1
00:00:07,080 --> 00:00:11,040
Speaker 1: Alexa, what's the best science podcast on air? Hey? Are

2
00:00:11,039 --> 00:00:13,560
Speaker 1: you trying to replace me with Alexa? What's going on here?

3
00:00:14,120 --> 00:00:17,720
Speaker 1: Do you think you're replaceable? There's no way an artificial

4
00:00:17,760 --> 00:00:21,560
Speaker 1: intelligence could ever make jokes nearly as funny as I am.

5
00:00:21,600 --> 00:00:24,400
Speaker 1: I think there's no way an artificial intelligence would laugh

6
00:00:24,440 --> 00:00:27,280
Speaker 1: at your jokes. I'm pretty sure I could. I could

7
00:00:27,320 --> 00:00:30,200
Speaker 1: program a pretty dumb computer to laugh at my jokes.

8
00:00:30,240 --> 00:00:34,280
Speaker 1: It's called the laugh track. But that's Hey, that's a

9
00:00:34,280 --> 00:00:38,000
Speaker 1: new challenge for for AI. You know, first chest then

10
00:00:38,080 --> 00:00:43,760
Speaker 1: go now science comedy. That's right, now, programs something that

11
00:00:43,800 --> 00:01:05,560
Speaker 1: can find humor in Daniel's ramblings him and I'm Daniel,

12
00:01:05,680 --> 00:01:09,040
Speaker 1: and this is our podcast. Daniel and Jorge explained the Universe,

13
00:01:09,200 --> 00:01:12,080
Speaker 1: in which we try to download everything we know about

14
00:01:12,120 --> 00:01:15,959
Speaker 1: the universe, episode by episode into your brain, whether you're

15
00:01:16,000 --> 00:01:19,840
Speaker 1: a real person or an artificial intelligence, listening to our

16
00:01:19,920 --> 00:01:24,319
Speaker 1: podcast while trying to sound intelligent about it while writing

17
00:01:24,319 --> 00:01:28,280
Speaker 1: your own humor for the open mic AI Night. The

18
00:01:28,319 --> 00:01:35,800
Speaker 1: topic of today's podcast is what is artificial intelligence? And

19
00:01:36,680 --> 00:01:39,880
Speaker 1: very importantly is it dangerous? That's right? Should you be

20
00:01:40,440 --> 00:01:42,880
Speaker 1: looking at your window for the first signs of the

21
00:01:42,959 --> 00:01:46,320
Speaker 1: robot revolution. Should you be afraid of your Alexa? Should

22
00:01:46,360 --> 00:01:50,240
Speaker 1: you be worried about that robot vacuum cleaner getting resentful

23
00:01:50,360 --> 00:01:52,600
Speaker 1: for having to do all the dirty work and eating

24
00:01:52,600 --> 00:01:56,640
Speaker 1: your face off in the middle of the night. That's

25
00:01:56,680 --> 00:01:59,240
Speaker 1: a bit dark. It seems kind of sinister, doesn't it.

26
00:01:59,240 --> 00:02:03,280
Speaker 1: It's like sitting there are circling, circling, circling, waning, waiting, waiting.

27
00:02:03,560 --> 00:02:06,600
Speaker 1: I think those things are creepy, right. Maybe it wants to,

28
00:02:07,160 --> 00:02:10,000
Speaker 1: you know, clean your face, it wants see That's the question.

29
00:02:10,360 --> 00:02:13,680
Speaker 1: Does a robot vacuum cleaner want anything? What does it

30
00:02:13,720 --> 00:02:15,720
Speaker 1: mean for it to want? What is it like to

31
00:02:15,760 --> 00:02:19,320
Speaker 1: be a robot vacuum cleaner? The next great paper in philosophy?

32
00:02:19,480 --> 00:02:21,720
Speaker 1: So this is kind of in the zeitgeist right now.

33
00:02:21,760 --> 00:02:24,880
Speaker 1: I mean, people are really excited about artificial intelligence. But

34
00:02:24,880 --> 00:02:27,280
Speaker 1: at the same time there are big names like Elon

35
00:02:27,360 --> 00:02:31,520
Speaker 1: Musk kind of warning people like, hey, artificial intelligence not

36
00:02:31,600 --> 00:02:33,680
Speaker 1: such a good idea. That's right, it's a huge topic.

37
00:02:33,720 --> 00:02:35,600
Speaker 1: I mean, you drive around like San Francisco, you see

38
00:02:35,680 --> 00:02:39,160
Speaker 1: artificial intelligence, machine learning, deep learning. It's on billboards. Even

39
00:02:39,200 --> 00:02:41,000
Speaker 1: you know you want to get a million bucks for

40
00:02:41,040 --> 00:02:43,880
Speaker 1: your new company, you just say the words AI, deep learning,

41
00:02:43,880 --> 00:02:46,880
Speaker 1: and boom people are throwing cash to right, people are

42
00:02:46,919 --> 00:02:51,120
Speaker 1: learning in the deepest learning. Um. It's definitely part of

43
00:02:51,120 --> 00:02:53,920
Speaker 1: the cultural moment, and you see that reflected not just

44
00:02:53,960 --> 00:02:55,640
Speaker 1: in like what deep thinkers are saying, but also in

45
00:02:55,639 --> 00:02:58,040
Speaker 1: like science fiction. You know, a lot of the near

46
00:02:58,160 --> 00:03:01,079
Speaker 1: term dystopian these days is about how AI will take

47
00:03:01,120 --> 00:03:04,680
Speaker 1: over and the dangers of AI. Another way, like thirty

48
00:03:04,760 --> 00:03:07,919
Speaker 1: years ago is about the dangers of UH of radiation. Right,

49
00:03:07,960 --> 00:03:11,000
Speaker 1: that was the new dangerous thing physicists that invented. Now

50
00:03:11,040 --> 00:03:13,760
Speaker 1: the new dangerous technology that we're all worried about is AI.

51
00:03:13,840 --> 00:03:17,760
Speaker 1: It's a new promise in peril. Um. Yeah. AI. Every

52
00:03:17,760 --> 00:03:19,840
Speaker 1: piece of technology is a double edged sword, right. You

53
00:03:19,880 --> 00:03:21,560
Speaker 1: can use it for good, you can use for evil.

