:08.365>Gemini <00:02:08.565>Flash, <00:02:08.764>I <00:02:08.965>think <00:02:09.164>1.5.

57
00:02:11.825 --> 00:02:13.125
<00:02:11.825>And <00:02:12.019>so <00:02:12.215>if <00:02:12.410>I <00:02:12.605>ask <00:02:12.800>the <00:02:12.995>chat,

58
00:02:13.545 --> 00:02:15.605
<00:02:13.545>who <00:02:13.778>is <00:02:14.011>talking <00:02:14.244>about <00:02:14.478>retrieval <00:02:14.711>augmented <00:02:14.945>generation,

59
00:02:16.145 --> 00:02:17.205
<00:02:16.145>we <00:02:16.383>know <00:02:16.622>it <00:02:16.860>doesn't <00:02:17.099>know.

60
00:02:17.725 --> 00:02:20.045
<00:02:17.725>Um, <00:02:17.924>it <00:02:18.125>has <00:02:18.325>more, <00:02:18.525>it <00:02:18.725>needs <00:02:18.924>more <00:02:19.125>information.

61
00:02:20.425 --> 00:02:23.365
<00:02:20.425>Uh, <00:02:20.658>Gemini <00:02:20.891>1.5 <00:02:21.125>Flash <00:02:21.358>was <00:02:21.591>trained <00:02:21.825>well

62
00:02:23.365 --> 00:02:25.245
<00:02:23.365>before <00:02:23.520>we <00:02:23.676>had <00:02:23.831>the, <00:02:23.987>uh, <00:02:24.142>schedule <00:02:24.298>up <00:02:24.453>for <00:02:24.609>this <00:02:24.765>event.

63
00:02:25.825 --> 00:02:30.325
<00:02:25.825>Um, <00:02:25.952>one <00:02:26.079>of <00:02:26.206>the <00:02:26.334>other <00:02:26.461>things <00:02:26.588>is, <00:02:26.715>uh, <00:02:26.843>you <00:02:26.970>know, <00:02:27.097>uh, <00:02:27.225>the,

64
00:02:30.625 --> 00:02:32.885
<00:02:30.625>it <00:02:30.765>used <00:02:30.905>to <00:02:31.045>the <00:02:31.185>world <00:02:31.325>used <00:02:31.465>to <00:02:31.605>be <00:02:31.745>about <00:02:31.885>Googling <00:02:32.025>yourself

65
00:02:32.885 --> 00:02:34.605
<00:02:32.885>to <00:02:33.040>find <00:02:33.196>out <00:02:33.351>if <00:02:33.507>you <00:02:33.662>showed <00:02:33.818>up <00:02:33.973>in <00:02:34.129>those <00:02:34.285>results.

66
00:02:34.625 --> 00:02:36.205
<00:02:34.625>Now, <00:02:34.825>it's <00:02:35.025>about <00:02:35.225>asking <00:02:35.425>an <00:02:35.625>LLM <00:02:35.825>if <00:02:36.025>you

67
00:02:36.205 --> 00:02:37.245
<00:02:36.204>know, <00:02:36.361>if <00:02:36.517>they <00:02:36.673>know <00:02:36.829>who <00:02:36.985>you <00:02:37.141>are.

68
00:02:37.585 --> 00:02:38.805
<00:02:37.585>And <00:02:37.951>it's <00:02:38.317>really <00:02:38.683>disappointing

69
00:02:38.805 --> 00:02:40.885
<00:02:38.805>that <00:02:38.980>they <00:02:39.155>simply <00:02:39.330>does <00:02:39.505>not <00:02:39.680>know <00:02:39.855>who <00:02:40.030>I <00:02:40.205>am.

70
00:02:41.305 --> 00:02:44.765
<00:02:41.305>Uh, <00:02:41.432>I <00:02:41.559>intend <00:02:41.686>to <00:02:41.814>change <00:02:41.941>that. <00:02:42.068>Um, <00:02:42.195>and <00:02:42.323>so <00:02:42.450>that's <00:02:42.577>our <00:02:42.705>problem.

71
00:02:45.265 --> 00:02:48.005
<00:02:45.265>Uh, <00:02:45.381>we <00:02:45.498>have <00:02:45.614>this, <00:02:45.731>uh, <00:02:45.848>this <00:02:45.964>is <00:02:46.081>built <00:02:46.198>together <00:02:46.315>in <00:02:46.431>a, <00:02:46.548>in <00:02:46.665>a,

72
00:02:48.005 --> 00:02:50.125
<00:02:48.005>just <00:02:48.238>a <00:02:48.471>JavaScript <00:02:48.704>and <00:02:48.938>no <00:02:49.171>JS <00:02:49.405>application.

73
00:02:50.155 --> 00:02:51.565
<00:02:50.155>It's <00:02:50.789>express, <00:02:51.424>uh,

74
00:02:51.665 --> 00:02:55.245
<00:02:51.665>and <00:02:51.792>right <00:02:51.919>now <00:02:52.046>we <00:02:52.174>are <00:02:52.301>calling <00:02:52.428>on <00:02:52.555>this, <00:02:52.683>uh, <00:02:52.810>find <00:02:52.937>similar <00:02:53.065>talk,

75
00:02:55.825 --> 00:03:00.645
<00:02:55.825>um, <00:02:56.000>uh, <00:02:56.174>method, <00:02:56.350>uh, <00:02:56.524>which <00:02:56.700>is <00:02:56.875>currently <00:02:57.049>not <00:02:57.225>implemented.

76
00:03:00.785 --> 00:03:02.765
<00:03:00.785>So <00:03:00.940>we'll <00:03:01.096>come <00:03:01.251>back <00:03:01.407>to <00:03:01.562>that <00:03:01.718>in <00:03:01.873>just <00:03:02.029>a <00:03:02.185>bit.

77
00:03:02.765 --> 00:03:06.485
<00:03:02.765>That's <00:03:02.940>why <00:03:03.114>we <00:03:03.290>get <00:03:03.464>no <00:03:03.640>context. <00:03:03.815>And <00:03:03.989>so, <00:03:04.165>right.

78
00:03:06.505 --> 00:03:10.545
<00:03:06.505>The <00:03:06.738>solution, <00:03:06.971>uh, <00:03:07.204>for <00:03:07.438>large <00:03:07.671>language <00:03:07.905>models,

79
00:03:10.605 --> 00:03:13.945
<00:03:10.605>not <00:03:10.760>knowing <00:03:10.916>things, <00:03:11.071>is <00:03:11.227>to <00:03:11.382>tell <00:03:11.538>them <00:03:11.693>things <00:03:11.849>by <00:03:12.005>retrieving,

80
00:03:14.285 --> 00:03:17.985
<00:03:14.285>uh, <00:03:14.440>and <00:03:14.596>augmenting <00:03:14.751>our <00:03:14.907>prompts, <00:03:15.062>uh, <00:03:15.218>to <00:03:15.373>allow <00:03:15.529>for <00:03:15.685>generation

81
00:03:18.615 --> 00:03:22.105
<00:03:18.615>with, <00:03:18.895>uh, <00:03:19.175>more <00:03:19.455>knowledge <00:03:19.735>retrieval, <00:03:20.015>augment,

82
00:03:22.625 --> 00:03:23.705
<00:03:22.625>retrieval, <00:03:23.111>augmented <00:03:23.596>generation.

83
00:03:24.565 --> 00:03:27.985
<00:03:24.565>Um, <00:03:24.720>the <00:03:24.876>idea <00:03:25.031>behind <00:03:25.187>rag, <00:03:25.342>uh, <00:03:25.498>here <00:03:25.653>is <00:03:25.809>of <00:03:25.965>course

84
00:03:27.985 --> 00:03:29.385
<00:03:27.985>that <00:03:28.300>we <00:03:28.615>store <00:03:28.930>our <00:03:29.245>data.

85
00:03:29.525 --> 00:03:30.985
<00:03:29.525>We <00:03:29.689>take <00:03:29.853>all <00:03:30.017>the <00:03:30.182>data. <00:03:30.346>We <00:03:30.510>want <00:03:30.674>the <00:03:30.839>model

86
00:03:30.985 --> 00:03:33.465
<00:03:30.985>to <00:03:31.140>eventually <00:03:31.296>know <00:03:31.451>at <00:03:31.607>some <00:03:31.762>point, <00:03:31.918>uh, <00:03:32.073>store <00:03:32.229>that <00:03:32.385>away.

87
00:03:33.465 --> 00:03:36.585
<00:03:33.465>Somewhere <00:03:33.620>a <00:03:33.776>user <00:03:33.931>makes <00:03:34.087>a <00:03:34.242>query <00:03:34.398>in <00:03:34.553>the <00:03:34.709>natural <00:03:34.865>language.

88
00:03:36.845 --> 00:03:40.745
<00:03:36.845>Uh, <00:03:37.020>something <00:03:37.195>like, <00:03:37.370>who's <00:03:37.545>speaking <00:03:37.720>about, <00:03:37.895>uh, <00:03:38.070>augmented, <00:03:38.245>uh,

89
00:03:41.025 --> 00:03:43.065
<00:03:41.025>retrieval, <00:03:41.258>augmented <00:03:41.491>generation, <00:03:41.725>who <00:03:41.958>is <00:03:42.191>Phil <00:03:42.425>Ash?

90
00:03:43.645 --> 00:03:45.065
<00:03:43.645>We <00:03:43.964>retrieve <00:03:44.284>that <00:03:44.603>relevant <00:03:44.923>data.

91
00:03:45.685 --> 00:03:47.705
<00:03:45.685>We <00:03:45.860>provide <00:03:46.035>that <00:03:46.210>data <00:03:46.385>as <00:03:46.560>context <00:03:46.735>to <00:03:46.910>the <00:03:47.085>model

92
00:03:48.175 --> 00:03:49.265
<00:03:48.175>with <00:03:48.502>the <00:03:48.829>original <00:03:49.155>query.

93
00:03:49.265 --> 00:03:50.305
<00:03:49.265>So <00:03:49.421>the <00:03:49.577>model <00:03:49.733>knows <00:03:49.889>what <00:03:50.045>to <00:03:50.201>answer.

94
00:03:50.965 --> 00:03:53.745
<00:03:50.965>The <00:03:51.165>model <00:03:51.365>generates <00:03:51.565>a <00:03:51.765>response, <00:03:51.965>something <00:03:52.165>else <00:03:52.365>happens.

95
00:03:53.765 --> 00:03:55.505
<00:03:53.765>Profit, <00:03:54.231>right? <00:03:54.698>Everyone's <00:03:55.165>happy.

96
00:03:56.525 --> 00:04:00.385
<00:03:56.525>Um, <00:03:56.680>and <00:03:56.836>the <00:03:56.991>problem <00:03:57.147>with <00:03:57.302>this, <00:03:57.458>uh, <00:03:57.613>really <00:03:57.769>the <00:03:57.925>bit

97
00:04:00.385 --> 00:04:01.465
<00:04:00.385>that <00:04:00.628>I <00:04:00.871>didn't <00:04:01.114>quite <00:04:01.357>understand

98
00:04:01.465 --> 00:04:03.785
<00:04:01.465>before <00:04:01.640>kind <00:04:01.815>of <00:04:01.990>digging <00:04:02.165>into <00:04:02.340>it, <00:04:02.515>was <00:04:02.690>really <00:04:02.865>how

99
00:04:03.785 --> 00:04:06.705
<00:04:03.785>that <00:04:04.018>retrieving <00:04:04.251>relevant <00:04:04.484>data, <00:04:04.718>uh, <00:04:04.951>comes <00:04:05.185>about

100
00:04:06.765 --> 00:04:07.785
<00:04:06.765>how <00:04:06.994>do <00:04:07.224>we <00:04:07.453>achieve <00:04:07.683>that?

101
00:04:08.455 --> 00:04:11.345
<00:04:08.455>Because <00:04:08.688>we're <00:04:08.921>not <00:04:09.155>doing <00:04:09.388>regular <00:04:09.621>search <00:04:09.855>here.

102
00:04:12.085 --> 00:04:13.905
<00:04:12.085>We <00:04:12.365>have <00:04:12.645>a <00:04:12.925>natural <00:04:13.205>language <00:04:13.485>query.

103
00:04:14.965 --> 00:04:16.585
<00:04:14.965>So, <00:04:15.315>so <00:04:15.665>retrieval <00:04:16.015>is <00:04:16.365>search,

104
00:04:16.805 --> 00:04:19.145
<00:04:16.805>but <00:04:17.005>not <00:04:17.204>natural, <00:04:17.405>not <00:04:17.605>not <00:04:17.805>normal <00:04:18.005>keyword <00:04:18.204>search

105
00:04:19.145 --> 00:04:20.185
<00:04:19.144>or <00:04:19.378>other <00:04:19.613>kind <00:04:19.846>of <00:04:20.081>filtering.

106
00:04:20.725 --> 00:04:22.185
<00:04:20.725>Uh, <00:04:20.987>and <00:04:21.250>then <00:04:21.513>augmentation <00:04:21.776>is <00:04:22.039>prompt

107
00:04:22.205 --> 00:04:23.465
<00:04:22.205>and <00:04:22.488>generating <00:04:22.772>is <00:04:23.055>the <00:04:23.339>model.

108
00:04:24.045 --> 00:04:26.505
<00:04:24.045>Um, <00:04:24.278>it's <00:04:24.511>search, <00:04:24.745>but <00:04:24.978>it's <00:04:25.211>not <00:04:25.445>normal.

109
00:04:26.655 --> 00:04:29.305
<00:04:26.655>It's <00:04:26.854>not <00:04:27.054>keyword <00:04:27.255>search. <00:04:27.455>It's <00:04:27.655>search <00:04:27.854>by <00:04:28.054>similarity.

110
00:04:30.295 --> 00:04:31.985
<00:04:30.295>It's, <00:04:30.528>um, <00:04:30.761>I <00:04:30.995>must <00:04:31.228>think <00:04:31.461>about <00:04:31.695>it.

111
00:04:32.015 --> 00:04:33.785
<00:04:32.015>It's, <00:04:32.248>it's <00:04:32.481>like <00:04:32.715>searching <00:04:32.948>by <00:04:33.181>vibes, <00:04:33.414>right?

112
00:04:33.785 --> 00:04:35.905
<00:04:33.784>It's <00:04:33.901>not <00:04:34.018>like, <00:04:34.135>well, <00:04:34.251>what <00:04:34.368>do <00:04:34.484>we, <00:04:34.601>we <00:04:34.718>feel <00:04:34.835>like <00:04:34.951>this <00:04:35.068>is <00:04:35.184>about?

113
00:04:36.685 --> 00:04:39.345
<00:04:36.685>Um, <00:04:36.885>and <00:04:37.085>so <00:04:37.285>to <00:04:37.485>achieve <00:04:37.685>that, <00:04:37.885>the, <00:04:38.085>the,

114
00:04:39.345 --> 00:04:42.425
<00:04:39.345>the <00:04:39.500>real <00:04:39.656>question <00:04:39.811>then <00:04:39.967>becomes, <00:04:40.122>how <00:04:40.278>do <00:04:40.433>we <00:04:40.589>capture <00:04:40.745>meaning

115
00:04:43.045 --> 00:04:44.585
<00:04:43.045>in <00:04:43.218>a <00:04:43.391>way <00:04:43.564>that <00:04:43.738>we <00:04:43.911>can <00:04:44.084>search <00:04:44.257>for <00:04:44.431>similar?

116
00:04:44.585 --> 00:04:48.955
<00:04:44.585>Meaning, <00:04:44.760>how <00:04:44.935>do <00:04:45.109>we <00:04:45.284>turn <00:04:45.460>our <00:04:45.635>text <00:04:45.810>or <00:04:45.984>images

117
00:04:49.015 --> 00:04:53.075
<00:04:49.015>or <00:04:49.170>other <00:04:49.326>media <00:04:49.481>into <00:04:49.637>something <00:04:49.792>that <00:04:49.948>is <00:04:50.103>just <00:04:50.259>meaning <00:04:50.414>so

118
00:04:53.075 --> 00:04:54.915
<00:04:53.075>that <00:04:53.230>we <00:04:53.386>can <00:04:53.541>search <00:04:53.697>for <00:04:53.852>it <00:04:54.008>in <00:04:54.163>a <00:04:54.319>similar <00:04:54.474>way?

119
00:04:56.175 --> 00:05:00.315
<00:04:56.175>Uh, <00:04:56.330>and <00:04:56.486>that <00:04:56.641>brings <00:04:56.797>us, <00:04:56.952>uh, <00:04:57.108>to <00:04:57.263>vector <00:04:57.419>embeddings, <00:04:57.575>uh,

120
00:05:00.315 --> 00:05:03.115
<00:05:00.315>vector <00:05:00.548>embeddings <00:05:00.781>are <00:05:01.015>have <00:05:01.248>a <00:05:01.481>thing <00:05:01.715>there.

121
00:05:03.115 --> 00:05:05.115
<00:05:03.115>Yeah. <00:05:03.395>Basically <00:05:03.675>a <00:05:03.955>list <00:05:04.235>of <00:05:04.515>numbers

122
00:05:05.265 --> 00:05:07.595
<00:05:05.265>that <00:05:05.440>represents <00:05:05.615>the <00:05:05.789>meaning <00:05:05.965>of <00:05:06.140>a <00:05:06.315>body <00:05:06.490>of <00:05:06.664>text

123
00:05:07.855 --> 00:05:09.515
<00:05:07.855>or <00:05:08.321>indeed, <00:05:08.788>other <00:05:09.255>media.

124
00:05:09.855 --> 00:05:11.795
<00:05:09.855>I'm <00:05:10.030>mostly <00:05:10.205>talking <00:05:10.380>about <00:05:10.555>text <00:05:10.730>today. <00:05:10.905>That's <00:05:11.080>my <00:05:11.255>example.

125
00:05:11.975 --> 00:05:15.515
<00:05:11.975>So <00:05:12.102>if <00:05:12.229>I <00:05:12.356>say <00:05:12.484>text, <00:05:12.611>I <00:05:12.738>mean <00:05:12.865>text <00:05:12.993>and <00:05:13.120>also <00:05:13.247>other <00:05:13.375>media.

126
00:05:16.415 --> 00:05:17.795
<00:05:16.415>But <00:05:16.570>a <00:05:16.725>veteran <00:05:16.880>beding <00:05:17.036>is <00:05:17.191>this <00:05:17.346>list <00:05:17.501>of <00:05:17.657>numbers

127
00:05:17.795 --> 00:05:19.835
<00:05:17.795>that <00:05:17.970>represents <00:05:18.145>the <00:05:18.320>meaning <00:05:18.495>of <00:05:18.670>a <00:05:18.845>body <00:05:19.020>of <00:05:19.195>text.

128
00:05:20.815 --> 00:05:22.115
<00:05:20.815>And <00:05:21.010>the <00:05:21.205>good <00:05:21.400>thing <00:05:21.595>about <00:05:21.790>numbers <00:05:21.985>is

129
00:05:22.115 --> 00:05:24.955
<00:05:22.115>that <00:05:22.290>computers <00:05:22.465>are <00:05:22.640>really <00:05:22.815>good <00:05:22.990>at <00:05:23.165>dealing <00:05:23.340>with <00:05:23.515>those.

130
00:05:25.115 --> 00:05:28.475
<00:05:25.115>I <00:05:25.270>appreciate <00:05:25.426>that <00:05:25.581>from <00:05:25.737>them. <00:05:25.892>So <00:05:26.048>let's <00:05:26.203>create <00:05:26.359>our <00:05:26.515>own,

131
00:05:28.855 --> 00:05:29.995
<00:05:28.855>uh, <00:05:29.368>vector <00:05:29.881>embedding.

132
00:05:29.995 --> 00:05:31.435
<00:05:29.995>That's <00:05:30.254>the <00:05:30.513>kind <00:05:30.772>of <00:05:31.031>idea <00:05:31.291>here.