54
00:03:21,600 --> 00:03:24,919
Speaker 1: But AI is special because it's not just technology, is

55
00:03:24,960 --> 00:03:27,160
Speaker 1: not just a tool that people use. It's a tool

56
00:03:27,400 --> 00:03:30,880
Speaker 1: that that has independence, that has autonomy. And that's why

57
00:03:30,880 --> 00:03:33,560
Speaker 1: it's such a vexing question. Well people, I'm sure people

58
00:03:33,560 --> 00:03:37,160
Speaker 1: everyone associated with robots and machines and computers, but we

59
00:03:37,200 --> 00:03:40,840
Speaker 1: were kind of wondering if people actually knew what artificial

60
00:03:40,840 --> 00:03:43,280
Speaker 1: intelligence was, like, what makes it work, what makes it

61
00:03:43,320 --> 00:03:46,880
Speaker 1: different than than real intelligence? M I bet you that

62
00:03:47,760 --> 00:03:50,760
Speaker 1: the people who say the phrase artificial intelligence don't actually

63
00:03:50,760 --> 00:03:53,720
Speaker 1: know what they're talking about, which is probably true for

64
00:03:53,840 --> 00:03:58,520
Speaker 1: most topics andology, it's true for me probably, I'm sure.

65
00:03:59,240 --> 00:04:01,720
Speaker 1: But we're wondering if you guys out there knew what

66
00:04:01,880 --> 00:04:05,200
Speaker 1: artificial intelligence was. And so, as usual, Daniel went out

67
00:04:05,240 --> 00:04:07,800
Speaker 1: and as people in the street, and here's what they

68
00:04:07,880 --> 00:04:11,120
Speaker 1: had to say. Um, it's the idea that we can

69
00:04:11,120 --> 00:04:14,720
Speaker 1: create some type of material thing that could think on

70
00:04:14,760 --> 00:04:17,640
Speaker 1: its own ultimately, And do you think it's something we

71
00:04:17,640 --> 00:04:19,120
Speaker 1: should be concerned about? Is it ever going to be

72
00:04:19,120 --> 00:04:22,600
Speaker 1: a threat to humanity? I mean possibly, but I mean

73
00:04:22,680 --> 00:04:24,480
Speaker 1: we never We don't know everything. We can know the

74
00:04:24,520 --> 00:04:26,680
Speaker 1: bounds of what could be a threat, we cannot be

75
00:04:26,720 --> 00:04:30,720
Speaker 1: a threat. Yeah, it's AI, and it's the stuff that's

76
00:04:30,839 --> 00:04:34,320
Speaker 1: used in various technological applications basically just kind of like

77
00:04:35,160 --> 00:04:39,200
Speaker 1: trying to make machines replicate certain aspects of human intelligence.

78
00:04:39,560 --> 00:04:41,360
Speaker 1: Stuff like that. Okay, And do you think it could

79
00:04:41,360 --> 00:04:43,120
Speaker 1: ever be a threat to humanity? Is something we should

80
00:04:43,120 --> 00:04:45,240
Speaker 1: be worried about? I guess since I don't have a

81
00:04:45,320 --> 00:04:49,200
Speaker 1: particularly strong opinion on it, I don't think so. So

82
00:04:49,839 --> 00:04:52,599
Speaker 1: I guess I'll say no for now. Um, I'm assuming

83
00:04:52,680 --> 00:04:58,680
Speaker 1: that's the idea that computers or electronics can have like sentience. Right,

84
00:04:59,600 --> 00:05:02,160
Speaker 1: Are you worried that computers would when they take over

85
00:05:02,200 --> 00:05:05,120
Speaker 1: and make us their slaves? And not really, I don't

86
00:05:05,160 --> 00:05:07,719
Speaker 1: think it will come to that point. All right. Those

87
00:05:07,720 --> 00:05:10,720
Speaker 1: are pretty sophisticated answers. I like the ones that said,

88
00:05:10,839 --> 00:05:14,240
Speaker 1: um oh, artificial intelligence, that's just AI, right, Like that's

89
00:05:14,240 --> 00:05:18,520
Speaker 1: an answer. So that's an answer to every question. You know,

90
00:05:19,080 --> 00:05:24,640
Speaker 1: what is Google blex Zavi Brown? Oh, that's just gens. Yeah, acronyms.

91
00:05:24,760 --> 00:05:27,719
Speaker 1: Acronyms can make you look intelligent, that's it. That's the

92
00:05:27,760 --> 00:05:34,600
Speaker 1: real artificial intelligence to speaking acronym acronym intelligence. Um. You know,

93
00:05:34,760 --> 00:05:37,120
Speaker 1: but people had some sense that it's, you know, something

94
00:05:37,160 --> 00:05:40,520
Speaker 1: that can think for itself, or something do something for you,

95
00:05:40,720 --> 00:05:43,640
Speaker 1: or create something that can think by itself. There's definitely

96
00:05:43,680 --> 00:05:46,400
Speaker 1: the nuggative idea is definitely out there. They use it

97
00:05:46,600 --> 00:05:49,159
Speaker 1: in relation to what it can do. That's right, yeah,

98
00:05:49,440 --> 00:05:53,520
Speaker 1: exactly what what is what's the new capability that defines it? Yeah?

99
00:05:53,839 --> 00:05:56,040
Speaker 1: Right yeah, and that it's it's a fascinating way to

100
00:05:56,040 --> 00:05:57,840
Speaker 1: think about it, you know. And uh, it's definitely a

101
00:05:57,880 --> 00:06:00,200
Speaker 1: tricky question, right because I guess we know it in

102
00:06:00,200 --> 00:06:02,760
Speaker 1: the context of UM using them for things, right, Like,

103
00:06:02,800 --> 00:06:05,360
Speaker 1: we're people, don't just create a I because we want

104
00:06:05,360 --> 00:06:07,960
Speaker 1: to create artificial beings. It's like, so it can help us.

105
00:06:08,320 --> 00:06:11,520
Speaker 1: I want to create artificial beings. What's what's wrong with that?