133
00:05:31.955 --> 00:05:33.635
<00:05:31.955>I <00:05:32.130>wanna <00:05:32.305>do <00:05:32.479>a <00:05:32.655>real <00:05:32.830>simple <00:05:33.005>version <00:05:33.180>of <00:05:33.354>this

134
00:05:33.635 --> 00:05:34.755
<00:05:33.635>that <00:05:33.887>is <00:05:34.139>going <00:05:34.391>to <00:05:34.643>work.

135
00:05:35.335 --> 00:05:36.955
<00:05:35.335>You'll <00:05:35.614>be <00:05:35.895>surprised <00:05:36.174>to <00:05:36.455>find <00:05:36.734>out.

136
00:05:37.295 --> 00:05:40.235
<00:05:37.295>Um, <00:05:37.422>but, <00:05:37.549>uh, <00:05:37.676>it <00:05:37.804>does <00:05:37.931>not <00:05:38.058>take <00:05:38.185>that <00:05:38.313>much <00:05:38.440>code <00:05:38.567>at <00:05:38.695>all.

137
00:05:40.935 --> 00:05:44.595
<00:05:40.935>Um, <00:05:41.110>so <00:05:41.285>this <00:05:41.460>is <00:05:41.635>my <00:05:41.810>hypothesis <00:05:41.985>for <00:05:42.160>this <00:05:42.335>demo,

138
00:05:44.655 --> 00:05:47.735
<00:05:44.655>and <00:05:44.810>for <00:05:44.966>what <00:05:45.121>we're <00:05:45.277>gonna <00:05:45.432>build <00:05:45.588>for <00:05:45.743>a <00:05:45.899>conference <00:05:46.054>bot.

139
00:05:47.735 --> 00:05:49.975
<00:05:47.735>Something <00:05:47.910>to <00:05:48.085>help <00:05:48.260>us <00:05:48.435>understand <00:05:48.610>things <00:05:48.785>about <00:05:48.960>the <00:05:49.135>talks

140
00:05:50.045 --> 00:05:51.575
<00:05:50.045>that <00:05:50.241>we're <00:05:50.438>going <00:05:50.635>to <00:05:50.831>see <00:05:51.028>today <00:05:51.225>and <00:05:51.421>tomorrow.

141
00:05:52.275 --> 00:05:54.935
<00:05:52.275>Uh, <00:05:52.508>the <00:05:52.741>titles <00:05:52.974>and <00:05:53.208>descriptions <00:05:53.441>contain <00:05:53.674>the

142
00:05:54.935 --> 00:05:56.175
<00:05:54.935>meaning <00:05:55.307>of <00:05:55.679>those <00:05:56.051>talks.

143
00:05:56.715 --> 00:05:59.405
<00:05:56.715>It <00:05:56.855>seems <00:05:56.995>reasonable. <00:05:57.135>It <00:05:57.275>would <00:05:57.415>be <00:05:57.555>pretty <00:05:57.695>rough <00:05:57.835>if <00:05:57.975>the <00:05:58.115>title

144
00:05:59.405 --> 00:06:00.525
<00:05:59.405>and <00:05:59.909>description, <00:06:00.412>uh,

145
00:06:00.525 --> 00:06:01.925
<00:06:00.525>was <00:06:00.777>describing <00:06:01.029>a <00:06:01.281>completely <00:06:01.533>different <00:06:01.785>talk.

146
00:06:02.785 --> 00:06:04.445
<00:06:02.785>Um, <00:06:03.018>those <00:06:03.251>are <00:06:03.485>made <00:06:03.718>up <00:06:03.951>of <00:06:04.185>words.

147
00:06:06.185 --> 00:06:07.525
<00:06:06.185>Uh, <00:06:06.426>and <00:06:06.667>the <00:06:06.908>title <00:06:07.149>and <00:06:07.390>descriptions

148
00:06:07.525 --> 00:06:09.445
<00:06:07.525>that <00:06:07.804>share <00:06:08.085>words <00:06:08.364>are <00:06:08.645>probably <00:06:08.924>similar.

149
00:06:10.185 --> 00:06:12.405
<00:06:10.185>Now, <00:06:10.340>this, <00:06:10.496>uh, <00:06:10.651>because <00:06:10.807>we <00:06:10.962>are <00:06:11.118>within <00:06:11.273>a <00:06:11.429>developer <00:06:11.585>conference,

150
00:06:12.785 --> 00:06:14.925
<00:06:12.785>uh, <00:06:12.985>the <00:06:13.185>context <00:06:13.385>comes <00:06:13.585>with <00:06:13.785>it <00:06:13.985>that <00:06:14.185>many

151
00:06:14.925 --> 00:06:17.005
<00:06:14.925>of <00:06:15.125>those <00:06:15.325>words <00:06:15.525>probably <00:06:15.725>mean <00:06:15.925>the <00:06:16.125>same <00:06:16.325>thing

152
00:06:17.005 --> 00:06:18.005
<00:06:17.005>because, <00:06:17.905>uh,

153
00:06:18.105 --> 00:06:20.645
<00:06:18.105>we <00:06:18.305>are <00:06:18.505>talking <00:06:18.705>about <00:06:18.905>technology, <00:06:19.105>uh, <00:06:19.305>and <00:06:19.505>development.

154
00:06:20.985 --> 00:06:22.325
<00:06:20.985>Uh, <00:06:21.186>and <00:06:21.387>so <00:06:21.588>the <00:06:21.789>titles <00:06:21.990>and <00:06:22.191>encryptions

155
00:06:22.325 --> 00:06:24.085
<00:06:22.325>that <00:06:22.604>share <00:06:22.885>words <00:06:23.164>are <00:06:23.445>probably <00:06:23.724>similar.

156
00:06:24.705 --> 00:06:26.565
<00:06:24.705>And <00:06:24.844>therefore, <00:06:24.984>we <00:06:25.125>should <00:06:25.265>be <00:06:25.405>able <00:06:25.544>to <00:06:25.685>collect <00:06:25.825>all <00:06:25.965>the <00:06:26.104>words

157
00:06:26.995 --> 00:06:29.965
<00:06:26.995>that <00:06:27.135>represent <00:06:27.275>each <00:06:27.415>talk <00:06:27.555>as <00:06:27.695>account <00:06:27.835>of <00:06:27.975>each <00:06:28.115>of <00:06:28.255>those <00:06:28.395>words.

158
00:06:31.385 --> 00:06:33.965
<00:06:31.385>Uh, <00:06:31.512>and <00:06:31.639>then <00:06:31.766>when <00:06:31.894>we <00:06:32.021>get <00:06:32.148>a <00:06:32.275>user <00:06:32.403>query, <00:06:32.530>we <00:06:32.657>can, <00:06:32.784>uh,

159
00:06:33.965 --> 00:06:35.045
<00:06:33.965>kind <00:06:34.159>of <00:06:34.353>do <00:06:34.548>the <00:06:34.742>same <00:06:34.937>thing.

160
00:06:35.045 --> 00:06:37.805
<00:06:35.045>Take <00:06:35.185>all <00:06:35.325>the <00:06:35.465>words <00:06:35.605>from <00:06:35.745>the <00:06:35.885>user <00:06:36.025>query, <00:06:36.165>uh, <00:06:36.305>uh, <00:06:36.445>uh,

161
00:06:37.805 --> 00:06:41.205
<00:06:37.805>turn <00:06:37.945>it <00:06:38.085>into <00:06:38.225>a <00:06:38.365>similar <00:06:38.505>vector, <00:06:38.645>uh, <00:06:38.785>to <00:06:38.925>our, <00:06:39.065>uh, <00:06:39.205>our,

162
00:06:41.225 --> 00:06:44.205
<00:06:41.225>our <00:06:41.425>talks, <00:06:41.625>uh, <00:06:41.825>and, <00:06:42.025>and, <00:06:42.225>and, <00:06:42.425>uh, <00:06:42.625>descriptions.

163
00:06:44.545 --> 00:06:47.085
<00:06:44.545>Uh, <00:06:44.685>and <00:06:44.825>then <00:06:44.965>compare <00:06:45.105>them. <00:06:45.245>And <00:06:45.385>we'll <00:06:45.525>get <00:06:45.665>to <00:06:45.805>comparing <00:06:45.945>later.

164
00:06:47.385 --> 00:06:52.365
<00:06:47.385>But <00:06:47.585>let's <00:06:47.784>build <00:06:47.985>this <00:06:48.185>bit <00:06:48.385>now. <00:06:48.585>So <00:06:48.784>I

165
00:06:52.525 --> 00:06:56.485
<00:06:52.525>have <00:06:52.700>got, <00:06:52.875>um, <00:06:53.049>a <00:06:53.224>bunch <00:06:53.400>of <00:06:53.575>code <00:06:53.750>already, <00:06:53.924>uh,

166
00:06:56.485 --> 00:06:59.725
<00:06:56.485>but <00:06:56.601>all <00:06:56.718>this <00:06:56.835>is <00:06:56.951>doing <00:06:57.068>right <00:06:57.185>now <00:06:57.301>is, <00:06:57.418>uh, <00:06:57.535>is <00:06:57.651>cutting <00:06:57.768>out <00:06:57.885>sort

167
00:06:59.725 --> 00:07:02.285
<00:06:59.725>of <00:06:59.865>common <00:07:00.005>words, <00:07:00.145>uh, <00:07:00.285>that <00:07:00.425>are <00:07:00.565>not <00:07:00.705>gonna <00:07:00.845>be <00:07:00.985>very <00:07:01.125>helpful

168
00:07:02.505 --> 00:07:03.645
<00:07:02.505>for <00:07:02.710>the <00:07:02.915>meaning <00:07:03.120>of <00:07:03.325>the <00:07:03.531>text.

169
00:07:04.265 --> 00:07:06.965
<00:07:04.265>Uh, <00:07:04.498>also <00:07:04.731>cutting <00:07:04.965>out <00:07:05.198>punctuation <00:07:05.431>and, <00:07:05.664>uh,

170
00:07:07.145 --> 00:07:09.165
<00:07:07.145>and <00:07:07.495>cutting <00:07:07.844>out <00:07:08.195>like <00:07:08.544>spaces

171
00:07:09.265 --> 00:07:11.605
<00:07:09.265>and <00:07:09.420>other <00:07:09.576>things <00:07:09.731>that <00:07:09.887>are <00:07:10.042>un <00:07:10.198>interesting, <00:07:10.353>unimportant <00:07:10.509>to <00:07:10.664>us.

172
00:07:12.105 --> 00:07:13.885
<00:07:12.105>Uh, <00:07:12.280>so <00:07:12.455>I've <00:07:12.630>got <00:07:12.805>this <00:07:12.980>text <00:07:13.155>of <00:07:13.330>words <00:07:13.505>function

173
00:07:13.885 --> 00:07:18.205
<00:07:13.885>that <00:07:14.001>just <00:07:14.118>takes <00:07:14.235>a, <00:07:14.351>uh, <00:07:14.468>a, <00:07:14.585>a <00:07:14.701>a <00:07:14.818>a <00:07:14.935>normally <00:07:15.051>written, <00:07:15.168>uh, <00:07:15.284>piece

174
00:07:18.205 --> 00:07:20.845
<00:07:18.205>of, <00:07:18.344>um, <00:07:18.484>text, <00:07:18.625>and <00:07:18.765>then, <00:07:18.905>uh, <00:07:19.044>turns <00:07:19.185>it <00:07:19.325>into <00:07:19.465>a <00:07:19.604>bunch

175
00:07:20.845 --> 00:07:24.445
<00:07:20.845>of <00:07:21.125>lowercase, <00:07:21.405>um, <00:07:21.685>meaningful <00:07:21.965>words, <00:07:22.245>hopefully.

176
00:07:25.025 --> 00:07:28.405
<00:07:25.025>Uh, <00:07:25.164>and <00:07:25.304>then <00:07:25.445>I <00:07:25.585>have <00:07:25.724>this, <00:07:25.864>uh, <00:07:26.005>Jason, <00:07:26.145>uh, <00:07:26.284>uh, <00:07:26.424>list

177
00:07:28.865 --> 00:07:31.565
<00:07:28.865>of <00:07:29.005>all <00:07:29.145>the <00:07:29.285>talks <00:07:29.425>and <00:07:29.565>descriptions, <00:07:29.705>uh, <00:07:29.845>that <00:07:29.985>we <00:07:30.125>have <00:07:30.265>today.

178
00:07:32.105 --> 00:07:34.685
<00:07:32.105>So <00:07:32.338>first <00:07:32.571>up, <00:07:32.805>we're <00:07:33.038>gonna <00:07:33.271>go <00:07:33.505>ahead

179
00:07:34.685 --> 00:07:38.485
<00:07:34.685>and <00:07:34.965>create <00:07:35.245>ourselves <00:07:35.525>a, <00:07:35.805>uh, <00:07:36.085>list

180
00:07:38.485 --> 00:07:41.085
<00:07:38.485>of <00:07:38.612>all <00:07:38.739>the <00:07:38.866>words <00:07:38.994>in <00:07:39.121>all <00:07:39.248>of <00:07:39.375>the <00:07:39.503>talk <00:07:39.630>titles <00:07:39.757>and <00:07:39.885>descriptions.

181
00:07:41.625 --> 00:07:45.605
<00:07:41.625>Um, <00:07:41.800>that's <00:07:41.975>almost <00:07:42.150>it. <00:07:42.325>Um, <00:07:42.500>copilot <00:07:42.675>is <00:07:42.850>really <00:07:43.025>helpful.

182
00:07:45.785 --> 00:07:47.885
<00:07:45.784>Uh, <00:07:45.924>when, <00:07:46.064>uh, <00:07:46.205>when, <00:07:46.344>when <00:07:46.484>you <00:07:46.624>have <00:07:46.765>to <00:07:46.905>do <00:07:47.044>life <00:07:47.184>coding.

183
00:07:48.085 --> 00:07:49.845
<00:07:48.085>'cause <00:07:48.284>it <00:07:48.484>already <00:07:48.685>knows, <00:07:48.885>it's <00:07:49.085>actually <00:07:49.284>seen <00:07:49.484>it

184
00:07:49.845 --> 00:07:52.005
<00:07:49.845>before <00:07:50.045>a <00:07:50.245>couple <00:07:50.445>of <00:07:50.645>times, <00:07:50.845>although <00:07:51.045>not <00:07:51.245>quite.

185
00:07:52.035 --> 00:07:55.005
<00:07:52.034>This <00:07:52.190>is <00:07:52.346>incorrect. <00:07:52.501>Um, <00:07:52.657>so <00:07:52.812>we're <00:07:52.968>gonna <00:07:53.123>turn <00:07:53.279>the <00:07:53.434>text,

186
00:07:55.185 --> 00:07:57.285
<00:07:55.185>the <00:07:55.465>talk <00:07:55.745>into, <00:07:56.025>uh, <00:07:56.305>just <00:07:56.585>words.

187
00:07:57.375 --> 00:07:58.445
<00:07:57.375>We're <00:07:57.615>gonna <00:07:57.856>flat <00:07:58.097>map <00:07:58.337>that,

188
00:07:58.445 --> 00:08:01.925
<00:07:58.445>and <00:07:58.585>we <00:07:58.724>actually <00:07:58.865>want <00:07:59.005>that <00:07:59.145>to <00:07:59.284>be, <00:07:59.425>uh, <00:07:59.565>a <00:07:59.705>unique <00:07:59.844>array.

189
00:08:02.405 --> 00:08:03.885
<00:08:02.405>'cause <00:08:02.671>otherwise <00:08:02.937>that's <00:08:03.204>not <00:08:03.470>very <00:08:03.736>helpful.

190
00:08:04.385 --> 00:08:05.605
<00:08:04.385>Uh, <00:08:04.604>so <00:08:04.824>we'll <00:08:05.043>make <00:08:05.263>it, <00:08:05.483>uh,

191
00:08:05.605 --> 00:08:08.045
<00:08:05.605>an <00:08:05.780>array <00:08:05.955>from <00:08:06.130>a <00:08:06.305>set <00:08:06.480>of <00:08:06.655>all <00:08:06.830>the <00:08:07.005>words.

192
00:08:09.145 --> 00:08:10.765
<00:08:09.145>And <00:08:09.424>when <00:08:09.705>we <00:08:09.984>do <00:08:10.265>that, <00:08:10.544>um,

193
00:08:11.545 --> 00:08:16.545
<00:08:11.545>I'm <00:08:11.825>just <00:08:12.105>gonna <00:08:12.385>log <00:08:12.665>out, <00:08:12.945>uh,

194
00:08:17.465 --> 00:08:18.805
<00:08:17.465>If <00:08:17.706>I <00:08:17.947>re-save <00:08:18.188>the <00:08:18.429>thing <00:08:18.671>right

195
00:08:18.805 --> 00:08:19.845
<00:08:18.805>down <00:08:18.961>at <00:08:19.117>the <00:08:19.273>bottom <00:08:19.429>of <00:08:19.585>the <00:08:19.741>screen.

196
00:08:19.845 --> 00:08:21.085
<00:08:19.845>It's <00:08:19.984>a <00:08:20.124>little <00:08:20.263>bit <00:08:20.403>small <00:08:20.542>there. <00:08:20.682>But <00:08:20.821>that <00:08:20.961>says,

197
00:08:21.085 --> 00:08:25.765
<00:08:21.085>we've <00:08:21.260>got <00:08:21.435>5,965 <00:08:21.609>unique <00:08:21.784>words, <00:08:21.960>uh, <00:08:22.135>in <00:08:22.310>this <00:08:22.484>set,

198
00:08:25.785 --> 00:08:28.525
<00:08:25.785>in <00:08:25.925>this <00:08:26.065>corpus <00:08:26.205>of <00:08:26.345>data, <00:08:26.485>in <00:08:26.625>this, <00:08:26.765>uh, <00:08:26.905>set <00:08:27.045>of <00:08:27.185>talks

199
00:08:28.525 --> 00:08:29.645
<00:08:28.525>and <00:08:28.777>descriptions <00:08:29.029>that <00:08:29.281>we <00:08:29.532>have.

200
00:08:30.385 --> 00:08:32.525
<00:08:30.385>Um, <00:08:31.784>uh,

201
00:08:32.545 --> 00:08:35.285
<00:08:32.544>and <00:08:32.672>so <00:08:32.799>actually, <00:08:32.926>we <00:08:33.054>can <00:08:33.181>just <00:08:33.308>log <00:08:33.435>out <00:08:33.563>the, <00:08:33.690>the <00:08:33.817>whole <00:08:33.944>lot.

202
00:08:35.785 --> 00:08:37.525
<00:08:35.784>Um, <00:08:35.940>this <00:08:36.096>is <00:08:36.251>gonna <00:08:36.407>show <00:08:36.562>me <00:08:36.718>the <00:08:36.873>first <00:08:37.029>a <00:08:37.184>hundred.

203
00:08:37.945 --> 00:08:39.605
<00:08:37.945>And <00:08:38.085>what's <00:08:38.225>quite <00:08:38.365>good <00:08:38.505>about <00:08:38.645>this <00:08:38.785>is <00:08:38.925>we <00:08:39.065>see <00:08:39.205>that <00:08:39.345>these,

204
00:08:39.605 --> 00:08:41.525
<00:08:39.605>these <00:08:39.838>first <00:08:40.071>ones <00:08:40.305>are <00:08:40.538>better <00:08:40.771>safe, <00:08:41.005>sorry,

205
00:08:41.525 --> 00:08:43.605
<00:08:41.525>threat <00:08:41.991>modeling <00:08:42.458>web <00:08:42.924>developers.