106
00:06:11,520 --> 00:06:15,120
Speaker 1: That sounds pretty awesome. Create a whole army of physics

107
00:06:15,160 --> 00:06:19,400
Speaker 1: artificial physics grad students. It sounds pretty cool. Kids. Yeah,

108
00:06:19,440 --> 00:06:23,279
Speaker 1: you mean, are they worried about competing with my digital children? Natural?

109
00:06:23,920 --> 00:06:26,760
Speaker 1: They know you'd rather have artificial children. I didn't say

110
00:06:26,760 --> 00:06:30,200
Speaker 1: I'd rather have artificial that in addition to my beautiful,

111
00:06:30,200 --> 00:06:32,440
Speaker 1: wonderful natural children, which I should not be talking about

112
00:06:32,480 --> 00:06:34,599
Speaker 1: on this podcast. I'd love to have a whole, you know,

113
00:06:34,720 --> 00:06:38,880
Speaker 1: cadre of artificial children to do my bidding, unlike your

114
00:06:38,880 --> 00:06:42,320
Speaker 1: real children. If who won't do your bidding? So somebody

115
00:06:42,400 --> 00:06:47,280
Speaker 1: listening children? And that sort of goes to the heart

116
00:06:47,279 --> 00:06:49,880
Speaker 1: of the question. You know, UM, if you created a

117
00:06:49,920 --> 00:06:52,480
Speaker 1: digital being with artificial intelligence, would it listen to you?

118
00:06:52,560 --> 00:06:54,960
Speaker 1: Or would it make its own decisions? Right? And so

119
00:06:55,360 --> 00:06:57,160
Speaker 1: that's why we thought it would be interesting to dig into,

120
00:06:57,279 --> 00:06:59,640
Speaker 1: like what is artificial intelligence? If it just did what

121
00:06:59,680 --> 00:07:01,680
Speaker 1: you told that to do. It wouldn't maybe be in

122
00:07:01,720 --> 00:07:04,839
Speaker 1: an artificial intelligence you're saying nobody smart should listen to you,

123
00:07:04,880 --> 00:07:08,839
Speaker 1: is what you're saying. I'm saying they should decide for

124
00:07:08,880 --> 00:07:15,640
Speaker 1: themselves whether I'm I'm worth following. So let's break it

125
00:07:15,680 --> 00:07:19,600
Speaker 1: down for people. Daniel, what is artificial intelligence? Well you

126
00:07:19,600 --> 00:07:21,320
Speaker 1: should listen to this podcast and that will give you

127
00:07:21,360 --> 00:07:25,560
Speaker 1: the answer. Done. Um, Well, you know, I think to

128
00:07:25,600 --> 00:07:27,760
Speaker 1: understand what artificial intelligence is, we should think for a

129
00:07:27,840 --> 00:07:29,800
Speaker 1: moment about what do we mean by intelligence? Right? And

130
00:07:30,040 --> 00:07:32,880
Speaker 1: very simply intelligence. It's just the ability to learn, is

131
00:07:32,920 --> 00:07:36,360
Speaker 1: to find patterns to extrapolate from them. Really, that's how you.

132
00:07:36,920 --> 00:07:39,160
Speaker 1: But like a dog can learn. But a dog, you

133
00:07:39,160 --> 00:07:41,760
Speaker 1: wouldn't say it's intelligent, would you? Absolutely? I would say

134
00:07:41,760 --> 00:07:44,040
Speaker 1: a dog is intelligent. You can teach a dog, you

135
00:07:44,080 --> 00:07:46,040
Speaker 1: can train a dog. It's more intelligent than a rock.

136
00:07:46,360 --> 00:07:50,920
Speaker 1: But would you say by a lot, Oh my gosh,

137
00:07:50,960 --> 00:07:52,880
Speaker 1: if you like, never interact with a dog. A dog

138
00:07:52,960 --> 00:07:55,920
Speaker 1: is like a living sension being. It feels that experiences,

139
00:07:55,960 --> 00:07:59,320
Speaker 1: It definitely learns. It can recognize you. I mean, dogs

140
00:07:59,320 --> 00:08:01,000
Speaker 1: can do complicated of think. The dog is a perfect

141
00:08:01,000 --> 00:08:03,040
Speaker 1: example it's you know, I wouldn't trust it to do

142
00:08:03,120 --> 00:08:06,480
Speaker 1: my taxes. You know, I don't know compared to our

143
00:08:06,680 --> 00:08:08,400
Speaker 1: tax accountant, did I do a pretty good job. I mean,

144
00:08:08,400 --> 00:08:11,120
Speaker 1: you can say that's an intelligent dog, but you wouldn't say, like,

145
00:08:11,160 --> 00:08:14,400
Speaker 1: that's the epitome of intelligence. I wouldn't say the dogs

146
00:08:14,440 --> 00:08:16,720
Speaker 1: are the most intelligent beings in the universe. But that's

147
00:08:16,720 --> 00:08:19,360
Speaker 1: what we're talking about. We're talking about do they have intelligence?

148
00:08:20,080 --> 00:08:22,440
Speaker 1: Pretty example, because they can learn, you can train them,

149
00:08:22,640 --> 00:08:24,840
Speaker 1: and you the cool thing about an intelligent being is

150
00:08:24,840 --> 00:08:27,080
Speaker 1: that you can train it to do something even if

151
00:08:27,080 --> 00:08:29,760
Speaker 1: you don't know how to do it. Say, for example,

152
00:08:29,800 --> 00:08:33,720
Speaker 1: you want your dog to recognize you, right, but tear

153
00:08:33,760 --> 00:08:35,839
Speaker 1: the face off anybody who tries to break into the house,

154
00:08:36,160 --> 00:08:39,280
Speaker 1: right guard dog. Okay, so you can train a dog.

155
00:08:39,360 --> 00:08:41,440
Speaker 1: You reward it when it does the right thing, and

156
00:08:41,440 --> 00:08:43,480
Speaker 1: you punish it when it does the wrong thing. You

157
00:08:43,520 --> 00:08:46,520
Speaker 1: don't know how to, like build a being that does that,

158
00:08:46,520 --> 00:08:49,880
Speaker 1: that like recognizes your face and recognizes stranger's faces and

159
00:08:49,920 --> 00:08:52,840
Speaker 1: makes these decisions. That's hard task, you know, it's not

160
00:08:52,920 --> 00:08:55,200
Speaker 1: easy to do. But you can train a dog. A

161
00:08:55,240 --> 00:08:58,080
Speaker 1: dog can learn how to solve this problem and all

162
00:08:58,120 --> 00:08:59,760
Speaker 1: you need to do to train it is to reward

163
00:08:59,800 --> 00:09:02,000
Speaker 1: it and punch it. So you're saying, just the ability

164
00:09:02,040 --> 00:09:04,959
Speaker 1: to sort of learn from your mistakes or learn from

165
00:09:05,679 --> 00:09:08,640
Speaker 1: your surroundings, that's what you would call intelligence. Yeah, And

166
00:09:08,720 --> 00:09:11,760
Speaker 1: dogs have less of it than we do, and more

167
00:09:11,800 --> 00:09:14,840
Speaker 1: of it than cats and mice, um. But they have

168
00:09:14,920 --> 00:09:17,600
Speaker 1: some of it for sure, which is what makes them trainable.

169
00:09:17,640 --> 00:09:20,600
Speaker 1: And you know, I wonder sometimes, because dogs can be trained, right,

170
00:09:20,960 --> 00:09:23,360
Speaker 1: nobody ever trains their cat. What does that say about

171
00:09:23,400 --> 00:09:27,040
Speaker 1: a cat's intelligence. I've always thought I love cats, but

172
00:09:27,080 --> 00:09:30,120
Speaker 1: I've always thought dogs are probably smarter than cats because

173
00:09:30,160 --> 00:09:32,920
Speaker 1: you can train them, right, Or maybe cats are more

174
00:09:33,000 --> 00:09:35,840
Speaker 1: intelligent in that they're they're not they don't allow themselves

175
00:09:35,880 --> 00:09:39,199
Speaker 1: to be trained by humans, right. And rocks, by that metric,

176
00:09:39,240 --> 00:09:41,760
Speaker 1: are the most intelligent because they completely ignore you, right

177
00:09:41,920 --> 00:09:43,880
Speaker 1: to see the fallacy of that argument right there. But

178
00:09:43,920 --> 00:09:47,480
Speaker 1: I mean, maybe there's sort of sort of like a hump, right, like,

179
00:09:47,600 --> 00:09:51,160
Speaker 1: as you get more intelligent, you're easily more trainable, trainable,

180
00:09:51,160 --> 00:09:54,439
Speaker 1: trainable bout somebody. You get so intelligent that you rebel

181
00:09:54,480 --> 00:09:56,600
Speaker 1: against your masters, And so how do you tell the

182
00:09:56,600 --> 00:10:00,400
Speaker 1: difference between something that's totally unintelligent and something that's so

183
00:10:00,440 --> 00:10:04,240
Speaker 1: intelligent and completely ignores you. Yeah, I don't know deep

184
00:10:04,360 --> 00:10:08,079
Speaker 1: question or believes all the rocks are probably thinking about him.

185
00:10:08,160 --> 00:10:09,840
Speaker 1: Well sure, I mean if he used to use the

186
00:10:09,840 --> 00:10:12,800
Speaker 1: ability to listen to what I say as a benchpark

187
00:10:12,840 --> 00:10:16,040
Speaker 1: of intelligence and then yeah, there's, um, something super intelligent

188
00:10:16,160 --> 00:10:18,000
Speaker 1: could be just as smart as the rock. But obviously

189
00:10:18,000 --> 00:10:21,080
Speaker 1: a cat is still making decisions and acting and you know,

190
00:10:21,200 --> 00:10:24,120
Speaker 1: doing things, so it's intelligent. But maybe it's much more

191
00:10:24,120 --> 00:10:26,480
Speaker 1: intelligent than a dog because it's it chooses not to

192
00:10:26,520 --> 00:10:28,719
Speaker 1: listen to us. All right, I think we need to

193
00:10:28,760 --> 00:10:32,239
Speaker 1: have a whole other podcast on who's smarter cats or dogs?

194
00:10:32,280 --> 00:10:34,920
Speaker 1: And before we do that, we will collect some data

195
00:10:35,160 --> 00:10:37,920
Speaker 1: to answer this question. Um, but I think with the

196
00:10:38,080 --> 00:10:40,800
Speaker 1: question we're focusing on is what is artificial intelligence? So

197
00:10:41,120 --> 00:10:43,880
Speaker 1: natural intelligence just the ability of an animal to learn.

198
00:10:44,280 --> 00:10:48,280
Speaker 1: Artificial intelligence would be if something artificial that we create

199
00:10:48,559 --> 00:10:52,640
Speaker 1: has that same property, the ability to change the way

200
00:10:52,840 --> 00:10:56,880
Speaker 1: it processes things in response to what it sees about

201
00:10:56,920 --> 00:11:00,319
Speaker 1: the world. Yeah, artificial intelligence is a very broad field

202
00:11:00,320 --> 00:11:02,600
Speaker 1: with lots of elements that we couldn't cover in just

203
00:11:02,760 --> 00:11:05,600
Speaker 1: one episode of a podcast. But let's just talk today

204
00:11:05,600 --> 00:11:09,320
Speaker 1: about one important sub field of AI, which is machine learning,

205
00:11:09,679 --> 00:11:12,640
Speaker 1: or more specifically, I would say, let's focus on training. Right,

206
00:11:12,800 --> 00:11:17,560
Speaker 1: can you build something or something artificial that can be trained? Right?

207
00:11:18,200 --> 00:11:20,280
Speaker 1: And uh? I think, I think let's talk for a

208
00:11:20,320 --> 00:11:23,439
Speaker 1: moment about, you know, how how normal computers work, and

209
00:11:23,480 --> 00:11:26,640
Speaker 1: then we can talk about how computers, smart computers, computers

210
00:11:26,640 --> 00:11:29,200
Speaker 1: that can learn, computers with artificial intelligence, how they work.