206
00:08:44.145 --> 00:08:47.005
<00:08:44.145>Uh, <00:08:44.300>and <00:08:44.456>if <00:08:44.611>we <00:08:44.767>actually <00:08:44.922>look <00:08:45.078>at, <00:08:45.233>uh, <00:08:45.389>the <00:08:45.545>torques

207
00:08:47.065 --> 00:08:49.565
<00:08:47.065>and <00:08:47.240>the <00:08:47.415>first <00:08:47.590>torque, <00:08:47.765>uh, <00:08:47.940>we <00:08:48.115>can <00:08:48.290>see <00:08:48.465>that

208
00:08:49.565 --> 00:08:51.605
<00:08:49.565>that <00:08:49.740>one <00:08:49.915>is <00:08:50.090>the <00:08:50.265>torque <00:08:50.440>better <00:08:50.615>safe <00:08:50.790>than, <00:08:50.965>sorry,

209
00:08:51.665 --> 00:08:54.085
<00:08:51.665>threat <00:08:51.865>modeling <00:08:52.064>for <00:08:52.265>web <00:08:52.464>developers <00:08:52.665>by <00:08:52.865>Clemen <00:08:53.064>Soner.

210
00:08:54.745 --> 00:08:57.845
<00:08:54.745>Um, <00:08:54.900>also, <00:08:55.056>if <00:08:55.211>we <00:08:55.367>look <00:08:55.522>at <00:08:55.678>number <00:08:55.833>329, <00:08:55.989>that's <00:08:56.145>mine.

211
00:08:58.225 --> 00:09:00.205
<00:08:58.225>Uh, <00:08:58.365>so <00:08:58.505>I'm <00:08:58.645>definitely <00:08:58.785>in <00:08:58.925>there, <00:08:59.065>which <00:08:59.205>is <00:08:59.345>good <00:08:59.485>to <00:08:59.625>know.

212
00:09:01.105 --> 00:09:04.165
<00:09:01.105>Um, <00:09:01.232>so <00:09:01.359>we <00:09:01.486>have <00:09:01.614>all <00:09:01.741>the <00:09:01.868>words <00:09:01.995>that <00:09:02.123>are <00:09:02.250>in <00:09:02.377>them, <00:09:02.505>three,

213
00:09:04.705 --> 00:09:05.885
<00:09:04.705>almost <00:09:05.236>6,000 <00:09:05.767>words.

214
00:09:06.825 --> 00:09:09.525
<00:09:06.825>So <00:09:06.965>let's <00:09:07.105>turn, <00:09:07.245>let's <00:09:07.385>take <00:09:07.525>one <00:09:07.665>of <00:09:07.805>those <00:09:07.945>text, <00:09:08.085>uh, <00:09:08.225>blocks

215
00:09:09.525 --> 00:09:11.765
<00:09:09.525>that <00:09:09.641>we <00:09:09.758>have <00:09:09.875>and <00:09:09.991>turn <00:09:10.108>it <00:09:10.225>into <00:09:10.341>a <00:09:10.458>vector <00:09:10.574>based <00:09:10.691>on <00:09:10.808>those <00:09:10.925>words.

216
00:09:13.465 --> 00:09:17.885
<00:09:13.465>Um, <00:09:13.592>we're <00:09:13.719>going <00:09:13.846>to <00:09:13.974>do <00:09:14.101>that <00:09:14.228>by, <00:09:14.355>uh, <00:09:14.483>creating <00:09:14.610>ourselves <00:09:14.737>a, <00:09:14.865>uh,

217
00:09:17.965 --> 00:09:21.215
<00:09:17.965>a <00:09:18.431>word <00:09:18.898>map, <00:09:19.365>uh,

218
00:09:22.065 --> 00:09:23.695
<00:09:22.065>using <00:09:22.220>an <00:09:22.376>object <00:09:22.531>just <00:09:22.687>to <00:09:22.842>make <00:09:22.998>it <00:09:23.153>nice <00:09:23.309>and <00:09:23.465>easy.

219
00:09:23.955 --> 00:09:25.455
<00:09:23.955>And <00:09:24.180>we're <00:09:24.405>gonna <00:09:24.630>take <00:09:24.855>the, <00:09:25.080>uh, <00:09:25.305>text.

220
00:09:25.505 --> 00:09:26.535
<00:09:25.505>We're <00:09:25.690>gonna <00:09:25.875>turn <00:09:26.061>it <00:09:26.246>to <00:09:26.432>words.

221
00:09:26.535 --> 00:09:27.855
<00:09:26.535>Yeah, <00:09:26.704>it's <00:09:26.874>doing <00:09:27.044>that <00:09:27.213>for <00:09:27.383>me, <00:09:27.553>isn't <00:09:27.723>it?

222
00:09:28.045 --> 00:09:29.775
<00:09:28.045>That'll, <00:09:28.324>that'll <00:09:28.604>work. <00:09:28.885>So <00:09:29.165>we <00:09:29.444>take

223
00:09:29.775 --> 00:09:30.895
<00:09:29.775>the <00:09:29.976>text, <00:09:30.178>turn <00:09:30.379>it <00:09:30.581>into <00:09:30.783>words.

224
00:09:31.435 --> 00:09:34.375
<00:09:31.435>Um, <00:09:31.574>and <00:09:31.714>for <00:09:31.854>each <00:09:31.994>of <00:09:32.135>those <00:09:32.275>words, <00:09:32.415>we're <00:09:32.555>going <00:09:32.694>to, <00:09:32.834>uh,

225
00:09:34.375 --> 00:09:36.575
<00:09:34.375>either <00:09:34.550>log <00:09:34.725>the <00:09:34.900>word <00:09:35.075>in <00:09:35.250>the <00:09:35.425>word <00:09:35.600>map <00:09:35.775>as

226
00:09:36.595 --> 00:09:38.135
<00:09:36.595>as <00:09:36.793>one, <00:09:36.991>or <00:09:37.189>we're <00:09:37.387>gonna <00:09:37.585>add <00:09:37.783>to <00:09:37.981>it.

227
00:09:38.135 --> 00:09:39.215
<00:09:38.135>So <00:09:38.297>we're <00:09:38.459>just <00:09:38.621>counting <00:09:38.783>the <00:09:38.945>number <00:09:39.107>of

228
00:09:39.215 --> 00:09:40.855
<00:09:39.215>words, <00:09:39.495>uh, <00:09:39.775>here <00:09:40.055>in <00:09:40.335>this <00:09:40.615>thing.

229
00:09:40.855 --> 00:09:43.135
<00:09:40.855>And <00:09:41.135>I'm <00:09:41.415>gonna <00:09:41.695>return, <00:09:41.975>then, <00:09:42.255>uh,

230
00:09:43.135 --> 00:09:47.055
<00:09:43.135>we're <00:09:43.310>gonna <00:09:43.485>return <00:09:43.660>the <00:09:43.835>all <00:09:44.010>words, <00:09:44.185>um, <00:09:44.360>the <00:09:44.535>6,000,

231
00:09:47.055 --> 00:09:50.255
<00:09:47.055>almost <00:09:47.210>6,000 <00:09:47.366>long <00:09:47.521>array, <00:09:47.677>uh, <00:09:47.832>where <00:09:47.988>we <00:09:48.143>map, <00:09:48.299>uh, <00:09:48.454>each

232
00:09:50.255 --> 00:09:53.095
<00:09:50.255>of <00:09:50.395>the <00:09:50.535>spots <00:09:50.675>in <00:09:50.814>it <00:09:50.955>to, <00:09:51.095>uh, <00:09:51.235>the <00:09:51.375>number <00:09:51.515>of <00:09:51.655>words

233
00:09:53.095 --> 00:09:56.255
<00:09:53.095>that <00:09:53.235>we <00:09:53.375>had <00:09:53.515>in <00:09:53.655>our <00:09:53.795>talk, <00:09:53.935>uh, <00:09:54.075>in <00:09:54.215>our <00:09:54.355>talk <00:09:54.495>description.

234
00:09:56.835 --> 00:09:59.895
<00:09:56.835>And <00:09:56.935>so <00:09:57.035>if <00:09:57.135>we <00:09:57.235>take, <00:09:57.335>uh, <00:09:57.435>if <00:09:57.535>we <00:09:57.635>now <00:09:57.735>have <00:09:57.835>a <00:09:57.935>look <00:09:58.035>at <00:09:58.135>console <00:09:58.235>log,

235
00:10:00.435 --> 00:10:04.695
<00:10:00.435>uh, <00:10:00.609>text <00:10:00.785>to <00:10:00.959>vector <00:10:01.135>of <00:10:01.310>that <00:10:01.484>first <00:10:01.660>talk, <00:10:01.834>uh,

236
00:10:04.695 --> 00:10:06.095
<00:10:04.695>we <00:10:04.947>get <00:10:05.199>this <00:10:05.451>big <00:10:05.703>long <00:10:05.955>array,

237
00:10:06.095 --> 00:10:08.535
<00:10:06.095>which <00:10:06.328>is <00:10:06.561>actually <00:10:06.795>almost <00:10:07.028>6,000 <00:10:07.261>elements <00:10:07.495>long.

238
00:10:08.955 --> 00:10:10.575
<00:10:08.955>Uh, <00:10:09.130>but <00:10:09.305>we <00:10:09.480>can <00:10:09.655>kind <00:10:09.830>of <00:10:10.005>see <00:10:10.180>that <00:10:10.355>actually,

239
00:10:10.575 --> 00:10:12.015
<00:10:10.575>because <00:10:10.791>this <00:10:11.007>talk <00:10:11.223>is <00:10:11.439>the <00:10:11.655>first <00:10:11.871>one.

240
00:10:12.395 --> 00:10:14.295
<00:10:12.395>Uh, <00:10:12.595>all <00:10:12.795>of <00:10:12.995>its <00:10:13.194>words <00:10:13.395>are <00:10:13.595>represented <00:10:13.795>at

241
00:10:14.295 --> 00:10:15.735
<00:10:14.295>the <00:10:14.554>absolute <00:10:14.813>start <00:10:15.072>of <00:10:15.331>the <00:10:15.591>array.

242
00:10:16.235 --> 00:10:20.255
<00:10:16.235>Uh, <00:10:16.375>and, <00:10:16.515>and <00:10:16.655>in <00:10:16.795>fact, <00:10:16.935>that's <00:10:17.075>not <00:10:17.215>only <00:10:17.355>500,865 <00:10:17.495>more <00:10:17.635>times,

243
00:10:20.525 --> 00:10:22.255
<00:10:20.525>more <00:10:20.758>items, <00:10:20.991>they're <00:10:21.225>all <00:10:21.458>zeros <00:10:21.691>as <00:10:21.925>well,

244
00:10:22.255 --> 00:10:23.695
<00:10:22.255>because <00:10:22.471>we <00:10:22.687>used <00:10:22.903>all <00:10:23.119>the <00:10:23.335>words <00:10:23.551>immediately.

245
00:10:24.075 --> 00:10:26.095
<00:10:24.075>And <00:10:24.215>so <00:10:24.355>this <00:10:24.495>is <00:10:24.635>kind <00:10:24.775>of <00:10:24.915>known <00:10:25.055>as <00:10:25.195>a <00:10:25.335>sparse <00:10:25.475>vector.

246
00:10:26.485 --> 00:10:28.375
<00:10:26.485>Like, <00:10:26.660>it, <00:10:26.835>it <00:10:27.010>doesn't <00:10:27.185>show <00:10:27.360>us <00:10:27.535>an <00:10:27.710>awful <00:10:27.885>lot.

247
00:10:28.875 --> 00:10:31.855
<00:10:28.875>Um, <00:10:29.002>but <00:10:29.129>let's <00:10:29.256>have <00:10:29.384>a <00:10:29.511>look <00:10:29.638>at <00:10:29.765>3 <00:10:29.893>29. <00:10:30.020>That's <00:10:30.147>my <00:10:30.275>talk.

248
00:10:32.235 --> 00:10:34.015
<00:10:32.235>If <00:10:32.410>we <00:10:32.585>look <00:10:32.760>at <00:10:32.935>that, <00:10:33.110>we <00:10:33.285>see <00:10:33.460>that, <00:10:33.635>uh,

249
00:10:34.685 --> 00:10:35.895
<00:10:34.685>that <00:10:34.957>there <00:10:35.229>are <00:10:35.501>some <00:10:35.774>ones,

250
00:10:35.895 --> 00:10:37.535
<00:10:35.895>there's <00:10:36.095>still <00:10:36.295>a <00:10:36.495>lot <00:10:36.694>of <00:10:36.895>zeros <00:10:37.095>in <00:10:37.295>there.

251
00:10:37.915 --> 00:10:40.895
<00:10:37.915>Uh, <00:10:38.089>but, <00:10:38.265>uh, <00:10:38.439>that's <00:10:38.615>probably <00:10:38.790>the <00:10:38.964>word <00:10:39.140>stage, <00:10:39.314>uh,

252
00:10:40.895 --> 00:10:44.055
<00:10:40.895>that <00:10:41.022>first <00:10:41.149>one <00:10:41.276>because <00:10:41.404>all <00:10:41.531>the <00:10:41.658>talks <00:10:41.785>say, <00:10:41.913>um, <00:10:42.040>it's <00:10:42.167>the <00:10:42.295>title

253
00:10:44.155 --> 00:10:45.415
<00:10:44.155>by <00:10:44.381>this <00:10:44.608>person <00:10:44.835>on <00:10:45.062>this <00:10:45.289>stage.

254
00:10:46.235 --> 00:10:47.455
<00:10:46.235>Uh, <00:10:46.391>and <00:10:46.548>then <00:10:46.705>we <00:10:46.862>get <00:10:47.019>some <00:10:47.176>other <00:10:47.333>words.

255
00:10:48.515 --> 00:10:51.415
<00:10:48.515>So <00:10:48.689>we <00:10:48.865>have <00:10:49.040>vectorized <00:10:49.215>at <00:10:49.390>this <00:10:49.564>point, <00:10:49.740>I <00:10:49.915>think.

256
00:10:51.415 --> 00:10:56.015
<00:10:51.415>And <00:10:51.531>so <00:10:51.648>we <00:10:51.765>can <00:10:51.881>now, <00:10:51.998>um, <00:10:52.115>create, <00:10:52.231>uh, <00:10:52.348>an <00:10:52.464>array, <00:10:52.581>which <00:10:52.698>has <00:10:52.814>a,

257
00:10:56.135 --> 00:10:57.575
<00:10:56.135>a <00:10:56.394>combination <00:10:56.653>of <00:10:56.912>both <00:10:57.171>the <00:10:57.431>torque

258
00:10:57.575 --> 00:10:59.695
<00:10:57.575>and <00:10:57.750>the <00:10:57.925>vector <00:10:58.100>alongside <00:10:58.275>it <00:10:58.450>that's <00:10:58.625>gonna <00:10:58.800>be <00:10:58.975>useful.

259
00:11:00.195 --> 00:11:02.695
<00:11:00.195>Uh, <00:11:00.395>we <00:11:00.595>actually <00:11:00.795>just <00:11:00.995>take <00:11:01.195>all <00:11:01.395>the <00:11:01.595>talks

260
00:11:02.695 --> 00:11:07.295
<00:11:02.695>and <00:11:02.835>we <00:11:02.975>map <00:11:03.115>it <00:11:03.255>to, <00:11:03.395>uh, <00:11:03.535>torque <00:11:03.675>and <00:11:03.815>a <00:11:03.955>rule <00:11:04.095>return

261
00:11:07.295 --> 00:11:10.255
<00:11:07.295>and <00:11:07.450>array <00:11:07.606>like <00:11:07.761>a <00:11:07.917>topple, <00:11:08.072>basically, <00:11:08.228>um, <00:11:08.383>of <00:11:08.539>the <00:11:08.694>torque

262
00:11:10.255 --> 00:11:11.735
<00:11:10.255>and <00:11:10.477>text <00:11:10.699>to <00:11:10.921>vector <00:11:11.143>of <00:11:11.365>the <00:11:11.587>torque.

263
00:11:12.075 --> 00:11:14.935
<00:11:12.075>So <00:11:12.191>when <00:11:12.308>we <00:11:12.425>do <00:11:12.541>that, <00:11:12.658>we <00:11:12.775>can <00:11:12.891>now <00:11:13.008>have <00:11:13.125>a <00:11:13.241>look <00:11:13.358>at, <00:11:13.475>uh,

264
00:11:14.935 --> 00:11:18.615
<00:11:14.935>talk <00:11:15.062>vectors <00:11:15.189>number <00:11:15.316>3 <00:11:15.444>29, <00:11:15.571>uh, <00:11:15.698>which <00:11:15.825>is <00:11:15.953>my <00:11:16.080>talk <00:11:16.207>about <00:11:16.334>rag,

265
00:11:18.955 --> 00:11:19.975
<00:11:18.955>and <00:11:19.184>then <00:11:19.414>also <00:11:19.643>its <00:11:19.873>vector.

266
00:11:20.435 --> 00:11:23.415
<00:11:20.435>So <00:11:20.590>we <00:11:20.746>have <00:11:20.901>that <00:11:21.057>kind <00:11:21.212>of <00:11:21.368>representation <00:11:21.523>of, <00:11:21.679>uh, <00:11:21.834>all

267
00:11:23.535 --> 00:11:27.535
<00:11:23.535>of <00:11:23.675>our <00:11:23.814>talks <00:11:23.954>and <00:11:24.094>all <00:11:24.235>of <00:11:24.375>the <00:11:24.515>data <00:11:24.655>that <00:11:24.795>they <00:11:24.935>contain.

268
00:11:30.485 --> 00:11:33.065
<00:11:30.485>So <00:11:30.685>that's <00:11:30.885>a, <00:11:31.085>that's <00:11:31.285>a <00:11:31.485>vector <00:11:31.685>of <00:11:31.885>sorts.

269
00:11:33.375 --> 00:11:35.145
<00:11:33.375>It's <00:11:33.550>sort <00:11:33.725>of <00:11:33.900>known <00:11:34.075>as <00:11:34.250>a <00:11:34.425>bag <00:11:34.600>of <00:11:34.775>words,

270
00:11:35.145 --> 00:11:36.985
<00:11:35.145>because <00:11:35.300>what <00:11:35.456>we <00:11:35.611>have <00:11:35.767>done <00:11:35.922>is <00:11:36.078>collected <00:11:36.233>all <00:11:36.389>the <00:11:36.545>words

271
00:11:36.985 --> 00:11:38.265
<00:11:36.985>and <00:11:37.113>said, <00:11:37.241>okay, <00:11:37.369>how <00:11:37.497>many <00:11:37.625>have <00:11:37.753>we <00:11:37.881>got <00:11:38.009>of <00:11:38.137>these?

272
00:11:39.925 --> 00:11:41.105
<00:11:39.925>But <00:11:40.102>now <00:11:40.279>we <00:11:40.456>need <00:11:40.633>to <00:11:40.810>compare <00:11:40.987>them.

273
00:11:42.125 --> 00:11:45.745
<00:11:42.125>Uh, <00:11:42.358>comparing <00:11:42.591>vectors, <00:11:42.825>uh, <00:11:43.058>is <00:11:43.291>surprisingly <00:11:43.525>good

274
00:11:45.745 --> 00:11:47.985
<00:11:45.745>because <00:11:45.920>as, <00:11:46.095>as <00:11:46.270>I <00:11:46.445>said <00:11:46.620>earlier, <00:11:46.795>computers <00:11:46.970>love <00:11:47.145>numbers.

275
00:11:48.005 --> 00:11:50.025
<00:11:48.005>And <00:11:48.132>now <00:11:48.259>we <00:11:48.386>just <00:11:48.514>have <00:11:48.641>a <00:11:48.768>bunch <00:11:48.895>of <00:11:49.023>numbers <00:11:49.150>to <00:11:49.277>compare <00:11:49.405>things.

276
00:11:50.925 --> 00:11:53.245
<00:11:50.925>But <00:11:51.064>to <00:11:51.204>understand <00:11:51.344>how <00:11:51.484>we <00:11:51.625>want <00:11:51.765>to <00:11:51.905>compare <00:11:52.045>this, <00:11:52.185>we <00:11:52.324>have

277
00:11:53.245 --> 00:11:55.005
<00:11:53.245>to <00:11:53.385>go <00:11:53.525>back <00:11:53.665>to <00:11:53.805>the <00:11:53.945>idea <00:11:54.085>that <00:11:54.225>this <00:11:54.365>vector <00:11:54.505>is <00:11:54.645>meaning.