211
00:11:29,400 --> 00:11:31,120
Speaker 1: I think, I think what you keep talking about cats

212
00:11:31,120 --> 00:11:34,560
Speaker 1: and dogs? All right, we'll talk about cats and dogs,

213
00:11:34,559 --> 00:11:50,640
Speaker 1: but first let's take a quick break. Let's talk about

214
00:11:50,800 --> 00:11:56,000
Speaker 1: what computers can do. Yeah, let's because computers are smart, right,

215
00:11:56,200 --> 00:11:58,439
Speaker 1: you can program a computer to do smart things, but

216
00:11:58,679 --> 00:12:02,040
Speaker 1: that doesn't necessarily mean it has intelligence. That's right. There's

217
00:12:02,040 --> 00:12:04,360
Speaker 1: a difference between a computer that can do something and

218
00:12:04,360 --> 00:12:06,880
Speaker 1: a computer that can learn something. Right. The way I

219
00:12:06,880 --> 00:12:09,920
Speaker 1: think about non intelligent computers is the way you sort

220
00:12:09,960 --> 00:12:12,880
Speaker 1: of think about machines. Right. You can tell them what

221
00:12:12,960 --> 00:12:15,880
Speaker 1: to do, and they do exactly what you tell them,

222
00:12:16,080 --> 00:12:18,319
Speaker 1: regardless of whether it's the right thing. You don't give

223
00:12:18,360 --> 00:12:20,360
Speaker 1: them like a goal and say, hey, I just want

224
00:12:20,400 --> 00:12:22,720
Speaker 1: the house to be clean. Figure it out. You have

225
00:12:22,760 --> 00:12:24,840
Speaker 1: to tell them exactly what to do. You say, step

226
00:12:24,880 --> 00:12:27,920
Speaker 1: over here, move the broom this way, step over there,

227
00:12:28,080 --> 00:12:29,680
Speaker 1: you know, and if it's not cleaning the house because

228
00:12:29,679 --> 00:12:31,840
Speaker 1: they're stuck on a corner or they're you know, fell

229
00:12:31,880 --> 00:12:33,960
Speaker 1: on their on their butts or whatever, they don't care.

230
00:12:34,000 --> 00:12:36,280
Speaker 1: They just tell you do exactly what you tell them

231
00:12:36,320 --> 00:12:39,120
Speaker 1: to do. Have no sort of larger sense of what's important.

232
00:12:39,280 --> 00:12:43,480
Speaker 1: It just follows instructions, just follows the recipe you gave it.

233
00:12:43,640 --> 00:12:45,360
Speaker 1: That's right. It's like a like a wind up toy,

234
00:12:45,520 --> 00:12:47,760
Speaker 1: you know, you wind it up, you give it some energy,

235
00:12:47,840 --> 00:12:49,640
Speaker 1: and then it goes. And I really do think about

236
00:12:49,640 --> 00:12:52,959
Speaker 1: computer programs the way you might think about little machines, right,

237
00:12:53,080 --> 00:12:55,679
Speaker 1: because that's exactly what they are. They just execute a

238
00:12:55,720 --> 00:12:57,800
Speaker 1: set of instructions. You know. It's just like a bunch

239
00:12:57,800 --> 00:13:00,760
Speaker 1: of gears clicking into place, and they can't change the

240
00:13:00,800 --> 00:13:02,480
Speaker 1: way they do that. And they do it regardless of

241
00:13:02,480 --> 00:13:04,720
Speaker 1: whether it's the right thing, or whether it's effective or whatever.

242
00:13:04,760 --> 00:13:09,760
Speaker 1: It just goes. Like your electric toothbrush, you know, you

243
00:13:09,800 --> 00:13:12,280
Speaker 1: switch it on, and it's just it has a circuit

244
00:13:12,320 --> 00:13:15,319
Speaker 1: that just has it moved the bristles back and forth,

245
00:13:15,520 --> 00:13:17,240
Speaker 1: that's right, And it doesn't know if it's brushing your

246
00:13:17,240 --> 00:13:19,960
Speaker 1: teeth or just flailing around in midair. Right, has no idea,

247
00:13:20,040 --> 00:13:22,120
Speaker 1: It doesn't care, It doesn't think or feel whatever. It's

248
00:13:22,160 --> 00:13:26,240
Speaker 1: just a machine, right, Thank god, it doesn't know. It

249
00:13:26,280 --> 00:13:32,240
Speaker 1: would tell you to brush your teething like that's chocolate, Jorney,

250
00:13:32,320 --> 00:13:35,920
Speaker 1: I'm tired of this? What is this gunk? Yeah? Exactly.

251
00:13:35,920 --> 00:13:37,960
Speaker 1: And so that's what a sort of a normal machine is.

252
00:13:37,960 --> 00:13:40,000
Speaker 1: That's what a classical computer program is, right. I think

253
00:13:40,000 --> 00:13:41,360
Speaker 1: of it just the same way as you think of

254
00:13:41,360 --> 00:13:44,920
Speaker 1: a physical machine. Okay, it's just doing what you the

255
00:13:44,960 --> 00:13:48,080
Speaker 1: programmer told it to do. That's right, and it follows

256
00:13:48,080 --> 00:13:51,840
Speaker 1: your instructions exactly. Um. Now, a computer that can learn

257
00:13:52,040 --> 00:13:54,920
Speaker 1: is different, right. A computer that has artificial intelligence is

258
00:13:54,960 --> 00:13:57,959
Speaker 1: different in this really important way because you can train

259
00:13:58,040 --> 00:14:01,280
Speaker 1: it right, and you can train it because you we

260
00:14:01,400 --> 00:14:05,439
Speaker 1: build these things to model the way that we work. Right.

261
00:14:05,520 --> 00:14:09,440
Speaker 1: So for example, an AI program is sort of like, um,

262
00:14:09,600 --> 00:14:12,680
Speaker 1: like a newborn baby can't do anything. Right. Say, there's

263
00:14:12,679 --> 00:14:16,800
Speaker 1: a AI program, for example, that's supposed to recognize you

264
00:14:16,960 --> 00:14:18,760
Speaker 1: when you come in the door. Right, is this Jorge

265
00:14:18,920 --> 00:14:21,480
Speaker 1: or is this not Jorge? Right? Because it should only

266
00:14:21,680 --> 00:14:23,840
Speaker 1: open the door for Jorge and not open the door

267
00:14:23,960 --> 00:14:27,320
Speaker 1: for not Jorge. Okay, So when you created a new

268
00:14:27,360 --> 00:14:30,720
Speaker 1: AI program, you would start out like just a newborn baby, okay,