278
00:11:55.385 --> 00:11:56.485
<00:11:55.385>Uh, <00:11:55.583>it's, <00:11:55.781>it's <00:11:55.979>kind <00:11:56.177>of <00:11:56.375>trying

279
00:11:56.485 --> 00:11:57.805
<00:11:56.485>to <00:11:56.683>capture <00:11:56.881>the <00:11:57.079>meaning <00:11:57.276>of <00:11:57.474>the <00:11:57.673>content.

280
00:11:59.385 --> 00:12:03.485
<00:11:59.385>Uh, <00:11:59.525>and <00:11:59.665>so <00:11:59.805>it's <00:11:59.944>doing <00:12:00.085>so <00:12:00.225>in <00:12:00.365>a <00:12:00.505>multi-dimensional <00:12:00.645>space, <00:12:00.785>uh,

281
00:12:03.545 --> 00:12:05.645
<00:12:03.545>in <00:12:03.685>the <00:12:03.824>case <00:12:03.964>of <00:12:04.104>our <00:12:04.245>vector <00:12:04.385>that <00:12:04.525>we <00:12:04.665>just <00:12:04.805>created <00:12:04.944>there,

282
00:12:05.805 --> 00:12:10.605
<00:12:05.805>a <00:12:05.979>5,965 <00:12:06.155>dimension <00:12:06.329>space, <00:12:06.505>which <00:12:06.680>is <00:12:06.854>way <00:12:07.030>too, <00:12:07.204>uh,

283
00:12:10.785 --> 00:12:12.485
<00:12:10.785>way <00:12:11.064>too <00:12:11.344>much <00:12:11.625>to <00:12:11.905>possibly <00:12:12.185>imagine.

284
00:12:12.945 --> 00:12:15.685
<00:12:12.945>Uh, <00:12:13.072>so <00:12:13.199>I <00:12:13.326>have <00:12:13.454>decreased <00:12:13.581>that <00:12:13.708>down <00:12:13.835>to <00:12:13.963>a <00:12:14.090>two <00:12:14.217>dimensional <00:12:14.345>space

285
00:12:15.715 --> 00:12:16.885
<00:12:15.715>because <00:12:16.066>that's <00:12:16.417>much <00:12:16.768>easier.

286
00:12:17.675 --> 00:12:19.365
<00:12:17.675>I've <00:12:18.025>also <00:12:18.375>reduced <00:12:18.724>things <00:12:19.074>down

287
00:12:19.365 --> 00:12:21.725
<00:12:19.365>to <00:12:19.540>having <00:12:19.715>three <00:12:19.890>vectors <00:12:20.065>in <00:12:20.240>my <00:12:20.415>two <00:12:20.590>dimensional <00:12:20.765>space,

288
00:12:22.025 --> 00:12:23.165
<00:12:22.025>uh, <00:12:22.281>for <00:12:22.538>the <00:12:22.794>example <00:12:23.050>here.

289
00:12:23.785 --> 00:12:25.605
<00:12:23.785>And <00:12:23.912>so <00:12:24.039>we <00:12:24.166>see, <00:12:24.294>we've <00:12:24.421>got <00:12:24.548>this <00:12:24.675>one <00:12:24.803>vector <00:12:24.930>down <00:12:25.057>the <00:12:25.185>bottom

290
00:12:26.015 --> 00:12:28.765
<00:12:26.015>who's, <00:12:26.189>uh, <00:12:26.365>you <00:12:26.540>know, <00:12:26.715>not <00:12:26.890>very <00:12:27.064>strongly <00:12:27.240>about <00:12:27.415>anything.

291
00:12:28.765 --> 00:12:30.845
<00:12:28.765>It <00:12:29.045>seems <00:12:29.324>this <00:12:29.605>one <00:12:29.885>out <00:12:30.165>here

292
00:12:30.845 --> 00:12:32.325
<00:12:30.845>that's <00:12:30.993>doing <00:12:31.141>a <00:12:31.289>bunch <00:12:31.437>of <00:12:31.585>stuff, <00:12:31.733>this <00:12:31.881>one <00:12:32.029>out <00:12:32.177>here.

293
00:12:32.665 --> 00:12:34.805
<00:12:32.665>And <00:12:32.805>you <00:12:32.944>might <00:12:33.084>think, <00:12:33.224>alright, <00:12:33.365>we, <00:12:33.505>we <00:12:33.645>want <00:12:33.785>the, <00:12:33.925>you <00:12:34.064>know,

294
00:12:34.805 --> 00:12:37.805
<00:12:34.805>the <00:12:34.960>closest <00:12:35.116>meaning, <00:12:35.271>the <00:12:35.427>closest <00:12:35.582>things <00:12:35.738>like <00:12:35.893>to <00:12:36.049>each <00:12:36.204>other.

295
00:12:38.545 --> 00:12:41.445
<00:12:38.545>Uh, <00:12:38.672>and <00:12:38.799>you <00:12:38.926>might <00:12:39.054>think <00:12:39.181>that <00:12:39.308>that <00:12:39.435>means <00:12:39.563>the, <00:12:39.690>the <00:12:39.817>two <00:12:39.944>vectors

296
00:12:41.445 --> 00:12:42.885
<00:12:41.445>that <00:12:41.661>are <00:12:41.877>kind <00:12:42.093>of <00:12:42.309>out <00:12:42.525>further <00:12:42.741>away,

297
00:12:42.885 --> 00:12:45.965
<00:12:42.885>but <00:12:43.025>clearly <00:12:43.165>the <00:12:43.305>two <00:12:43.444>closest <00:12:43.585>together <00:12:43.725>on <00:12:43.865>the, <00:12:44.005>on <00:12:44.145>the <00:12:44.285>map.

298
00:12:47.225 --> 00:12:50.125
<00:12:47.225>But <00:12:47.380>that's <00:12:47.536>not <00:12:47.691>the <00:12:47.847>case. <00:12:48.002>'cause <00:12:48.158>vectors <00:12:48.313>aren't <00:12:48.469>just <00:12:48.625>points

299
00:12:50.185 --> 00:12:53.365
<00:12:50.185>in <00:12:50.340>space, <00:12:50.496>and <00:12:50.651>they're <00:12:50.807>not <00:12:50.962>points <00:12:51.118>in <00:12:51.273>this <00:12:51.429>dimensional <00:12:51.584>space.

300
00:12:53.675 --> 00:12:56.005
<00:12:53.675>They're <00:12:54.025>indeed <00:12:54.375>direction <00:12:54.724>and <00:12:55.074>magnitude.

301
00:12:56.955 --> 00:13:01.525
<00:12:56.955>It's <00:12:57.082>better <00:12:57.209>to <00:12:57.336>show <00:12:57.464>them <00:12:57.591>as <00:12:57.718>like <00:12:57.845>a, <00:12:57.973>a, <00:12:58.100>a, <00:12:58.227>a <00:12:58.355>line,

302
00:13:01.525 --> 00:13:04.245
<00:13:01.525>basically <00:13:01.680>going <00:13:01.836>out <00:13:01.991>from <00:13:02.147>zero <00:13:02.302>zero <00:13:02.458>from <00:13:02.613>the <00:13:02.769>origin, <00:13:02.925>uh,

303
00:13:04.425 --> 00:13:07.565
<00:13:04.425>to <00:13:04.564>the <00:13:04.704>point <00:13:04.844>in <00:13:04.984>space <00:13:05.125>that <00:13:05.265>they <00:13:05.405>eventually <00:13:05.545>kind <00:13:05.685>of <00:13:05.824>reach.

304
00:13:09.265 --> 00:13:13.365
<00:13:09.265>And <00:13:09.420>that <00:13:09.576>kind <00:13:09.731>of, <00:13:09.887>uh, <00:13:10.042>direction <00:13:10.198>is <00:13:10.353>really <00:13:10.509>the <00:13:10.665>meaning

305
00:13:13.515 --> 00:13:15.605
<00:13:13.515>that <00:13:13.715>we <00:13:13.915>are <00:13:14.115>looking <00:13:14.314>for <00:13:14.515>in <00:13:14.715>these <00:13:14.915>vectors.

306
00:13:16.265 --> 00:13:18.925
<00:13:16.265>Um, <00:13:16.439>it <00:13:16.615>turns <00:13:16.790>out <00:13:16.965>the <00:13:17.140>magnitude <00:13:17.314>just <00:13:17.490>tells <00:13:17.665>you

307
00:13:18.925 --> 00:13:20.805
<00:13:18.925>how <00:13:19.204>much <00:13:19.484>it <00:13:19.765>means <00:13:20.045>that, <00:13:20.324>uh,

308
00:13:20.825 --> 00:13:24.285
<00:13:20.825>and <00:13:20.932>so, <00:13:21.040>um, <00:13:21.148>in <00:13:21.255>our, <00:13:21.363>uh, <00:13:21.471>in <00:13:21.578>our <00:13:21.686>example <00:13:21.794>of <00:13:21.901>a <00:13:22.009>bunch <00:13:22.117>of <00:13:22.225>talks,

309
00:13:24.355 --> 00:13:26.685
<00:13:24.355>like <00:13:24.588>the <00:13:24.821>first <00:13:25.055>talk, <00:13:25.288>uh, <00:13:25.521>the, <00:13:25.755>the,

310
00:13:26.785 --> 00:13:28.565
<00:13:26.785>the <00:13:26.959>first <00:13:27.135>talk <00:13:27.310>in <00:13:27.485>my <00:13:27.660>list <00:13:27.834>about <00:13:28.010>threat <00:13:28.185>modeling,

311
00:13:28.785 --> 00:13:30.485
<00:13:28.785>and <00:13:28.959>she <00:13:29.135>says <00:13:29.310>threat <00:13:29.485>modeling <00:13:29.660>in <00:13:29.834>it <00:13:30.010>three <00:13:30.185>times.

312
00:13:31.065 --> 00:13:33.685
<00:13:31.065>Uh, <00:13:31.205>and <00:13:31.345>so <00:13:31.485>it's <00:13:31.625>kind <00:13:31.765>of <00:13:31.905>like <00:13:32.045>strongly <00:13:32.185>about <00:13:32.325>threat <00:13:32.465>modeling,

313
00:13:34.025 --> 00:13:35.885
<00:13:34.025>but <00:13:34.199>if <00:13:34.375>another <00:13:34.550>talk <00:13:34.725>just <00:13:34.900>said <00:13:35.074>threat <00:13:35.250>modeling, <00:13:35.425>once,

314
00:13:36.315 --> 00:13:38.245
<00:13:36.315>it's <00:13:36.515>still <00:13:36.715>about <00:13:36.915>threat <00:13:37.115>modeling, <00:13:37.315>just <00:13:37.515>less <00:13:37.715>so.

315
00:13:39.105 --> 00:13:43.645
<00:13:39.105>And <00:13:39.260>so <00:13:39.416>the <00:13:39.571>closest <00:13:39.727>vectors <00:13:39.882>to <00:13:40.038>each <00:13:40.193>other <00:13:40.349>in <00:13:40.505>terms

316
00:13:43.645 --> 00:13:46.325
<00:13:43.645>of <00:13:43.800>meaning, <00:13:43.956>in <00:13:44.111>terms <00:13:44.267>of <00:13:44.422>their <00:13:44.578>direction, <00:13:44.733>are <00:13:44.889>the <00:13:45.045>ones

317
00:13:46.325 --> 00:13:47.645
<00:13:46.325>that <00:13:46.523>are <00:13:46.721>pointing <00:13:46.919>in <00:13:47.117>the <00:13:47.314>same <00:13:47.513>direction.

318
00:13:47.645 --> 00:13:49.805
<00:13:47.645>So <00:13:47.772>this <00:13:47.899>one <00:13:48.026>down <00:13:48.154>at <00:13:48.281>the <00:13:48.408>bottom <00:13:48.535>and <00:13:48.663>the <00:13:48.790>one <00:13:48.917>out <00:13:49.045>here.

319
00:13:51.625 --> 00:13:55.765
<00:13:51.625>And <00:13:51.752>so <00:13:51.879>what <00:13:52.006>we <00:13:52.134>need <00:13:52.261>to <00:13:52.388>do <00:13:52.515>is <00:13:52.643>calculate, <00:13:52.770>uh, <00:13:52.897>the, <00:13:53.025>uh,

320
00:13:55.765 --> 00:13:57.605
<00:13:55.765>the <00:13:55.998>vectors <00:13:56.231>that <00:13:56.465>are <00:13:56.698>indeed <00:13:56.931>closest <00:13:57.165>in

321
00:13:57.605 --> 00:13:58.965
<00:13:57.605>that <00:13:57.911>respect, <00:13:58.217>uh, <00:13:58.523>by <00:13:58.829>angle.

322
00:13:59.345 --> 00:14:02.325
<00:13:59.345>And <00:13:59.520>that <00:13:59.695>brings <00:13:59.870>us <00:14:00.045>to <00:14:00.220>maths, <00:14:00.395>uh, <00:14:00.570>cosign <00:14:00.745>similarity.

323
00:14:03.385 --> 00:14:05.445
<00:14:03.385>Uh, <00:14:03.585>I'm <00:14:03.785>not <00:14:03.985>going <00:14:04.185>to <00:14:04.385>explain <00:14:04.585>coastline <00:14:04.785>similarity,

324
00:14:05.445 --> 00:14:07.325
<00:14:05.445>but <00:14:05.645>I <00:14:05.845>am <00:14:06.045>gonna <00:14:06.245>throw <00:14:06.445>a, <00:14:06.645>uh, <00:14:06.845>a,

325
00:14:07.805 --> 00:14:10.725
<00:14:07.805>a <00:14:07.960>formula <00:14:08.116>up <00:14:08.271>on <00:14:08.427>the <00:14:08.582>strain <00:14:08.738>just <00:14:08.893>to, <00:14:09.049>uh, <00:14:09.204>just

326
00:14:10.725 --> 00:14:12.965
<00:14:10.725>to <00:14:10.865>scare <00:14:11.005>anybody <00:14:11.145>who <00:14:11.285>wasn't <00:14:11.425>paying <00:14:11.565>attention <00:14:11.705>kind <00:14:11.845>of <00:14:11.985>in <00:14:12.125>the,

327
00:14:12.965 --> 00:14:13.965
<00:14:12.965>uh, <00:14:13.190>in <00:14:13.415>the <00:14:13.640>expo <00:14:13.865>still.

328
00:14:15.105 --> 00:14:17.285
<00:14:15.105>So, <00:14:15.338>coastline <00:14:15.571>similarity <00:14:15.805>of <00:14:16.038>two <00:14:16.271>vectors, <00:14:16.505>uh,

329
00:14:17.285 --> 00:14:19.165
<00:14:17.285>is <00:14:17.518>the <00:14:17.751>dot <00:14:17.985>product <00:14:18.218>of <00:14:18.451>two <00:14:18.685>vectors.

330
00:14:19.165 --> 00:14:22.245
<00:14:19.165>That <00:14:19.339>is, <00:14:19.515>uh, <00:14:19.689>each <00:14:19.865>item <00:14:20.040>in <00:14:20.214>the <00:14:20.390>vector <00:14:20.564>multiplied

331
00:14:22.245 --> 00:14:23.285
<00:14:22.245>by <00:14:22.557>each <00:14:22.869>other, <00:14:23.180>uh,

332
00:14:23.285 --> 00:14:25.565
<00:14:23.285>and <00:14:23.564>added <00:14:23.844>up <00:14:24.125>over <00:14:24.405>time, <00:14:24.685>divided

333
00:14:25.705 --> 00:14:28.685
<00:14:25.705>by <00:14:25.938>the <00:14:26.171>respective <00:14:26.405>magnitudes <00:14:26.638>of <00:14:26.871>each <00:14:27.105>vector.

334
00:14:28.705 --> 00:14:32.525
<00:14:28.705>That's <00:14:28.832>this, <00:14:28.959>um, <00:14:29.086>double <00:14:29.214>bar <00:14:29.341>A <00:14:29.468>and <00:14:29.595>B <00:14:29.723>at <00:14:29.850>the <00:14:29.977>bottom <00:14:30.105>here.

335
00:14:33.105 --> 00:14:36.205
<00:14:33.105>And <00:14:33.245>that's <00:14:33.385>just, <00:14:33.525>you <00:14:33.665>square <00:14:33.805>each <00:14:33.945>item <00:14:34.085>in, <00:14:34.225>uh, <00:14:34.365>the <00:14:34.505>array

336
00:14:36.205 --> 00:14:38.245
<00:14:36.205>and <00:14:36.345>add <00:14:36.485>the, <00:14:36.625>uh, <00:14:36.765>in <00:14:36.905>the <00:14:37.045>vector <00:14:37.185>and <00:14:37.325>add <00:14:37.465>them <00:14:37.605>up,

337
00:14:38.265 --> 00:14:40.245
<00:14:38.265>and <00:14:38.420>then <00:14:38.576>take <00:14:38.731>the <00:14:38.887>square <00:14:39.042>root <00:14:39.198>of <00:14:39.353>that <00:14:39.509>whole <00:14:39.665>thing.

338
00:14:41.185 --> 00:14:44.285
<00:14:41.185>But <00:14:41.340>like, <00:14:41.496>let's <00:14:41.651>see <00:14:41.807>that <00:14:41.962>in <00:14:42.118>Code <00:14:42.273>I <00:14:42.429>mathematical <00:14:42.584>formula.

339
00:14:46.105 --> 00:14:47.525
<00:14:46.105>I'd <00:14:46.247>say <00:14:46.389>it's <00:14:46.531>too <00:14:46.673>early <00:14:46.815>in <00:14:46.957>the <00:14:47.098>morning <00:14:47.241>for <00:14:47.383>it,

340
00:14:47.525 --> 00:14:48.525
<00:14:47.525>but <00:14:47.975>it's <00:14:48.425>enormous.

341
00:14:48.555 --> 00:14:50.725
<00:14:48.555>Four <00:14:48.729>o'clock, <00:14:48.905>still <00:14:49.079>feel <00:14:49.255>like <00:14:49.430>that's <00:14:49.604>too <00:14:49.780>early <00:14:49.954>though.

342
00:14:51.705 --> 00:14:54.765
<00:14:51.705>Um, <00:14:51.832>if <00:14:51.959>we <00:14:52.086>were <00:14:52.214>gonna <00:14:52.341>implement <00:14:52.468>that <00:14:52.595>as, <00:14:52.723>uh, <00:14:52.850>as <00:14:52.977>code, <00:14:53.105>uh,

343
00:14:54.765 --> 00:14:56.365
<00:14:54.765>we've <00:14:55.045>got <00:14:55.324>this <00:14:55.605>dot <00:14:55.885>product <00:14:56.165>magnitude,

344
00:14:56.365 --> 00:14:58.725
<00:14:56.365>and <00:14:56.505>then <00:14:56.645>a <00:14:56.785>co-sign <00:14:56.925>similarity <00:14:57.065>function <00:14:57.205>to <00:14:57.345>take <00:14:57.485>it <00:14:57.625>all <00:14:57.765>apart.

345
00:14:59.265 --> 00:15:02.045
<00:14:59.265>And <00:14:59.381>so <00:14:59.498>our <00:14:59.615>dot <00:14:59.731>product, <00:14:59.848>I'm <00:14:59.965>gonna <00:15:00.081>do <00:15:00.198>it. <00:15:00.314>Um, <00:15:00.431>there <00:15:00.548>we <00:15:00.665>go.