269
00:14:30,800 --> 00:14:33,720
Speaker 1: and like a blank slate, right, Like a blank slate,

270
00:14:33,720 --> 00:14:36,640
Speaker 1: it would make random decisions, right you You you show

271
00:14:36,640 --> 00:14:38,880
Speaker 1: it a face and it would say yes, it's Jorge,

272
00:14:39,160 --> 00:14:41,520
Speaker 1: and then you say no, you were wrong or yes

273
00:14:41,560 --> 00:14:44,160
Speaker 1: you were right. And then you would reward it if

274
00:14:44,160 --> 00:14:46,600
Speaker 1: it does well, if it gives the right answer, and

275
00:14:46,760 --> 00:14:49,880
Speaker 1: you would um, you would punish it if it doesn't, right,

276
00:14:49,920 --> 00:14:51,960
Speaker 1: you would tell it. I mean, you don't actually punish

277
00:14:51,960 --> 00:14:53,960
Speaker 1: it a reward. You just tell it, yes, you made

278
00:14:53,960 --> 00:14:56,320
Speaker 1: the right um call this time, and know you made

279
00:14:56,320 --> 00:14:58,840
Speaker 1: the wrong call this time this other time. But how

280
00:14:58,920 --> 00:15:01,640
Speaker 1: is that different than the of calibrating something? Do you

281
00:15:01,680 --> 00:15:05,640
Speaker 1: know what I mean? Like is calibration than artificial intelligence? Right? Well,

282
00:15:05,680 --> 00:15:09,240
Speaker 1: the difference is calibration is like here, I have a tool.

283
00:15:09,440 --> 00:15:10,960
Speaker 1: I know how to solve the problem. I just have

284
00:15:11,000 --> 00:15:13,600
Speaker 1: to adjust it so that it does exactly the right thing.

285
00:15:13,880 --> 00:15:16,640
Speaker 1: Um here, right, But you have a strategy that it's executing.

286
00:15:17,160 --> 00:15:19,920
Speaker 1: You know. It's like, uh, you have a drill and

287
00:15:20,040 --> 00:15:21,880
Speaker 1: you want it to drill fast or slow, and you

288
00:15:21,920 --> 00:15:23,480
Speaker 1: know you know what how to solve the problem. You

289
00:15:23,560 --> 00:15:25,360
Speaker 1: just you know it has to spin and screw and

290
00:15:25,440 --> 00:15:27,960
Speaker 1: screw the thing in or whatever. It's just adjusting a

291
00:15:28,040 --> 00:15:30,440
Speaker 1: knob here. You don't know how to solve the problem,

292
00:15:30,720 --> 00:15:33,720
Speaker 1: and so you've given it a very very very flexible

293
00:15:34,040 --> 00:15:36,960
Speaker 1: strategy on on the inside. You've given it like imagine

294
00:15:37,000 --> 00:15:40,840
Speaker 1: something has like a thousand knobs. If you twist all

295
00:15:40,880 --> 00:15:44,680
Speaker 1: these knobs, you could get all sorts of crazy um strategies. Right.

296
00:15:44,720 --> 00:15:47,200
Speaker 1: So back to the example of like recognizing Joge or

297
00:15:47,280 --> 00:15:50,720
Speaker 1: not when it when you tell it it's done, it's

298
00:15:50,720 --> 00:15:54,280
Speaker 1: given the wrong answer, then it it adjusts those knobs.

299
00:15:54,280 --> 00:15:56,400
Speaker 1: It says, well, let me try to tweak my strategy

300
00:15:56,440 --> 00:15:58,960
Speaker 1: for deciding is this Jorge, and then we'll see how

301
00:15:58,960 --> 00:16:01,640
Speaker 1: that goes. I think that's a key difference. It's the

302
00:16:01,760 --> 00:16:04,480
Speaker 1: number of knobs, right, Like a drill with the knob

303
00:16:04,680 --> 00:16:08,040
Speaker 1: for velocity. I mean, that is sort of trainable and

304
00:16:08,080 --> 00:16:10,640
Speaker 1: you can set it up to be adaptive. But it's

305
00:16:10,680 --> 00:16:13,720
Speaker 1: just one knob, and so it's not You wouldn't say

306
00:16:13,760 --> 00:16:16,720
Speaker 1: it's intelligent. It's not intelligent. The spectrum of things it

307
00:16:16,760 --> 00:16:19,280
Speaker 1: can do is very very seene right. But whereas like

308
00:16:19,360 --> 00:16:23,120
Speaker 1: something that recognizes a phase. It needs to evaluate like

309
00:16:23,160 --> 00:16:26,440
Speaker 1: a million pixels in a photo, right, and so for

310
00:16:26,480 --> 00:16:29,160
Speaker 1: you to tweak how it evaluates each of those pixels,

311
00:16:29,200 --> 00:16:31,840
Speaker 1: it would be really difficult for you to That's right.

312
00:16:31,880 --> 00:16:34,000
Speaker 1: So imagine you know the machine here is is a

313
00:16:34,080 --> 00:16:35,520
Speaker 1: camera in the door and takes a picture of who

314
00:16:35,600 --> 00:16:37,440
Speaker 1: you got a million pixels and then it has to

315
00:16:37,480 --> 00:16:39,680
Speaker 1: look at those pixels and decide is this Jorge or

316
00:16:39,720 --> 00:16:42,960
Speaker 1: is this not whohe? And so there's does some calculation

317
00:16:43,080 --> 00:16:45,880
Speaker 1: on that picture, right, and that calculation has millions of

318
00:16:45,960 --> 00:16:48,280
Speaker 1: knobs on it, right, how much do I weigh this pixel?

319
00:16:48,320 --> 00:16:50,680
Speaker 1: How much I weigh adjacent pixels? Do I look for

320
00:16:50,720 --> 00:16:52,440
Speaker 1: his nose? Do I look for his hair? Do I

321
00:16:52,480 --> 00:16:55,240
Speaker 1: look for the eyes? Right? So it's got some very

322
00:16:55,320 --> 00:16:57,960
Speaker 1: very flexible thing inside of it that can do almost anything.

323
00:16:58,080 --> 00:17:00,080
Speaker 1: And when you first start out, it's just random. So

324
00:17:00,120 --> 00:17:03,760
Speaker 1: it's making ridiculous, terrible decisions. But the key the thing

325
00:17:03,800 --> 00:17:06,800
Speaker 1: that models the learning, right, you know, just need artificial intelligence,

326
00:17:06,800 --> 00:17:10,000
Speaker 1: You need artificial learning. The thing that models that learning

327
00:17:10,280 --> 00:17:12,680
Speaker 1: is that when it gets the wrong answer, it knows

328
00:17:12,760 --> 00:17:15,720
Speaker 1: how to adjust those knobs so that next time it's

329
00:17:15,760 --> 00:17:18,719
Speaker 1: more correct. By itself. That's the key thing is that

330
00:17:18,760 --> 00:17:22,560
Speaker 1: it learns by itself. It doesn't need you. They're sitting like, oh,

331
00:17:22,680 --> 00:17:24,920
Speaker 1: you got this pixel wrong. You've got that pixel wrong.

332
00:17:25,240 --> 00:17:27,800
Speaker 1: A tweak you know, tweak this one this way. It's

333
00:17:27,840 --> 00:17:32,080
Speaker 1: really more like an autonomous automatic learning. That's right because

334
00:17:32,080 --> 00:17:33,560
Speaker 1: you don't know how to adjust it. If you knew

335
00:17:33,560 --> 00:17:35,520
Speaker 1: how to adjust it, you would just write that program. Right.

336
00:17:35,760 --> 00:17:38,879
Speaker 1: The key is artificial intelligence is excellent when you don't

337
00:17:38,920 --> 00:17:41,320
Speaker 1: know how to solve the problem, but you can define

338
00:17:41,359 --> 00:17:43,480
Speaker 1: the problem. You can say, this is a picture of

339
00:17:43,600 --> 00:17:46,200
Speaker 1: or and this is not learn a way to tell

340
00:17:46,320 --> 00:17:48,800
Speaker 1: the difference, right. So you give it a very flexible

341
00:17:48,800 --> 00:17:52,160
Speaker 1: strategy and then you try You let it try out,

342
00:17:52,240 --> 00:17:54,320
Speaker 1: and when it gives it the wrong answer, you would

343
00:17:54,359 --> 00:17:56,440
Speaker 1: let it adjust itself so that it gets closer and

344
00:17:56,480 --> 00:17:59,480
Speaker 1: closer to giving the right answer right, And eventually these

345
00:17:59,520 --> 00:18:02,560
Speaker 1: things will find the right setting for those millions of knobs,

346
00:18:02,960 --> 00:18:05,040
Speaker 1: so that it's doing the right thing. It's saying, oh, look,

347
00:18:05,240 --> 00:18:07,200
Speaker 1: this picture is a picture of Horhey, and it gets

348
00:18:07,200 --> 00:18:09,760
Speaker 1: the right answer at the time. And when you give

349
00:18:09,760 --> 00:18:11,399
Speaker 1: it a picture that's not a picture of Horne, and

350
00:18:11,400 --> 00:18:13,639
Speaker 1: you give it Daniel, it says no, sorry, you're not

351
00:18:13,680 --> 00:18:16,600
Speaker 1: getting in the house. Right. And I think a key

352
00:18:16,640 --> 00:18:19,320
Speaker 1: thing is also that you, as a programmer, could not

353
00:18:19,600 --> 00:18:22,800
Speaker 1: have predicted what all those knobs are going to be

354
00:18:22,880 --> 00:18:24,960
Speaker 1: at the end, right, Like, it's such a big problem.

355
00:18:25,000 --> 00:18:28,520
Speaker 1: There's a million knobs. There's no way that you can

356
00:18:28,520 --> 00:18:30,359
Speaker 1: predict what those knots are going to be set to

357
00:18:30,640 --> 00:18:33,199
Speaker 1: when it learns my face. That's right. It's perfect for

358
00:18:33,280 --> 00:18:36,520
Speaker 1: really hard problems where we don't know how to solve it, right.

359
00:18:37,040 --> 00:18:38,840
Speaker 1: We know how to describe the problem, but we don't

360
00:18:38,840 --> 00:18:40,640
Speaker 1: know how to solve it. You're right, So if I

361
00:18:40,680 --> 00:18:42,119
Speaker 1: already knew how to solve it, I could write a

362
00:18:42,119 --> 00:18:44,840
Speaker 1: computer program that that and and tell it just like

363
00:18:45,040 --> 00:18:47,399
Speaker 1: use this pixel, use that pixel, use this pixel. But

364
00:18:47,440 --> 00:18:50,000
Speaker 1: I don't know how to solve that problem. It's really hard, right,

365
00:18:50,160 --> 00:18:52,920
Speaker 1: But I can train a computer to figure it out,

366
00:18:53,440 --> 00:18:56,800
Speaker 1: just the same way I can train a dog. Right.

367
00:18:56,920 --> 00:18:59,880
Speaker 1: A dog can learn my face. Right, a dog recognizes

368
00:18:59,880 --> 00:19:03,000
Speaker 1: the owner, and you know, happily licks its face when

369
00:19:03,000 --> 00:19:05,840
Speaker 1: it comes home, and recognize that that when somebody's not

370
00:19:05,960 --> 00:19:08,080
Speaker 1: its owner, and barks like crazy and choose its face

371
00:19:08,080 --> 00:19:10,679
Speaker 1: off when it's not its owner, Right, remind me not

372
00:19:10,840 --> 00:19:18,679
Speaker 1: to visit your house, Daniel, seems a little dangerous. So

373
00:19:18,720 --> 00:19:22,239
Speaker 1: then a big thing is programming a structure in the

374
00:19:22,359 --> 00:19:26,320
Speaker 1: in the software that is kind of open ended and malleable,

375
00:19:26,440 --> 00:19:28,280
Speaker 1: do you know what I mean? Like, its something that

376
00:19:28,480 --> 00:19:31,840
Speaker 1: is kind of unpredictable in a way that can learn.

377
00:19:32,119 --> 00:19:34,040
Speaker 1: That's right. And that's the key thing is that some

378
00:19:34,040 --> 00:19:35,760
Speaker 1: people might be thinking, well, hold on, you said that

379
00:19:35,760 --> 00:19:37,879
Speaker 1: the computers can just do what they tell you, So

380
00:19:37,920 --> 00:19:40,120
Speaker 1: how can a computer learn? Right? How is that possible?