346
00:15:02.045 --> 00:15:03.045
<00:15:02.045>Product <00:15:02.344>starts <00:15:02.645>at <00:15:02.944>zero

347
00:15:03.145 --> 00:15:06.605
<00:15:03.145>and <00:15:03.300>then <00:15:03.456>four, <00:15:03.611>uh, <00:15:03.767>for <00:15:03.922>every <00:15:04.078>item <00:15:04.233>in <00:15:04.389>the <00:15:04.545>vector,

348
00:15:06.945 --> 00:15:08.005
<00:15:06.945>uh, <00:15:07.135>they <00:15:07.326>are <00:15:07.517>the <00:15:07.708>same <00:15:07.899>length.

349
00:15:08.005 --> 00:15:09.965
<00:15:08.005>So <00:15:08.180>we <00:15:08.355>can <00:15:08.530>just <00:15:08.705>check <00:15:08.880>the <00:15:09.055>vector <00:15:09.230>a <00:15:09.405>length.

350
00:15:10.265 --> 00:15:12.685
<00:15:10.265>Uh, <00:15:10.465>we're <00:15:10.665>going <00:15:10.865>to <00:15:11.064>add <00:15:11.265>on <00:15:11.465>the <00:15:11.665>product

351
00:15:12.945 --> 00:15:15.325
<00:15:12.945>of <00:15:13.100>the <00:15:13.256>two <00:15:13.411>items <00:15:13.567>at <00:15:13.722>the <00:15:13.878>vector <00:15:14.033>at <00:15:14.189>that <00:15:14.345>point.

352
00:15:16.225 --> 00:15:18.285
<00:15:16.225>Uh, <00:15:16.458>and <00:15:16.691>then <00:15:16.925>we <00:15:17.158>return <00:15:17.391>the <00:15:17.625>product.

353
00:15:19.145 --> 00:15:20.165
<00:15:19.145>And <00:15:19.328>so <00:15:19.512>that's <00:15:19.695>our <00:15:19.879>dot <00:15:20.063>product.

354
00:15:22.185 --> 00:15:23.285
<00:15:22.185>The <00:15:22.382>magnitude, <00:15:22.580>on <00:15:22.779>the <00:15:22.977>other <00:15:23.175>hand,

355
00:15:23.285 --> 00:15:24.645
<00:15:23.285>only <00:15:23.488>takes <00:15:23.693>one <00:15:23.896>vector <00:15:24.101>at <00:15:24.305>a <00:15:24.509>time.

356
00:15:25.665 --> 00:15:28.205
<00:15:25.665>Uh, <00:15:25.805>and <00:15:25.944>in <00:15:26.084>this <00:15:26.224>case, <00:15:26.365>I'm <00:15:26.505>gonna <00:15:26.645>call <00:15:26.785>it <00:15:26.925>a <00:15:27.064>sum.

357
00:15:28.985 --> 00:15:30.485
<00:15:28.985>And, <00:15:29.210>uh, <00:15:29.435>again, <00:15:29.660>we're <00:15:29.885>gonna <00:15:30.110>loop <00:15:30.335>over

358
00:15:31.295 --> 00:15:32.445
<00:15:31.295>everything <00:15:31.640>in <00:15:31.985>the <00:15:32.330>vector

359
00:15:33.025 --> 00:15:35.725
<00:15:33.025>and <00:15:33.165>add <00:15:33.305>up <00:15:33.444>to <00:15:33.584>the <00:15:33.725>sum <00:15:33.865>the <00:15:34.005>square <00:15:34.145>of <00:15:34.285>each <00:15:34.425>item,

360
00:15:37.235 --> 00:15:41.975
<00:15:37.235>and <00:15:37.410>then <00:15:37.585>return, <00:15:37.760>uh, <00:15:37.935>math <00:15:38.110>squirt, <00:15:38.285>uh, <00:15:38.460>for <00:15:38.635>sum.

361
00:15:42.755 --> 00:15:43.815
<00:15:42.755>And <00:15:42.993>so <00:15:43.232>that's <00:15:43.470>our <00:15:43.709>magnitude.

362
00:15:44.155 --> 00:15:47.015
<00:15:44.155>And <00:15:44.310>then <00:15:44.466>putting <00:15:44.621>the <00:15:44.777>two <00:15:44.932>things <00:15:45.088>together, <00:15:45.243>uh, <00:15:45.399>brings <00:15:45.555>us

363
00:15:47.015 --> 00:15:48.735
<00:15:47.015>to <00:15:47.481>cosign <00:15:47.948>similarity, <00:15:48.415>uh,

364
00:15:48.735 --> 00:15:50.855
<00:15:48.735>which <00:15:48.910>is <00:15:49.085>gonna <00:15:49.260>return <00:15:49.435>us <00:15:49.610>ultimately, <00:15:49.785>uh, <00:15:49.960>a <00:15:50.135>number

365
00:15:50.855 --> 00:15:52.335
<00:15:50.855>between <00:15:51.188>one <00:15:51.521>and <00:15:51.854>minus <00:15:52.187>one.

366
00:15:52.955 --> 00:15:55.735
<00:15:52.955>Uh, <00:15:53.095>where <00:15:53.235>one <00:15:53.375>is <00:15:53.515>fully <00:15:53.655>in <00:15:53.795>agreement, <00:15:53.935>it's <00:15:54.075>the <00:15:54.215>same <00:15:54.355>thing.

367
00:15:56.155 --> 00:15:58.655
<00:15:56.155>Uh, <00:15:56.355>zero <00:15:56.555>is <00:15:56.755>pretty <00:15:56.954>much <00:15:57.155>nothing <00:15:57.355>in <00:15:57.555>common.

368
00:15:58.675 --> 00:15:59.975
<00:15:58.675>And <00:15:58.967>minus <00:15:59.260>one <00:15:59.552>the <00:15:59.845>opposite,

369
00:16:00.365 --> 00:16:01.375
<00:16:00.365>like, <00:16:00.516>we're <00:16:00.668>going <00:16:00.819>in <00:16:00.971>the <00:16:01.122>other <00:16:01.274>way.

370
00:16:01.595 --> 00:16:03.455
<00:16:01.595>Oh, <00:16:01.875>and <00:16:02.155>coast, <00:16:02.435>ah, <00:16:02.715>thanks <00:16:02.995>copilot.

371
00:16:03.745 --> 00:16:05.495
<00:16:03.745>Leave <00:16:03.900>it <00:16:04.056>long <00:16:04.211>enough, <00:16:04.367>it <00:16:04.522>does <00:16:04.678>your <00:16:04.833>job <00:16:04.989>for <00:16:05.145>you.

372
00:16:05.835 --> 00:16:09.335
<00:16:05.835>Uh, <00:16:05.962>so <00:16:06.089>we <00:16:06.216>have <00:16:06.344>the <00:16:06.471>dot <00:16:06.598>product, <00:16:06.725>uh, <00:16:06.853>divided <00:16:06.980>by <00:16:07.107>the, <00:16:07.235>uh,

373
00:16:10.425 --> 00:16:12.805
<00:16:10.425>the <00:16:10.704>product <00:16:10.984>of <00:16:11.265>the <00:16:11.545>two <00:16:11.824>magnitudes.

374
00:16:13.745 --> 00:16:18.245
<00:16:13.745>And <00:16:13.885>so <00:16:14.025>using <00:16:14.165>this, <00:16:14.305>we <00:16:14.445>can <00:16:14.585>console <00:16:14.725>log, <00:16:14.865>uh, <00:16:15.005>our, <00:16:15.145>um,

375
00:16:18.835 --> 00:16:20.125
<00:16:18.835>talk <00:16:19.996>vectors.

376
00:16:21.695 --> 00:16:26.065
<00:16:21.695>What <00:16:21.835>are <00:16:21.975>they <00:16:22.115>called <00:16:22.255>again? <00:16:22.395>Talk <00:16:22.535>vectors. <00:16:22.675>Yeah, <00:16:22.815>there <00:16:22.955>we <00:16:23.095>go.

377
00:16:26.325 --> 00:16:30.785
<00:16:26.325>Um, <00:16:26.675>let's <00:16:27.025>co-sign <00:16:27.375>similarity, <00:16:27.725>uh,

378
00:16:31.735 --> 00:16:34.065
<00:16:31.735>that <00:16:32.435>with <00:16:33.135>itself.

379
00:16:37.335 --> 00:16:40.715
<00:16:37.335>And <00:16:37.451>if <00:16:37.568>I <00:16:37.685>do <00:16:37.801>that, <00:16:37.918>I <00:16:38.035>get <00:16:38.151>not <00:16:38.268>a <00:16:38.385>number. <00:16:38.501>Sweet. <00:16:38.618>It's <00:16:38.735>exciting.

380
00:16:42.655 --> 00:16:44.355
<00:16:42.655>Thanks. <00:16:43.355>JavaScript. <00:16:44.055>Um,

381
00:16:50.015 --> 00:16:52.095
<00:16:50.015>Ultimately <00:16:50.155>not <00:16:50.295>important, <00:16:50.435>but <00:16:50.574>I <00:16:50.715>was <00:16:50.855>just <00:16:50.995>trying <00:16:51.135>to <00:16:51.275>show <00:16:51.415>that

382
00:16:52.095 --> 00:16:53.095
<00:16:52.095>that <00:16:52.275>should <00:16:52.455>come <00:16:52.635>out <00:16:52.815>as <00:16:52.995>one.

383
00:16:53.595 --> 00:16:55.415
<00:16:53.595>Why <00:16:53.770>am <00:16:53.945>I <00:16:54.120>getting <00:16:54.295>that? <00:16:54.470>Come <00:16:54.645>on, <00:16:54.820>talk <00:16:54.995>vectors.

384
00:16:55.915 --> 00:16:59.055
<00:16:55.915>Um, <00:16:56.148>we <00:16:56.381>should <00:16:56.615>come <00:16:56.848>out <00:16:57.081>as <00:16:57.314>one.

385
00:16:59.355 --> 00:17:04.175
<00:16:59.355>Um, <00:16:59.821>which <00:17:00.288>is, <00:17:00.755>It's

386
00:17:04.175 --> 00:17:05.295
<00:17:04.175>gotta <00:17:04.511>sit, <00:17:04.847>is <00:17:05.183>it?

387
00:17:05.365 --> 00:17:06.215
<00:17:05.365>It's <00:17:05.619>index <00:17:05.875>one. <00:17:06.129>Ah,

388
00:17:06.215 --> 00:17:07.375
<00:17:06.214>you're <00:17:06.364>right, <00:17:06.513>you're <00:17:06.662>right, <00:17:06.811>you're <00:17:06.960>right, <00:17:07.109>you're <00:17:07.259>right.

389
00:17:08.045 --> 00:17:10.695
<00:17:08.045>Copilot. <00:17:08.220>Trying <00:17:08.395>to <00:17:08.570>help <00:17:08.745>too <00:17:08.920>much <00:17:09.095>there. <00:17:09.270>Thank <00:17:09.445>you.

390
00:17:11.835 --> 00:17:15.175
<00:17:11.835>Not <00:17:12.035>too, <00:17:12.235>not <00:17:12.435>too. <00:17:12.635>Uh, <00:17:12.835>there <00:17:13.035>we <00:17:13.235>go.

391
00:17:15.235 --> 00:17:17.375
<00:17:15.234>So, <00:17:15.468>um, <00:17:15.701>again, <00:17:15.935>actually <00:17:16.168>due <00:17:16.401>to <00:17:16.635>JavaScript,

392
00:17:17.435 --> 00:17:20.615
<00:17:17.435>the <00:17:17.575>similarity <00:17:17.714>is <00:17:17.855>now <00:17:17.994>1.0 <00:17:18.135>0 <00:17:18.274>0 <00:17:18.415>0 <00:17:18.554>0 <00:17:18.694>0 <00:17:18.835>2.

393
00:17:21.075 --> 00:17:23.215
<00:17:21.075>Um, <00:17:21.355>thanks. <00:17:21.635>Thank <00:17:21.915>you <00:17:22.194>to <00:17:22.475>JavaScript.

394
00:17:23.355 --> 00:17:25.015
<00:17:23.355>But <00:17:23.555>if <00:17:23.755>we <00:17:23.954>now <00:17:24.155>compare <00:17:24.355>my <00:17:24.555>talk, <00:17:24.755>say

395
00:17:25.015 --> 00:17:27.495
<00:17:25.015>to <00:17:25.155>the <00:17:25.295>first <00:17:25.435>one <00:17:25.575>about <00:17:25.715>threat <00:17:25.855>modeling, <00:17:25.995>uh, <00:17:26.135>we <00:17:26.275>expect <00:17:26.415>that

396
00:17:27.495 --> 00:17:29.375
<00:17:27.494>to <00:17:27.844>be <00:17:28.194>less, <00:17:28.544>um, <00:17:28.895>similar,

397
00:17:29.475 --> 00:17:32.735
<00:17:29.475>and <00:17:29.630>it <00:17:29.786>is, <00:17:29.941>it <00:17:30.097>comes <00:17:30.252>out <00:17:30.408>as <00:17:30.563>0.04, <00:17:30.719>et <00:17:30.875>cetera.

398
00:17:33.475 --> 00:17:36.935
<00:17:33.475>Um, <00:17:33.708>so <00:17:33.941>our <00:17:34.175>cosign <00:17:34.408>is <00:17:34.641>working, <00:17:34.875>uh,

399
00:17:36.935 --> 00:17:38.735
<00:17:36.935>even <00:17:37.135>though <00:17:37.335>I <00:17:37.534>wasn't <00:17:37.735>helping <00:17:37.935>myself <00:17:38.135>show <00:17:38.335>that.

400
00:17:39.235 --> 00:17:42.095
<00:17:39.235>Uh, <00:17:39.390>so <00:17:39.546>finally, <00:17:39.701>we <00:17:39.857>need <00:17:40.012>to, <00:17:40.168>uh, <00:17:40.323>find <00:17:40.479>similar <00:17:40.635>talks.

401
00:17:42.095 --> 00:17:44.055
<00:17:42.095>When <00:17:42.235>we're <00:17:42.375>given <00:17:42.515>a <00:17:42.655>talk, <00:17:42.795>we <00:17:42.935>need <00:17:43.075>to <00:17:43.215>find <00:17:43.355>similar <00:17:43.495>ones.

402
00:17:44.085 --> 00:17:45.815
<00:17:44.085>Currently <00:17:44.365>we're <00:17:44.645>returning <00:17:44.925>an <00:17:45.205>empty <00:17:45.485>list.

403
00:17:46.235 --> 00:17:48.495
<00:17:46.235>Uh, <00:17:46.362>now <00:17:46.489>I <00:17:46.616>want <00:17:46.744>to <00:17:46.871>take, <00:17:46.998>I <00:17:47.125>wanna <00:17:47.253>get <00:17:47.380>the <00:17:47.507>query <00:17:47.635>vector

404
00:17:48.685 --> 00:17:50.525
<00:17:48.685>that <00:17:48.965>takes <00:17:49.245>the, <00:17:49.524>the <00:17:49.804>incoming <00:17:50.085>query.

405
00:17:50.545 --> 00:17:52.565
<00:17:50.545>The <00:17:50.745>user <00:17:50.945>query <00:17:51.145>turns <00:17:51.345>that <00:17:51.545>into <00:17:51.745>a <00:17:51.945>vector,

406
00:17:52.985 --> 00:17:55.165
<00:17:52.985>and <00:17:53.140>then <00:17:53.296>we're <00:17:53.451>gonna <00:17:53.607>map <00:17:53.762>over <00:17:53.918>the <00:17:54.073>torques <00:17:54.229>and <00:17:54.385>vectors.

407
00:17:55.705 --> 00:17:59.325
<00:17:55.705>I'm <00:17:55.845>actually <00:17:55.985>gonna <00:17:56.125>return, <00:17:56.264>uh, <00:17:56.405>a <00:17:56.544>map <00:17:56.685>of <00:17:56.824>the <00:17:56.965>torque <00:17:57.105>vector

408
00:17:59.385 --> 00:18:01.205
<00:17:59.385>and <00:17:59.665>the <00:17:59.945>co-sign <00:18:00.225>similarity <00:18:00.504>of <00:18:00.785>the,

409
00:18:01.305 --> 00:18:02.685
<00:18:01.305>of <00:18:01.553>the <00:18:01.801>query <00:18:02.050>and <00:18:02.298>the <00:18:02.547>vector.

410
00:18:03.195 --> 00:18:05.485
<00:18:03.195>Then <00:18:03.428>we're <00:18:03.661>gonna <00:18:03.895>sort <00:18:04.128>by <00:18:04.361>the <00:18:04.595>similarity,

411
00:18:05.485 --> 00:18:07.325
<00:18:05.485>which <00:18:05.718>is <00:18:05.951>actually <00:18:06.185>in <00:18:06.418>the <00:18:06.651>second <00:18:06.885>section,

412
00:18:07.945 --> 00:18:09.165
<00:18:07.945>and <00:18:08.128>we're <00:18:08.311>gonna <00:18:08.494>slice <00:18:08.677>by <00:18:08.860>the <00:18:09.043>limit.

413
00:18:09.165 --> 00:18:10.605
<00:18:09.165>So <00:18:09.350>I'm <00:18:09.535>just <00:18:09.720>gonna <00:18:09.905>try <00:18:10.090>and <00:18:10.275>return <00:18:10.461>five.

414
00:18:11.265 --> 00:18:14.285
<00:18:11.265>And <00:18:11.420>so <00:18:11.576>if <00:18:11.731>I <00:18:11.887>look <00:18:12.042>to, <00:18:12.198>well, <00:18:12.353>rather <00:18:12.509>than <00:18:12.665>try

415
00:18:14.285 --> 00:18:15.765
<00:18:14.285>and <00:18:14.475>now <00:18:14.665>call <00:18:14.855>this <00:18:15.046>out <00:18:15.236>in <00:18:15.426>console, <00:18:15.617>uh,

416
00:18:15.765 --> 00:18:17.605
<00:18:15.765>this <00:18:15.965>is <00:18:16.165>now <00:18:16.365>hooked <00:18:16.565>up <00:18:16.765>to <00:18:16.965>my <00:18:17.165>application.

417
00:18:18.545 --> 00:18:22.565
<00:18:18.545>Um, <00:18:18.661>and <00:18:18.778>so <00:18:18.895>I, <00:18:19.011>I've <00:18:19.128>got <00:18:19.245>a <00:18:19.361>search, <00:18:19.478>uh, <00:18:19.595>kind <00:18:19.711>of <00:18:19.828>bar <00:18:19.945>here,

418
00:18:22.905 --> 00:18:25.085
<00:18:22.905>so <00:18:23.060>I <00:18:23.216>can <00:18:23.371>ask <00:18:23.527>like, <00:18:23.682>who <00:18:23.838>is, <00:18:23.993>uh, <00:18:24.149>speaking <00:18:24.305>about

419
00:18:25.725 --> 00:18:26.885
<00:18:25.725>retrieval <00:18:26.246>augmented <00:18:26.769>generation?

420
00:18:26.885 --> 00:18:28.965
<00:18:26.885>And <00:18:27.060>this <00:18:27.235>is <00:18:27.410>gonna <00:18:27.585>return <00:18:27.760>the <00:18:27.935>five <00:18:28.110>most <00:18:28.285>similar

421
00:18:29.975 --> 00:18:32.485
<00:18:29.975>talks <00:18:30.208>according <00:18:30.441>to <00:18:30.675>my <00:18:30.908>bag <00:18:31.141>of <00:18:31.375>words,

422
00:18:32.485 --> 00:18:34.845
<00:18:32.485>vector <00:18:32.834>representation, <00:18:33.185>uh, <00:18:33.534>of <00:18:33.885>that.

423
00:18:35.385 --> 00:18:36.685
<00:18:35.385>And, <00:18:35.619>uh, <00:18:35.853>that's <00:18:36.087>pretty <00:18:36.321>good, <00:18:36.554>right?

424
00:18:36.685 --> 00:18:39.285
<00:18:36.685>So <00:18:36.840>our <00:18:36.996>first <00:18:37.151>one <00:18:37.307>is <00:18:37.462>exploring <00:18:37.618>advanced <00:18:37.773>patterns <00:18:37.929>in <00:18:38.085>retrieval

425
00:18:39.285 --> 00:18:42.765
<00:18:39.285>augmented <00:18:39.485>generation, <00:18:39.685>uh, <00:18:39.885>by <00:18:40.085>Carl <00:18:40.285>Lapier <00:18:40.485>similarity <00:18:40.685>of,

426
00:18:42.765 --> 00:18:43.925
<00:18:42.765>uh, <00:18:43.287>north <00:18:43.809>0.28.

427
00:18:44.575 --> 00:18:47.485
<00:18:44.575>Seems <00:18:44.715>like <00:18:44.855>it <00:18:44.995>should <00:18:45.135>be <00:18:45.275>more <00:18:45.415>similar, <00:18:45.555>but <00:18:45.695>I <00:18:45.835>think <00:18:45.975>it's

428
00:18:47.485 --> 00:18:49.205
<00:18:47.485>because <00:18:47.685>it's <00:18:47.885>a <00:18:48.084>bad <00:18:48.284>vector, <00:18:48.485>but <00:18:48.685>that's <00:18:48.885>okay.

429
00:18:49.205 --> 00:18:50.485
<00:18:49.205>We <00:18:49.397>meant <00:18:49.589>to <00:18:49.781>build <00:18:49.973>a <00:18:50.165>bad <00:18:50.357>vector.

430
00:18:50.985 --> 00:18:52.885
<00:18:50.985>Uh, <00:18:51.140>we <00:18:51.296>then <00:18:51.451>have <00:18:51.607>my <00:18:51.762>talk <00:18:51.918>build <00:18:52.073>rag <00:18:52.229>from <00:18:52.385>scratch.

431
00:18:52.885 --> 00:18:55.165
<00:18:52.885>That's <00:18:53.165>cool. <00:18:53.445>Uh, <00:18:53.725>similarity <00:18:54.004>of <00:18:54.285>0.20.

432
00:18:55.695 --> 00:18:58.845
<00:18:55.695>Large <00:18:55.895>language <00:18:56.095>models, <00:18:56.294>love <00:18:56.495>knowledge <00:18:56.695>graphs. <00:18:56.895>Um, <00:18:57.095>cool.

433
00:18:58.905 --> 00:19:00.205
<00:18:58.905>And <00:18:59.072>so <00:18:59.239>actually, <00:18:59.406>if <00:18:59.573>I <00:18:59.740>were <00:18:59.907>gonna <00:19:00.074>go

434
00:19:00.205 --> 00:19:03.205
<00:19:00.205>and <00:19:00.379>ask <00:19:00.554>the <00:19:00.730>same <00:19:00.905>question, <00:19:01.080>uh, <00:19:01.254>over <00:19:01.429>here, <00:19:01.605>uh,

435
00:19:03.385 --> 00:19:05.405
<00:19:03.385>who <00:19:03.618>is <00:19:03.851>speaking <00:19:04.085>about <00:19:04.318>retrieval <00:19:04.551>augmented <00:19:04.785>generation?

436
00:19:06.465 --> 00:19:08.205
<00:19:06.465>We <00:19:06.665>are <00:19:06.865>now <00:19:07.064>feeding <00:19:07.264>our <00:19:07.465>model <00:19:07.665>the <00:19:07.865>data

437
00:19:08.205 --> 00:19:09.325
<00:19:08.205>that <00:19:08.373>we <00:19:08.541>saw <00:19:08.709>there <00:19:08.877>in <00:19:09.045>the <00:19:09.213>search.

438
00:19:09.865 --> 00:19:12.805
<00:19:09.865>And, <00:19:10.005>uh, <00:19:10.145>Gemini <00:19:10.285>in <00:19:10.425>this <00:19:10.565>case <00:19:10.705>can <00:19:10.845>return <00:19:10.985>a <00:19:11.125>whole <00:19:11.265>list

439
00:19:12.805 --> 00:19:13.885
<00:19:12.805>of <00:19:12.967>talks <00:19:13.129>that <00:19:13.291>we <00:19:13.453>might <00:19:13.615>wanna <00:19:13.777>go

440
00:19:13.885 --> 00:19:15.325
<00:19:13.885>and <00:19:14.144>see <00:19:14.403>about <00:19:14.662>retrieval <00:19:14.921>augmented <00:19:15.181>generation.

441
00:19:15.925 --> 00:19:17.965
<00:19:15.925>I <00:19:16.080>also <00:19:16.236>particularly <00:19:16.391>like <00:19:16.547>how <00:19:16.702>it <00:19:16.858>calls <00:19:17.013>out <00:19:17.169>that, <00:19:17.325>uh,

442
00:19:18.025 --> 00:19:19.645
<00:19:18.025>the <00:19:18.258>large <00:19:18.491>language <00:19:18.725>models <00:19:18.958>and <00:19:19.191>model <00:19:19.425>graph

443
00:19:19.825 --> 00:19:20.925
<00:19:19.825>did <00:19:20.072>not <00:19:20.320>provide <00:19:20.567>a <00:19:20.815>speaker.

444
00:19:21.265 --> 00:19:23.045
<00:19:21.265>And <00:19:21.465>it's <00:19:21.665>pointed <00:19:21.865>that <00:19:22.065>out, <00:19:22.265>hopefully. <00:19:22.465>Thank <00:19:22.665>you.

445
00:19:25.345 --> 00:19:28.925
<00:19:25.345>So <00:19:25.472>the <00:19:25.599>first <00:19:25.726>time <00:19:25.854>I <00:19:25.981>did <00:19:26.108>that, <00:19:26.235>I <00:19:26.363>was <00:19:26.490>actually <00:19:26.617>quite <00:19:26.745>impressed

446
00:19:28.925 --> 00:19:31.285
<00:19:28.925>that <00:19:29.080>it <00:19:29.236>kind <00:19:29.391>of <00:19:29.547>worked <00:19:29.702>like <00:19:29.858>it's <00:19:30.013>a <00:19:30.169>bad <00:19:30.325>vector

447
00:19:31.285 --> 00:19:33.485
<00:19:31.285>representation, <00:19:31.518>and <00:19:31.751>the <00:19:31.985>similarity <00:19:32.218>is <00:19:32.451>not <00:19:32.685>high,

448
00:19:34.065 --> 00:19:37.965
<00:19:34.065>but <00:19:34.345>it <00:19:34.625>works <00:19:34.905>sort <00:19:35.185>of, <00:19:35.465>right?

449
00:19:38.225 --> 00:19:39.565
<00:19:38.225>Um, <00:19:38.627>you <00:19:39.029>see, <00:19:39.431>actually

450
00:19:39.565 --> 00:19:41.285
<00:19:39.565>what <00:19:39.765>we <00:19:39.965>needed <00:19:40.165>in <00:19:40.365>this <00:19:40.565>particular <00:19:40.765>case <00:19:40.965>was

451
00:19:41.285 --> 00:19:42.285
<00:19:41.285>to <00:19:41.465>know <00:19:41.645>all <00:19:41.825>the <00:19:42.005>data <00:19:42.185>upfront.

452
00:19:42.425 --> 00:19:45.565
<00:19:42.425>We <00:19:42.541>knew <00:19:42.658>all <00:19:42.774>the <00:19:42.891>talks, <00:19:43.008>so <00:19:43.125>we <00:19:43.241>could <00:19:43.358>build <00:19:43.475>that <00:19:43.591>map <00:19:43.708>of <00:19:43.825>all

453
00:19:45.565 --> 00:19:47.925
<00:19:45.565>of <00:19:45.740>the <00:19:45.915>words <00:19:46.090>involved <00:19:46.265>in <00:19:46.440>them, <00:19:46.615>uh, <00:19:46.790>and, <00:19:46.965>uh,

454
00:19:47.925 --> 00:19:51.525
<00:19:47.925>and <00:19:48.125>use <00:19:48.325>that <00:19:48.524>the <00:19:48.725>query <00:19:48.925>itself <00:19:49.125>is <00:19:49.325>actually

455
00:19:51.805 --> 00:19:52.845
<00:19:51.805>sensitive <00:19:52.117>to <00:19:52.429>the <00:19:52.741>vocabulary.

456
00:19:52.915 --> 00:19:55.245
<00:19:52.915>Like <00:19:53.031>if <00:19:53.148>you <00:19:53.264>use <00:19:53.381>a <00:19:53.498>bunch <00:19:53.615>of <00:19:53.731>words <00:19:53.848>that <00:19:53.965>were <00:19:54.081>not <00:19:54.198>in <00:19:54.315>the

457
00:19:55.795 --> 00:19:59.005
<00:19:55.795>talk <00:19:55.995>descriptions, <00:19:56.195>you <00:19:56.395>get <00:19:56.595>nothing <00:19:56.795>out <00:19:56.995>of <00:19:57.195>it.

458
00:19:59.385 --> 00:20:03.285
<00:19:59.385>It <00:19:59.585>doesn't <00:19:59.785>help, <00:19:59.985>it <00:20:00.185>can't <00:20:00.385>know <00:20:00.585>More <00:20:00.785>words.

459
00:20:03.465 --> 00:20:04.965
<00:20:03.465>And <00:20:03.633>if <00:20:03.802>we <00:20:03.971>did <00:20:04.139>this <00:20:04.308>on <00:20:04.477>a <00:20:04.646>bigger <00:20:04.814>dataset,

460
00:20:04.965 --> 00:20:06.845
<00:20:04.965>this <00:20:05.165>is <00:20:05.365>only <00:20:05.564>like <00:20:05.764>350 <00:20:05.965>talks <00:20:06.165>or <00:20:06.365>so.

461
00:20:06.845 --> 00:20:09.285
<00:20:06.845>If <00:20:06.985>we <00:20:07.125>had <00:20:07.265>more <00:20:07.405>words, <00:20:07.545>we <00:20:07.685>had <00:20:07.825>to <00:20:07.965>do <00:20:08.105>more <00:20:08.245>calculations.

462
00:20:09.545 --> 00:20:11.605
<00:20:09.545>I'm <00:20:09.720>actually <00:20:09.895>quite <00:20:10.070>impressed <00:20:10.245>with <00:20:10.420>how <00:20:10.595>quickly <00:20:10.770>JavaScript <00:20:10.945>is,

463
00:20:11.605 --> 00:20:13.605
<00:20:11.605>are <00:20:11.760>able <00:20:11.916>to <00:20:12.071>dock <00:20:12.227>product <00:20:12.382>a <00:20:12.538>whole <00:20:12.693>bunch <00:20:12.849>of <00:20:13.005>things.

464
00:20:14.265 --> 00:20:15.765
<00:20:14.265>Um, <00:20:14.940>but, <00:20:15.615>uh,

465
00:20:15.765 --> 00:20:16.565
<00:20:15.765>more <00:20:16.005>words <00:20:16.245>mean <00:20:16.485>more

466
00:20:16.565 --> 00:20:18.125
<00:20:16.565>calculations <00:20:16.798>mean <00:20:17.031>it's <00:20:17.265>going <00:20:17.498>to <00:20:17.731>be <00:20:17.965>slower.

467
00:20:18.425 --> 00:20:21.165
<00:20:18.425>And <00:20:18.565>yet, <00:20:18.705>if <00:20:18.845>we <00:20:18.985>are, <00:20:19.125>uh, <00:20:19.264>doing <00:20:19.405>this <00:20:19.544>to <00:20:19.685>thousands, <00:20:19.825>hundreds

468
00:20:21.165 --> 00:20:23.245
<00:20:21.165>of <00:20:21.398>thousands <00:20:21.631>or <00:20:21.865>millions <00:20:22.098>of <00:20:22.331>documents, <00:20:22.565>uh,

469
00:20:23.245 --> 00:20:25.405
<00:20:23.245>that's <00:20:23.445>gonna <00:20:23.645>grind <00:20:23.844>to <00:20:24.044>a <00:20:24.245>halt <00:20:24.445>very <00:20:24.645>quickly.

470
00:20:27.265 --> 00:20:29.125
<00:20:27.265>And <00:20:27.498>then <00:20:27.731>there's <00:20:27.965>actual <00:20:28.198>meaning <00:20:28.431>of <00:20:28.665>things.

471
00:20:29.385 --> 00:20:31.085
<00:20:29.385>Now, <00:20:29.560>I <00:20:29.735>said <00:20:29.910>earlier, <00:20:30.085>we're <00:20:30.260>actually <00:20:30.435>kind <00:20:30.610>of <00:20:30.785>fortunate

472
00:20:31.085 --> 00:20:34.685
<00:20:31.085>that <00:20:31.212>in <00:20:31.339>a <00:20:31.466>list <00:20:31.594>of <00:20:31.721>350 <00:20:31.848>talks, <00:20:31.975>uh, <00:20:32.103>at <00:20:32.230>a <00:20:32.357>developer <00:20:32.485>conference,

473
00:20:34.935 --> 00:20:36.245
<00:20:34.935>we're <00:20:35.170>probably <00:20:35.406>gonna <00:20:35.642>find <00:20:35.878>that <00:20:36.113>most

474
00:20:36.245 --> 00:20:38.205
<00:20:36.245>of <00:20:36.445>the <00:20:36.645>words <00:20:36.844>in <00:20:37.044>them <00:20:37.245>mean <00:20:37.445>similar <00:20:37.645>things.

475
00:20:38.905 --> 00:20:41.085
<00:20:38.905>Um, <00:20:39.045>but <00:20:39.185>the <00:20:39.325>same <00:20:39.465>word <00:20:39.605>can <00:20:39.745>of <00:20:39.885>course <00:20:40.024>mean <00:20:40.165>different <00:20:40.305>things.

476
00:20:41.655 --> 00:20:42.445
<00:20:41.655>Let's <00:20:41.892>go <00:20:42.129>back <00:20:42.366>there.

477
00:20:43.025 --> 00:20:44.205
<00:20:43.025>The <00:20:43.202>same <00:20:43.379>word <00:20:43.556>can <00:20:43.733>mean <00:20:43.909>different <00:20:44.087>things.

478
00:20:44.705 --> 00:20:47.005
<00:20:44.705>Um, <00:20:45.054>different <00:20:45.405>phrasing <00:20:45.754>of, <00:20:46.105>uh,

479
00:20:47.105 --> 00:20:48.485
<00:20:47.105>of <00:20:47.353>words <00:20:47.601>can <00:20:47.850>mean <00:20:48.098>different <00:20:48.347>things.

480
00:20:48.645 --> 00:20:51.205
<00:20:48.645>I <00:20:48.785>like <00:20:48.925>the, <00:20:49.065>uh, <00:20:49.205>the <00:20:49.345>fact <00:20:49.485>that <00:20:49.625>man <00:20:49.764>bites <00:20:49.905>dog, <00:20:50.045>uh,

481
00:20:51.205 --> 00:20:53.525
<00:20:51.205>and <00:20:51.345>dog <00:20:51.485>bites <00:20:51.625>man <00:20:51.764>have <00:20:51.905>the <00:20:52.044>exact <00:20:52.185>same <00:20:52.324>words <00:20:52.465>in <00:20:52.605>him,

482
00:20:53.525 --> 00:20:54.725
<00:20:53.525>have <00:20:53.795>mean <00:20:54.065>very <00:20:54.335>different <00:20:54.605>things.

483
00:20:55.825 --> 00:20:57.005
<00:20:55.825>Um, <00:20:56.179>and <00:20:56.533>similarly, <00:20:56.887>like,

484
00:20:57.005 --> 00:20:57.925
<00:20:57.005>we're <00:20:57.212>actually <00:20:57.419>not <00:20:57.626>doing <00:20:57.833>anything

485
00:20:57.925 --> 00:20:59.085
<00:20:57.925>clever <00:20:58.133>with <00:20:58.342>any <00:20:58.551>of <00:20:58.760>those <00:20:58.969>words.

486
00:20:59.585 --> 00:21:02.805
<00:20:59.585>Uh, <00:20:59.785>and <00:20:59.985>so, <00:21:00.185>um, <00:21:00.385>you <00:21:00.585>know, <00:21:00.785>um, <00:21:00.985>the,

487
00:21:02.825 --> 00:21:04.325
<00:21:02.825>the <00:21:03.095>threat <00:21:03.365>modeling <00:21:03.635>talk <00:21:03.905>actually, <00:21:04.175>uh,

488
00:21:04.495 --> 00:21:06.085
<00:21:04.495>talks <00:21:04.774>about <00:21:05.054>threats <00:21:05.334>at <00:21:05.614>one <00:21:05.895>point,

489
00:21:06.345 --> 00:21:08.005
<00:21:06.345>but <00:21:06.545>threats <00:21:06.745>is <00:21:06.945>a <00:21:07.145>different <00:21:07.345>word <00:21:07.545>to <00:21:07.745>threat.

490
00:21:08.505 --> 00:21:09.805
<00:21:08.505>We <00:21:08.672>could <00:21:08.839>do <00:21:09.006>work <00:21:09.173>to <00:21:09.340>clear <00:21:09.507>that <00:21:09.675>up,

491
00:21:10.065 --> 00:21:13.205
<00:21:10.065>but <00:21:10.192>really <00:21:10.319>what <00:21:10.446>we <00:21:10.574>actually <00:21:10.701>want <00:21:10.828>to <00:21:10.955>do <00:21:11.083>is, <00:21:11.210>uh, <00:21:11.337>avoid <00:21:11.465>all

492
00:21:13.205 --> 00:21:14.285
<00:21:13.205>that <00:21:13.447>together <00:21:13.691>and <00:21:13.934>move <00:21:14.177>on

493
00:21:14.285 --> 00:21:15.805
<00:21:14.285>to <00:21:14.627>actual <00:21:14.969>embedding <00:21:15.311>models <00:21:15.653>themselves.

494
00:21:16.475 --> 00:21:20.685
<00:21:16.475>This <00:21:16.615>is <00:21:16.754>our, <00:21:16.895>um, <00:21:17.034>embedding <00:21:17.175>models <00:21:17.314>are <00:21:17.455>made <00:21:17.594>available <00:21:17.735>to <00:21:17.875>us

495
00:21:20.905 --> 00:21:24.365
<00:21:20.905>by, <00:21:21.060>uh, <00:21:21.216>people <00:21:21.371>who <00:21:21.527>have <00:21:21.682>built, <00:21:21.838>uh, <00:21:21.993>large <00:21:22.149>language <00:21:22.305>models,

496
00:21:24.705 --> 00:21:29.565
<00:21:24.705>uh, <00:21:24.845>and <00:21:24.985>give <00:21:25.125>us <00:21:25.264>a <00:21:25.405>much <00:21:25.544>better <00:21:25.685>version <00:21:25.824>of <00:21:25.965>meaning, <00:21:26.105>uh,

497
00:21:29.595 --> 00:21:31.685
<00:21:29.595>than <00:21:29.795>my, <00:21:29.995>uh, <00:21:30.195>silly <00:21:30.395>little <00:21:30.595>bag <00:21:30.795>of <00:21:30.995>words

498
00:21:31.685 --> 00:21:33.205
<00:21:31.685>that <00:21:31.913>we <00:21:32.140>had, <00:21:32.369>uh, <00:21:32.597>a <00:21:32.824>minute <00:21:33.053>ago.

499
00:21:34.025 --> 00:21:36.205
<00:21:34.025>Uh, <00:21:34.375>embedding <00:21:34.725>models <00:21:35.075>are, <00:21:35.425>um,

500
00:21:36.985 --> 00:21:40.605
<00:21:36.985>and <00:21:37.125>then <00:21:37.264>they <00:21:37.405>use <00:21:37.544>the, <00:21:37.685>uh, <00:21:37.824>attention <00:21:37.965>kind <00:21:38.104>of <00:21:38.245>mechanism, <00:21:38.385>uh,

501
00:21:40.605 --> 00:21:43.845
<00:21:40.605>that <00:21:40.780>large <00:21:40.955>language <00:21:41.130>models <00:21:41.305>use <00:21:41.480>to, <00:21:41.655>uh, <00:21:41.830>uh, <00:21:42.005>kind

502
00:21:43.845 --> 00:21:47.045
<00:21:43.845>of <00:21:44.020>track <00:21:44.195>meaning <00:21:44.370>through, <00:21:44.545>uh, <00:21:44.720>sentences, <00:21:44.895>through <00:21:45.070>texts, <00:21:45.245>uh,

503
00:21:47.065 --> 00:21:50.045
<00:21:47.065>in <00:21:47.205>order <00:21:47.345>to <00:21:47.485>position, <00:21:47.625>uh, <00:21:47.765>in <00:21:47.905>order <00:21:48.045>to <00:21:48.185>create <00:21:48.325>a <00:21:48.465>vector

504
00:21:50.045 --> 00:21:53.085
<00:21:50.045>that <00:21:50.220>ultimately <00:21:50.395>yeah, <00:21:50.570>captures <00:21:50.745>the <00:21:50.920>meaning <00:21:51.095>of <00:21:51.270>those <00:21:51.445>words.

505
00:21:53.345 --> 00:21:55.525
<00:21:53.345>Uh, <00:21:53.695>including <00:21:54.045>negatives, <00:21:54.395>including, <00:21:54.745>um,

506
00:21:55.535 --> 00:21:56.605
<00:21:55.535>words <00:21:55.856>in <00:21:56.177>opposite <00:21:56.498>orders

507
00:21:56.605 --> 00:21:57.805
<00:21:56.605>to <00:21:56.821>each <00:21:57.037>other, <00:21:57.253>meaning <00:21:57.469>different <00:21:57.685>things.

508
00:21:59.515 --> 00:22:01.525
<00:21:59.515>It's <00:21:59.690>nice <00:21:59.865>and <00:22:00.040>easy <00:22:00.215>to <00:22:00.390>use <00:22:00.565>these <00:22:00.740>embedding <00:22:00.915>models.

509
00:22:01.745 --> 00:22:03.565
<00:22:01.745>Uh, <00:22:01.919>at <00:22:02.094>least <00:22:02.270>if <00:22:02.445>you <00:22:02.620>pick <00:22:02.794>something <00:22:02.969>like <00:22:03.145>OpenAI,

510
00:22:03.565 --> 00:22:04.845
<00:22:03.565>which <00:22:03.757>has <00:22:03.949>a <00:22:04.141>great <00:22:04.333>one, <00:22:04.525>text <00:22:04.717>embedding

511
00:22:04.845 --> 00:22:05.965
<00:22:04.845>three <00:22:05.097>small, <00:22:05.349>it's <00:22:05.601>very <00:22:05.852>useful.

512
00:22:06.225 --> 00:22:09.965
<00:22:06.225>Uh, <00:22:06.380>Google <00:22:06.536>has <00:22:06.691>one. <00:22:06.847>Um, <00:22:07.002>actually <00:22:07.158>OpenAI <00:22:07.313>has <00:22:07.469>twice, <00:22:07.625>right?

513
00:22:10.185 --> 00:22:11.245
<00:22:10.185>Uh, <00:22:10.375>this <00:22:10.566>is <00:22:10.757>just <00:22:10.948>in <00:22:11.139>JavaScript,

514
00:22:11.245 --> 00:22:14.365
<00:22:11.245>and <00:22:11.372>if <00:22:11.499>you <00:22:11.626>were <00:22:11.754>to <00:22:11.881>use <00:22:12.008>an <00:22:12.135>OpenAI <00:22:12.263>embedding, <00:22:12.390>like, <00:22:12.517>it's, <00:22:12.645>it's

515
00:22:14.365 --> 00:22:15.725
<00:22:14.365>that <00:22:14.539>many <00:22:14.714>lines <00:22:14.889>of <00:22:15.064>code, <00:22:15.239>it's <00:22:15.414>job <00:22:15.589>done.

516
00:22:16.345 --> 00:22:20.165
<00:22:16.345>I'm, <00:22:16.461>uh, <00:22:16.578>I'm <00:22:16.695>excited <00:22:16.811>by <00:22:16.928>how <00:22:17.045>easy <00:22:17.161>stuff <00:22:17.278>like <00:22:17.395>this <00:22:17.511>is <00:22:17.628>for, <00:22:17.745>uh,

517
00:22:20.165 --> 00:22:22.765
<00:22:20.165>kind <00:22:20.320>of <00:22:20.476>web <00:22:20.631>developers <00:22:20.787>to <00:22:20.942>use, <00:22:21.098>uh, <00:22:21.253>once <00:22:21.409>you <00:22:21.565>understand

518
00:22:22.765 --> 00:22:23.805
<00:22:22.765>what <00:22:22.952>you're <00:22:23.139>actually <00:22:23.326>doing <00:22:23.513>with <00:22:23.701>it.

519
00:22:25.065 --> 00:22:27.515
<00:22:25.065>Um, <00:22:25.298>and <00:22:25.531>then, <00:22:25.765>so <00:22:25.998>the <00:22:26.231>second <00:22:26.465>part

520
00:22:27.515 --> 00:22:30.155
<00:22:27.515>of <00:22:27.670>this <00:22:27.826>embedding <00:22:27.981>model <00:22:28.137>like <00:22:28.292>that <00:22:28.448>is <00:22:28.603>gonna <00:22:28.759>capture, <00:22:28.915>uh,

521
00:22:30.155 --> 00:22:32.595
<00:22:30.155>that <00:22:30.310>data <00:22:30.466>into <00:22:30.621>a <00:22:30.777>vector <00:22:30.932>that <00:22:31.088>is <00:22:31.243>more <00:22:31.399>densely <00:22:31.555>packed.

522
00:22:32.895 --> 00:22:34.715
<00:22:32.895>Uh, <00:22:33.175>so <00:22:33.455>OpenAI, <00:22:33.735>for <00:22:34.014>example, <00:22:34.295>uh,

523
00:22:34.715 --> 00:22:37.355
<00:22:34.715>will <00:22:34.995>give <00:22:35.274>you <00:22:35.554>1,536 <00:22:35.834>dimensions, <00:22:36.115>uh,

524
00:22:37.355 --> 00:22:40.115
<00:22:37.355>for <00:22:37.635>text <00:22:37.915>embedding <00:22:38.195>three <00:22:38.475>small <00:22:38.755>3072

525
00:22:40.115 --> 00:22:41.635
<00:22:40.115>dimensions <00:22:40.457>for <00:22:40.799>their <00:22:41.141>large <00:22:41.483>one.

526
00:22:42.015 --> 00:22:44.595
<00:22:42.015>Uh, <00:22:42.170>but <00:22:42.326>those <00:22:42.481>models, <00:22:42.637>yeah, <00:22:42.792>tend <00:22:42.948>to <00:22:43.103>include <00:22:43.259>hundreds <00:22:43.415>up

527
00:22:44.595 --> 00:22:47.355
<00:22:44.595>to <00:22:44.828>a <00:22:45.061>few <00:22:45.295>thousand <00:22:45.528>of <00:22:45.761>dimensions, <00:22:45.995>um,

528
00:22:47.615 --> 00:22:50.315
<00:22:47.615>but <00:22:47.790>densely <00:22:47.965>packed <00:22:48.140>with <00:22:48.315>a <00:22:48.490>lot <00:22:48.665>more <00:22:48.840>meaning, <00:22:49.015>uh,

529
00:22:50.315 --> 00:22:52.235
<00:22:50.315>than <00:22:50.490>we <00:22:50.665>had <00:22:50.840>in <00:22:51.015>our <00:22:51.190>bag <00:22:51.365>of <00:22:51.540>words <00:22:51.715>model.

530
00:22:53.975 --> 00:22:57.515
<00:22:53.975>And <00:22:54.130>then <00:22:54.286>vector <00:22:54.441>databases, <00:22:54.597>uh, <00:22:54.752>allow <00:22:54.908>you <00:22:55.063>to, <00:22:55.219>uh, <00:22:55.375>capture

531
00:22:58.085 --> 00:23:02.075
<00:22:58.085>those <00:22:58.365>vectors, <00:22:58.645>um, <00:22:58.925>and <00:22:59.205>store <00:22:59.485>them

532
00:23:02.075 --> 00:23:04.435
<00:23:02.075>and <00:23:02.215>search <00:23:02.355>against <00:23:02.495>them <00:23:02.635>in <00:23:02.775>a <00:23:02.915>much <00:23:03.055>more <00:23:03.195>efficient <00:23:03.335>manner <00:23:03.475>than

533
00:23:04.435 --> 00:23:05.595
<00:23:04.435>just <00:23:04.956>running, <00:23:05.479>uh,

534
00:23:05.675 --> 00:23:08.115
<00:23:05.675>a <00:23:05.850>similarity <00:23:06.024>against <00:23:06.200>things <00:23:06.375>in <00:23:06.550>JavaScript <00:23:06.725>like <00:23:06.899>I've <00:23:07.075>been

535
00:23:08.115 --> 00:23:10.035
<00:23:08.115>doing <00:23:08.465>terrible <00:23:08.815>idea <00:23:09.165>for <00:23:09.515>production.

536
00:23:10.455 --> 00:23:13.835
<00:23:10.455>So <00:23:10.595>vector <00:23:10.735>databases <00:23:10.875>are <00:23:11.014>gonna <00:23:11.155>do <00:23:11.294>that <00:23:11.435>via, <00:23:11.574>um, <00:23:11.715>indexing, <00:23:11.855>uh,

537
00:23:13.875 --> 00:23:18.275
<00:23:13.875>a <00:23:14.030>vector, <00:23:14.186>uh, <00:23:14.341>to <00:23:14.497>return <00:23:14.652>you <00:23:14.808>the <00:23:14.963>nearest <00:23:15.119>neighbors, <00:23:15.275>uh,

538
00:23:18.335 --> 00:23:20.715
<00:23:18.335>the <00:23:18.510>nearest <00:23:18.685>most <00:23:18.860>similar <00:23:19.035>things <00:23:19.210>based <00:23:19.385>on <00:23:19.560>either <00:23:19.735>co-sign

539
00:23:21.205 --> 00:23:24.835
<00:23:21.205>similarity, <00:23:21.345>uh, <00:23:21.485>or <00:23:21.625>you <00:23:21.764>can <00:23:21.905>do, <00:23:22.044>uh, <00:23:22.185>other <00:23:22.324>kind <00:23:22.465>of <00:23:22.605>things.

540
00:23:25.695 --> 00:23:27.355
<00:23:25.695>Um, <00:23:25.975>at <00:23:26.254>DataStax <00:23:26.534>we <00:23:26.814>have, <00:23:27.095>uh,

541
00:23:27.355 --> 00:23:28.835
<00:23:27.355>Astro <00:23:27.577>db, <00:23:27.799>which <00:23:28.021>is <00:23:28.243>our <00:23:28.465>vector <00:23:28.687>store.

542
00:23:29.055 --> 00:23:30.995
<00:23:29.055>Uh, <00:23:29.288>it <00:23:29.521>actually <00:23:29.755>can <00:23:29.988>handle <00:23:30.221>doing <00:23:30.455>the

543
00:23:30.995 --> 00:23:32.355
<00:23:30.995>vectorization <00:23:31.301>of <00:23:31.607>data <00:23:31.913>for <00:23:32.219>you.

544
00:23:32.455 --> 00:23:35.835
<00:23:32.455>We, <00:23:32.582>uh, <00:23:32.709>have <00:23:32.836>a <00:23:32.964>bunch <00:23:33.091>of, <00:23:33.218>um, <00:23:33.345>providers <00:23:33.473>that <00:23:33.600>you <00:23:33.727>can, <00:23:33.855>uh,

545
00:23:35.905 --> 00:23:37.155
<00:23:35.905>hook <00:23:36.186>up <00:23:36.467>to <00:23:36.748>your <00:23:37.030>database.

546
00:23:37.695 --> 00:23:38.795
<00:23:37.695>And <00:23:37.942>so <00:23:38.190>when <00:23:38.437>you <00:23:38.685>ingest,

547
00:23:38.815 --> 00:23:41.795
<00:23:38.815>or <00:23:38.922>in <00:23:39.030>this <00:23:39.138>kind <00:23:39.245>of <00:23:39.353>line <00:23:39.461>of <00:23:39.568>code <00:23:39.676>here <00:23:39.784>where <00:23:39.891>you <00:23:39.999>do <00:23:40.107>the, <00:23:40.215>uh,

548
00:23:42.255 --> 00:23:43.755
<00:23:42.255>uh, <00:23:42.525>take <00:23:42.795>a <00:23:43.065>collection <00:23:43.335>and <00:23:43.605>try

549
00:23:43.755 --> 00:23:46.805
<00:23:43.755>and <00:23:43.871>find <00:23:43.988>against <00:23:44.105>it, <00:23:44.221>sorting <00:23:44.338>by <00:23:44.455>the <00:23:44.571>query, <00:23:44.688>if <00:23:44.805>you <00:23:44.921>pass <00:23:45.038>it <00:23:45.155>in

550
00:23:46.805 --> 00:23:48.245
<00:23:46.805>as <00:23:47.237>dollar <00:23:47.668>vectorize, <00:23:48.101>uh,

551
00:23:48.245 --> 00:23:50.805
<00:23:48.245>that's <00:23:48.445>gonna <00:23:48.645>get <00:23:48.844>automatically <00:23:49.044>vectorized <00:23:49.245>for <00:23:49.445>you, <00:23:49.645>uh,

552
00:23:50.825 --> 00:23:53.725
<00:23:50.825>and <00:23:51.000>return, <00:23:51.175>uh, <00:23:51.350>the <00:23:51.525>results <00:23:51.700>of <00:23:51.875>your <00:23:52.050>similarity <00:23:52.225>search.

553
00:23:54.625 --> 00:23:59.085
<00:23:54.625>And <00:23:54.752>so <00:23:54.879>what <00:23:55.006>I <00:23:55.134>can <00:23:55.261>do <00:23:55.388>is <00:23:55.515>now <00:23:55.643>jump <00:23:55.770>out <00:23:55.897>of, <00:23:56.025>uh,

554
00:23:59.105 --> 00:24:01.445
<00:23:59.105>our <00:23:59.338>existing, <00:23:59.571>uh, <00:23:59.805>bag <00:24:00.038>of <00:24:00.271>words <00:24:00.505>thing.

555
00:24:01.445 --> 00:24:04.485
<00:24:01.445>I've <00:24:01.600>got <00:24:01.756>a <00:24:01.911>vector <00:24:02.067>search, <00:24:02.222>uh, <00:24:02.378>version <00:24:02.533>of <00:24:02.689>this <00:24:02.845>instead.

556
00:24:04.905 --> 00:24:07.325
<00:24:04.905>Uh, <00:24:05.080>which, <00:24:05.254>um, <00:24:05.430>so <00:24:05.605>actually <00:24:05.780>I <00:24:05.955>already <00:24:06.129>have, <00:24:06.305>oops,

557
00:24:08.045 --> 00:24:10.725
<00:24:08.045>I <00:24:08.200>have <00:24:08.356>everything <00:24:08.511>in <00:24:08.667>my <00:24:08.822>Astro <00:24:08.978>DB <00:24:09.133>database <00:24:09.289>right <00:24:09.445>here.

558
00:24:11.225 --> 00:24:13.645
<00:24:11.225>Uh, <00:24:11.380>you <00:24:11.536>can <00:24:11.691>see <00:24:11.847>all <00:24:12.002>the <00:24:12.158>talks, <00:24:12.313>uh, <00:24:12.469>we've <00:24:12.625>got,

559
00:24:13.645 --> 00:24:14.805
<00:24:13.645>they're <00:24:13.853>in <00:24:14.062>a <00:24:14.271>different <00:24:14.480>order <00:24:14.689>here,

560
00:24:14.825 --> 00:24:19.005
<00:24:14.825>but <00:24:14.965>if <00:24:15.105>I <00:24:15.245>ask <00:24:15.385>my <00:24:15.525>database <00:24:15.665>here, <00:24:15.805>uh, <00:24:15.945>who <00:24:16.085>is <00:24:16.225>talking

561
00:24:19.735 --> 00:24:23.285
<00:24:19.735>about <00:24:20.201>retrieval <00:24:20.668>augmented <00:24:21.135>generation,

562
00:24:24.935 --> 00:24:27.745
<00:24:24.935>this <00:24:25.090>is <00:24:25.246>gonna <00:24:25.401>vectorize <00:24:25.557>that <00:24:25.712>live <00:24:25.868>in <00:24:26.023>the, <00:24:26.179>uh, <00:24:26.335>um,

563
00:24:28.925 --> 00:24:30.105
<00:24:28.925>uh, <00:24:29.190>in <00:24:29.456>the <00:24:29.721>dashboard <00:24:29.987>here.

564
00:24:30.445 --> 00:24:32.265
<00:24:30.445>And <00:24:30.725>then, <00:24:31.004>uh, <00:24:31.284>there <00:24:31.564>we <00:24:31.845>go.

565
00:24:32.265 --> 00:24:33.745
<00:24:32.265>We <00:24:32.487>get <00:24:32.709>the <00:24:32.931>sort <00:24:33.153>of <00:24:33.375>answers <00:24:33.597>here,

566
00:24:33.745 --> 00:24:35.385
<00:24:33.745>exploring <00:24:34.094>advanced <00:24:34.445>patterns <00:24:34.794>in <00:24:35.145>retrieval,

567
00:24:35.385 --> 00:24:36.705
<00:24:35.385>augmented <00:24:35.622>generation <00:24:35.860>build <00:24:36.097>drag <00:24:36.335>from <00:24:36.572>scratch.

568
00:24:36.705 --> 00:24:38.145
<00:24:36.705>Large <00:24:36.964>language <00:24:37.223>models, <00:24:37.482>love <00:24:37.741>knowledge <00:24:38.001>graphs.

569
00:24:38.255 --> 00:24:39.745
<00:24:38.255>They're <00:24:38.478>the <00:24:38.702>ones <00:24:38.925>we <00:24:39.149>expected <00:24:39.372>to <00:24:39.596>see.

570
00:24:40.605 --> 00:24:44.185
<00:24:40.605>Uh, <00:24:40.745>and <00:24:40.885>so <00:24:41.025>if <00:24:41.165>we, <00:24:41.305>oops, <00:24:41.445>if <00:24:41.585>we <00:24:41.725>swap <00:24:41.865>out, <00:24:42.005>um,

571
00:24:44.625 --> 00:24:46.865
<00:24:44.625>I <00:24:44.800>have <00:24:44.975>the <00:24:45.150>same <00:24:45.325>find <00:24:45.500>similar <00:24:45.675>torque <00:24:45.850>function <00:24:46.025>here.

572
00:24:47.165 --> 00:24:48.385
<00:24:47.165>If <00:24:47.348>we <00:24:47.531>head <00:24:47.714>back <00:24:47.897>to <00:24:48.080>our <00:24:48.263>server

573
00:24:48.805 --> 00:24:53.385
<00:24:48.805>and, <00:24:48.960>uh, <00:24:49.116>just <00:24:49.271>swap <00:24:49.427>out <00:24:49.582>the <00:24:49.738>implementation, <00:24:49.893>uh, <00:24:50.049>our, <00:24:50.205>uh,

574
00:24:53.385 --> 00:24:54.905
<00:24:53.385>search <00:24:53.613>should <00:24:53.841>actually <00:24:54.069>go <00:24:54.297>and <00:24:54.525>get <00:24:54.753>better.

575
00:24:56.035 --> 00:24:58.745
<00:24:56.035>Let's <00:24:56.385>go <00:24:56.735>to <00:24:57.085>Firefox. <00:24:57.435>Uh,

576
00:24:58.745 --> 00:25:02.545
<00:24:58.745>and <00:24:58.872>so <00:24:58.999>when <00:24:59.126>we <00:24:59.254>looked <00:24:59.381>at <00:24:59.508>the <00:24:59.635>original <00:24:59.763>search, <00:24:59.890>um, <00:25:00.017>we <00:25:00.145>saw

577
00:25:02.545 --> 00:25:04.105
<00:25:02.545>that <00:25:03.011>asking <00:25:03.478>that, <00:25:03.945>uh,

578
00:25:04.195 --> 00:25:06.465
<00:25:04.195>who's <00:25:04.428>talking <00:25:04.661>about <00:25:04.895>retrieval <00:25:05.128>augmented <00:25:05.361>generation, <00:25:05.595>uh,

579
00:25:06.485 --> 00:25:08.305
<00:25:06.485>is <00:25:06.718>about <00:25:06.951>2.88 <00:25:07.185>was <00:25:07.418>the <00:25:07.651>most <00:25:07.885>similar,

580
00:25:08.805 --> 00:25:13.175
<00:25:08.805>if <00:25:09.005>I <00:25:09.205>ask <00:25:09.405>the <00:25:09.605>same <00:25:09.805>thing <00:25:10.005>Now, <00:25:10.205>uh,

581
00:25:13.195 --> 00:25:16.495
<00:25:13.195>the <00:25:13.370>similarity <00:25:13.544>has <00:25:13.720>gone <00:25:13.895>up <00:25:14.070>to <00:25:14.245>0.779, <00:25:14.419>uh, <00:25:14.595>with,

582
00:25:16.605 --> 00:25:19.455
<00:25:16.605>with <00:25:16.780>this <00:25:16.955>talk, <00:25:17.130>uh, <00:25:17.305>my <00:25:17.480>talk <00:25:17.655>comes <00:25:17.830>out <00:25:18.005>0.77,

583
00:25:19.875 --> 00:25:22.015
<00:25:19.875>and <00:25:20.075>we <00:25:20.275>also <00:25:20.475>get <00:25:20.675>slightly <00:25:20.875>different, <00:25:21.075>uh, <00:25:21.275>results.

584
00:25:22.395 --> 00:25:24.615
<00:25:22.395>Uh, <00:25:22.595>this <00:25:22.795>was <00:25:22.995>actually <00:25:23.195>a <00:25:23.395>talk <00:25:23.595>about <00:25:23.795>blockchain

585
00:25:24.615 --> 00:25:25.855
<00:25:24.615>and <00:25:24.987>retrieval <00:25:25.359>augmented <00:25:25.731>generation

586
00:25:25.855 --> 00:25:27.095
<00:25:25.855>somehow <00:25:26.134>in <00:25:26.413>the <00:25:26.692>previous <00:25:26.971>results.

587
00:25:27.315 --> 00:25:29.495
<00:25:27.315>Uh, <00:25:27.470>but <00:25:27.626>now <00:25:27.781>it's <00:25:27.937>about <00:25:28.092>building <00:25:28.248>things <00:25:28.403>with <00:25:28.559>spring <00:25:28.715>and,

588
00:25:29.495 --> 00:25:30.655
<00:25:29.495>and <00:25:29.842>AI <00:25:30.191>much <00:25:30.539>nicer.

589
00:25:31.275 --> 00:25:33.495
<00:25:31.275>Uh, <00:25:31.475>and, <00:25:31.675>um, <00:25:31.875>and <00:25:32.075>yeah, <00:25:32.275>final <00:25:32.475>one <00:25:32.675>there.

590
00:25:33.755 --> 00:25:35.295
<00:25:33.755>And <00:25:33.909>so <00:25:34.063>we <00:25:34.217>can <00:25:34.371>take <00:25:34.525>that <00:25:34.679>back <00:25:34.833>to <00:25:34.987>our <00:25:35.141>chat.

591
00:25:35.915 --> 00:25:37.695
<00:25:35.915>Um, <00:25:36.615>ask <00:25:37.315>it,

592
00:25:37.825 --> 00:25:39.575
<00:25:37.825>who's <00:25:38.105>talking <00:25:38.385>about <00:25:38.665>retrieval <00:25:38.945>augmented <00:25:39.225>generation,

593
00:25:39.595 --> 00:25:42.735
<00:25:39.595>and <00:25:39.750>now <00:25:39.906>using, <00:25:40.061>uh, <00:25:40.217>a <00:25:40.372>full <00:25:40.528>on, <00:25:40.683>uh, <00:25:40.839>embedding <00:25:40.995>model,

594
00:25:43.055 --> 00:25:45.055
<00:25:43.055>a <00:25:43.288>vex <00:25:43.521>database <00:25:43.755>in <00:25:43.988>the <00:25:44.221>background, <00:25:44.455>uh,

595
00:25:45.155 --> 00:25:47.695
<00:25:45.155>and, <00:25:45.355>uh, <00:25:45.555>Gemini <00:25:45.754>to, <00:25:45.955>uh, <00:25:46.155>to <00:25:46.355>generate <00:25:46.555>things.

596
00:25:48.035 --> 00:25:49.815
<00:25:48.035>Uh, <00:25:48.210>we <00:25:48.385>get <00:25:48.560>this <00:25:48.735>kind <00:25:48.910>of <00:25:49.085>list <00:25:49.260>of <00:25:49.435>people,

597
00:25:50.275 --> 00:25:52.815
<00:25:50.275>and <00:25:50.430>I <00:25:50.586>can <00:25:50.741>also <00:25:50.897>ask <00:25:51.052>it, <00:25:51.208>um, <00:25:51.363>who <00:25:51.519>I <00:25:51.675>am.

598
00:25:52.815 --> 00:25:54.215
<00:25:52.815>And <00:25:52.995>I <00:25:53.175>think <00:25:53.355>it <00:25:53.535>might <00:25:53.715>know <00:25:53.895>better <00:25:54.074>now

599
00:25:54.565 --> 00:25:57.295
<00:25:54.565>that <00:25:54.740>at <00:25:54.915>least <00:25:55.090>I'm <00:25:55.265>talking <00:25:55.440>about <00:25:55.615>something, <00:25:55.790>oh, <00:25:55.965>okay,

600
00:25:57.315 --> 00:25:58.335
<00:25:57.315>it <00:25:57.498>still <00:25:57.682>doesn't <00:25:57.865>know <00:25:58.049>about <00:25:58.233>me.

601
00:25:58.355 --> 00:26:02.255
<00:25:58.355>That's <00:25:58.510>fine. <00:25:58.666>I <00:25:58.821>never <00:25:58.977>expected <00:25:59.132>Gemini <00:25:59.288>to <00:25:59.443>be <00:25:59.599>so <00:25:59.755>cruel.

602
00:26:04.515 --> 00:26:06.295
<00:26:04.515>So <00:26:04.795>that's <00:26:05.075>sort <00:26:05.355>of <00:26:05.635>the <00:26:05.915>basics

603
00:26:06.295 --> 00:26:07.535
<00:26:06.295>of <00:26:06.667>retrieval, <00:26:07.039>augmented <00:26:07.411>generation.

604
00:26:07.635 --> 00:26:09.495
<00:26:07.635>We <00:26:07.835>shouldn't <00:26:08.035>be <00:26:08.235>building <00:26:08.435>our <00:26:08.635>own <00:26:08.835>embedding <00:26:09.035>models

605
00:26:09.495 --> 00:26:12.455
<00:26:09.495>unless <00:26:09.650>you <00:26:09.806>are, <00:26:09.961>uh, <00:26:10.117>also, <00:26:10.272>uh, <00:26:10.428>doing <00:26:10.583>the <00:26:10.739>machine <00:26:10.895>learning,

606
00:26:12.455 --> 00:26:14.535
<00:26:12.455>doing <00:26:12.688>the <00:26:12.921>kind <00:26:13.155>of <00:26:13.388>building <00:26:13.621>models <00:26:13.855>yourself.

607
00:26:14.835 --> 00:26:17.295
<00:26:14.835>Uh, <00:26:14.975>don't <00:26:15.115>use <00:26:15.255>a <00:26:15.395>bag <00:26:15.535>of <00:26:15.675>words, <00:26:15.815>it's <00:26:15.955>not <00:26:16.095>that <00:26:16.235>useful.

608
00:26:18.035 --> 00:26:21.095
<00:26:18.035>Um, <00:26:18.175>but <00:26:18.315>there's <00:26:18.455>more <00:26:18.595>we <00:26:18.735>can <00:26:18.875>do <00:26:19.015>with <00:26:19.155>this, <00:26:19.295>like <00:26:19.435>that,

609
00:26:21.125 --> 00:26:23.895
<00:26:21.125>that <00:26:21.300>very <00:26:21.475>simple <00:26:21.650>version <00:26:21.825>of <00:26:22.000>that, <00:26:22.175>uh, <00:26:22.350>uses <00:26:22.525>kind

610
00:26:23.895 --> 00:26:26.095
<00:26:23.895>of <00:26:24.128>these <00:26:24.361>generic <00:26:24.595>models, <00:26:24.828>uh, <00:26:25.061>and, <00:26:25.295>um,

611
00:26:26.915 --> 00:26:28.895
<00:26:26.915>and <00:26:27.115>only <00:26:27.315>kind <00:26:27.514>of <00:26:27.715>does <00:26:27.915>that <00:26:28.115>similarity <00:26:28.315>search.

612
00:26:29.555 --> 00:26:32.015
<00:26:29.555>Uh, <00:26:29.788>I've <00:26:30.021>written, <00:26:30.255>uh, <00:26:30.488>I <00:26:30.721>recently, <00:26:30.955>uh,

613
00:26:32.015 --> 00:26:33.615
<00:26:32.015>published <00:26:32.365>this <00:26:32.715>highly <00:26:33.065>accurate <00:26:33.415>retrieval

614
00:26:33.615 --> 00:26:35.015
<00:26:33.615>for <00:26:33.867>your <00:26:34.119>rag <00:26:34.371>application <00:26:34.623>with <00:26:34.875>Colbert

615
00:26:35.015 --> 00:26:38.575
<00:26:35.015>and <00:26:35.142>Astro <00:26:35.269>DB <00:26:35.396>Colbert <00:26:35.524>is <00:26:35.651>a <00:26:35.778>different <00:26:35.905>way <00:26:36.033>to, <00:26:36.160>uh, <00:26:36.287>both <00:26:36.415>store

616
00:26:38.635 --> 00:26:39.855
<00:26:38.635>and <00:26:39.001>retrieve, <00:26:39.367>uh, <00:26:39.733>things

617
00:26:39.855 --> 00:26:42.855
<00:26:39.855>by <00:26:39.995>similarity <00:26:40.135>in <00:26:40.275>which <00:26:40.415>you <00:26:40.555>actually <00:26:40.695>make <00:26:40.835>every <00:26:40.975>token <00:26:41.115>in <00:26:41.255>a

618
00:26:42.855 --> 00:26:46.125
<00:26:42.855>string <00:26:43.088>a <00:26:43.321>vector, <00:26:43.555>which, <00:26:43.788>um, <00:26:44.021>sounds <00:26:44.255>mad,

619
00:26:46.125 --> 00:26:47.765
<00:26:46.125>but <00:26:46.325>it's <00:26:46.525>totally <00:26:46.725>possible, <00:26:46.925>and <00:26:47.125>it <00:26:47.325>makes <00:26:47.525>it

620
00:26:47.765 --> 00:26:49.125
<00:26:47.765>everything <00:26:48.009>a <00:26:48.254>little <00:26:48.499>bit <00:26:48.744>more <00:26:48.989>accurate.

621
00:26:49.905 --> 00:26:52.325
<00:26:49.905>Um, <00:26:50.060>and <00:26:50.216>we've <00:26:50.371>also <00:26:50.527>started <00:26:50.682>writing <00:26:50.838>at <00:26:50.993>datas <00:26:51.149>stacks <00:26:51.305>about,

622
00:26:52.345 --> 00:26:55.685
<00:26:52.345>uh, <00:26:52.520>how, <00:26:52.695>uh, <00:26:52.870>content-centric <00:26:53.045>knowledge <00:26:53.220>graphs <00:26:53.395>can <00:26:53.570>also <00:26:53.745>help,

623
00:26:56.025 --> 00:26:58.925
<00:26:56.025>uh, <00:26:56.225>get <00:26:56.425>better <00:26:56.625>context <00:26:56.825>for <00:26:57.025>LLMs, <00:26:57.225>uh, <00:26:57.425>for

624
00:26:58.925 --> 00:27:00.325
<00:26:58.925>that <00:26:59.240>augmentation <00:26:59.555>for <00:26:59.870>that <00:27:00.185>generation.

625
00:27:01.385 --> 00:27:02.445
<00:27:01.385>And <00:27:01.623>finally, <00:27:01.862>I <00:27:02.100>just <00:27:02.339>wanted

626
00:27:02.445 --> 00:27:04.205
<00:27:02.445>to <00:27:02.620>share <00:27:02.794>other <00:27:02.970>things <00:27:03.145>you <00:27:03.320>could <00:27:03.495>do <00:27:03.669>with, <00:27:03.845>uh,

627
00:27:04.205 --> 00:27:05.845
<00:27:04.205>vector <00:27:04.671>databases, <00:27:05.138>uh, <00:27:05.605>or,

628
00:27:05.845 --> 00:27:09.045
<00:27:05.845>or <00:27:05.985>vector <00:27:06.125>kind <00:27:06.265>of <00:27:06.405>similarity, <00:27:06.545>uh, <00:27:06.685>in <00:27:06.825>which <00:27:06.965>I, <00:27:07.105>I, <00:27:07.245>uh,

629
00:27:09.045 --> 00:27:11.925
<00:27:09.045>created <00:27:09.185>a, <00:27:09.325>uh, <00:27:09.465>feature <00:27:09.605>for <00:27:09.745>my <00:27:09.885>own <00:27:10.025>blog <00:27:10.165>in <00:27:10.305>which <00:27:10.445>I,

630
00:27:12.085 --> 00:27:14.925
<00:27:12.085>I <00:27:12.318>could <00:27:12.551>show, <00:27:12.785>um, <00:27:13.018>related <00:27:13.251>blog <00:27:13.485>posts,

631
00:27:15.065 --> 00:27:16.205
<00:27:15.065>uh, <00:27:15.321>underneath <00:27:15.578>a <00:27:15.834>blog <00:27:16.091>post.

632
00:27:16.565 --> 00:27:17.725
<00:27:16.565>'cause <00:27:16.773>that <00:27:16.982>is <00:27:17.191>the <00:27:17.400>absolute <00:27:17.609>bread

633
00:27:17.725 --> 00:27:18.725
<00:27:17.725>and <00:27:17.949>butter <00:27:18.175>for <00:27:18.399>these <00:27:18.625>vectors.

634
00:27:18.945 --> 00:27:22.405
<00:27:18.945>If <00:27:19.072>you <00:27:19.199>have <00:27:19.326>vectorized, <00:27:19.454>uh, <00:27:19.581>uh, <00:27:19.708>the <00:27:19.835>content <00:27:19.963>of <00:27:20.090>a <00:27:20.217>post, <00:27:20.345>uh,

635
00:27:22.405 --> 00:27:24.245
<00:27:22.405>then <00:27:22.754>a <00:27:23.105>similarity <00:27:23.455>score, <00:27:23.805>uh,

636
00:27:24.245 --> 00:27:26.725
<00:27:24.245>or <00:27:24.478>a <00:27:24.711>similarity <00:27:24.945>sort <00:27:25.178>brings <00:27:25.411>back <00:27:25.645>definitely

637
00:27:26.725 --> 00:27:27.805
<00:27:26.725>things <00:27:26.967>that <00:27:27.211>are, <00:27:27.454>are <00:27:27.697>related.

638
00:27:28.865 --> 00:27:31.365
<00:27:28.865>Uh, <00:27:28.992>I <00:27:29.119>am <00:27:29.246>going <00:27:29.374>to <00:27:29.501>go <00:27:29.628>to <00:27:29.755>the <00:27:29.883>last <00:27:30.010>thing <00:27:30.137>that, <00:27:30.265>uh,

639
00:27:31.545 --> 00:27:33.245
<00:27:31.545>QR <00:27:31.720>code <00:27:31.895>there <00:27:32.070>will <00:27:32.245>take <00:27:32.420>you <00:27:32.595>to <00:27:32.770>the <00:27:32.945>slides

640
00:27:33.245 --> 00:27:34.245
<00:27:33.245>and <00:27:33.395>all <00:27:33.544>the <00:27:33.695>links <00:27:33.844>from <00:27:33.995>the <00:27:34.145>talk.

641
00:27:34.745 --> 00:27:37.765
<00:27:34.745>Uh, <00:27:34.872>and <00:27:34.999>so, <00:27:35.126>um, <00:27:35.254>hopefully <00:27:35.381>you <00:27:35.508>can, <00:27:35.635>uh, <00:27:35.763>check <00:27:35.890>that <00:27:36.017>out <00:27:36.145>there.

642
00:27:38.135 --> 00:27:40.045
<00:27:38.135>Other <00:27:38.335>than <00:27:38.535>that, <00:27:38.735>uh, <00:27:38.935>my <00:27:39.135>name's <00:27:39.335>Von <00:27:39.535>Ash.

643
00:27:40.045 --> 00:27:42.085
<00:27:40.045>Uh, <00:27:40.245>I'm <00:27:40.445>a <00:27:40.645>developer <00:27:40.845>relations <00:27:41.045>engineer <00:27:41.245>at <00:27:41.445>DataStax.

644
00:27:42.385 --> 00:27:45.085
<00:27:42.385>Uh, <00:27:42.540>we'd <00:27:42.696>love <00:27:42.851>to <00:27:43.007>help <00:27:43.162>you <00:27:43.318>build, <00:27:43.473>uh, <00:27:43.629>very <00:27:43.785>awesome,

645
00:27:45.155 --> 00:27:47.765
<00:27:45.155>very <00:27:45.355>cool, <00:27:45.555>uh, <00:27:45.754>generative <00:27:45.955>AI <00:27:46.155>things <00:27:46.355>using <00:27:46.555>retrieval,

646
00:27:47.765 --> 00:27:49.045
<00:27:47.765>augmented <00:27:48.053>generation <00:27:48.341>and <00:27:48.629>other <00:27:48.917>stuff.

647
00:27:49.705 --> 00:27:52.125
<00:27:49.705>Uh, <00:27:49.832>so <00:27:49.959>do <00:27:50.086>please <00:27:50.214>come <00:27:50.341>and, <00:27:50.468>uh, <00:27:50.595>come <00:27:50.723>and <00:27:50.850>talk <00:27:50.977>to <00:27:51.105>us.

648
00:27:52.155 --> 00:27:53.885
<00:27:52.155>Come <00:27:52.282>and <00:27:52.409>find <00:27:52.536>me. <00:27:52.664>Come <00:27:52.791>and <00:27:52.918>find <00:27:53.045>the <00:27:53.173>rest <00:27:53.300>of <00:27:53.427>the <00:27:53.555>team.

649
00:27:53.885 --> 00:27:56.405
<00:27:53.885>We're <00:27:54.040>up <00:27:54.196>in <00:27:54.351>the <00:27:54.507>other <00:27:54.662>hall, <00:27:54.818>uh, <00:27:54.973>the <00:27:55.129>DataStax <00:27:55.285>booth.

650
00:27:56.705 --> 00:27:58.445
<00:27:56.705>Uh, <00:27:56.938>and <00:27:57.171>otherwise, <00:27:57.405>thank <00:27:57.638>you <00:27:57.871>very <00:27:58.105>much

651
00:27:58.445 --> 00:28:00.605
<00:27:58.445>and <00:27:58.600>enjoy <00:27:58.756>the <00:27:58.911>rest <00:27:59.067>of <00:27:59.222>the, <00:27:59.378>uh, <00:27:59.533>of <00:27:59.689>the <00:27:59.845>conference.

652
00:28:00.605 --> 00:28:00.925
<00:28:00.605>Thank <00:28:00.893>you.