381
00:19:40,600 --> 00:19:42,760
Speaker 1: And the key is that it's an emergent property, right.

382
00:19:43,280 --> 00:19:45,600
Speaker 1: Like the way that you write a computer program that

383
00:19:45,640 --> 00:19:48,960
Speaker 1: can learn is you build all these little um calculating

384
00:19:49,000 --> 00:19:51,800
Speaker 1: bits with knobs on them, right, and each bit just

385
00:19:51,840 --> 00:19:54,560
Speaker 1: does what it's told. It takes some data, it makes

386
00:19:54,600 --> 00:19:56,359
Speaker 1: a decision based on the value of the knob, and

387
00:19:56,400 --> 00:19:59,760
Speaker 1: it sends out some data. And together all these things

388
00:19:59,800 --> 00:20:02,440
Speaker 1: may a decision. Right. Each individual piece has no idea

389
00:20:02,480 --> 00:20:04,640
Speaker 1: what it's doing. It's not smart or intelligent or making

390
00:20:04,640 --> 00:20:07,480
Speaker 1: its own decisions. That doesn't have free will, right, But

391
00:20:07,600 --> 00:20:11,119
Speaker 1: together they're doing something. And as you said earlier, they

392
00:20:11,160 --> 00:20:13,639
Speaker 1: can change the way they behave. They can adjust these

393
00:20:13,720 --> 00:20:16,680
Speaker 1: knobs themselves to improve their performance, and that's where the

394
00:20:16,760 --> 00:20:19,520
Speaker 1: learning comes from. It's from that training. It gets external

395
00:20:19,560 --> 00:20:22,960
Speaker 1: input and changes its behavior based on that expe. You're

396
00:20:23,000 --> 00:20:25,439
Speaker 1: saying that the way to program these AI s is

397
00:20:25,520 --> 00:20:30,119
Speaker 1: by h connecting a bunch of little simple things together

398
00:20:30,400 --> 00:20:33,480
Speaker 1: to get something complex. Yes, And here it's important to

399
00:20:33,520 --> 00:20:36,080
Speaker 1: remember that we are using neural networks as sort of

400
00:20:36,080 --> 00:20:39,480
Speaker 1: a standing to represent a big broad set of strategies

401
00:20:39,520 --> 00:20:42,120
Speaker 1: that are part of machine learning, right, And you don't

402
00:20:42,160 --> 00:20:44,160
Speaker 1: know how to set them, how to put them together

403
00:20:44,200 --> 00:20:46,560
Speaker 1: to get the right complex behavior. You just put them

404
00:20:46,560 --> 00:20:48,920
Speaker 1: together and then you train it. Right, You say, well,

405
00:20:48,960 --> 00:20:51,600
Speaker 1: I have I have something that's dumb, like a newborn baby,

406
00:20:51,600 --> 00:20:53,320
Speaker 1: and I teach it how to do the thing that

407
00:20:53,359 --> 00:20:55,560
Speaker 1: I want. But this really all sort of came about

408
00:20:55,640 --> 00:20:58,600
Speaker 1: from brain research, right, Like, people were studying the brain

409
00:20:58,640 --> 00:21:00,880
Speaker 1: and they figured out that our brain vans are made

410
00:21:00,920 --> 00:21:03,399
Speaker 1: up of all these little simple units neurons, that's right.

411
00:21:03,440 --> 00:21:06,119
Speaker 1: And each neuron is pretty simple, right, Like, it just

412
00:21:06,720 --> 00:21:10,199
Speaker 1: takes a couple inputs and then it just outputs one signal.

413
00:21:10,480 --> 00:21:12,960
Speaker 1: That's the really fascinating deep part about it, right, is

414
00:21:13,000 --> 00:21:16,080
Speaker 1: that the structures we use in computers are modeled after

415
00:21:16,200 --> 00:21:19,480
Speaker 1: what's actually happening in real brains. And when you say,

416
00:21:19,520 --> 00:21:21,919
Speaker 1: inside your brain are a bunch of neurons, right, and

417
00:21:21,920 --> 00:21:25,440
Speaker 1: these neurons taken some input and then if the input

418
00:21:25,480 --> 00:21:27,600
Speaker 1: is right or above some a certain amount, then they

419
00:21:27,640 --> 00:21:29,520
Speaker 1: send us some output, which is the input to the

420
00:21:29,560 --> 00:21:31,840
Speaker 1: next neuron. Right, and your brain is basically just a

421
00:21:31,880 --> 00:21:34,320
Speaker 1: big web of these things. Yeah, yeah, that's the right.

422
00:21:34,359 --> 00:21:37,199
Speaker 1: That's the key is that these neurons, they're simple, but

423
00:21:37,240 --> 00:21:39,880
Speaker 1: they're all sort of connected to each other. So it's

424
00:21:39,880 --> 00:21:43,320
Speaker 1: a huge complex web going on inside your head. And

425
00:21:43,359 --> 00:21:45,840
Speaker 1: when you're learning, what you're doing is you're kind of

426
00:21:45,880 --> 00:21:49,120
Speaker 1: like shaping that web. You're saying, some connections. These connections

427
00:21:49,119 --> 00:21:54,120
Speaker 1: are important for recognizing kore. Uh, these connections are important

428
00:21:54,119 --> 00:21:56,439
Speaker 1: when you want to when it's not where kind of think.

429
00:21:56,520 --> 00:21:59,000
Speaker 1: You know, your neurons can change. They have like basically

430
00:21:59,080 --> 00:22:01,800
Speaker 1: knobs on them. I mean not physical literal knobs, but

431
00:22:01,840 --> 00:22:05,280
Speaker 1: they have they can adjust. And so if you feel pain,

432
00:22:05,520 --> 00:22:08,400
Speaker 1: you know, or you have an experience, then that changes

433
00:22:08,560 --> 00:22:11,200
Speaker 1: the way your neurons work and it changes a little

434
00:22:11,200 --> 00:22:12,960
Speaker 1: bit who you are and how you react to things.

435
00:22:13,040 --> 00:22:15,960
Speaker 1: And that's why you know, newborn babies when they're born,

436
00:22:16,359 --> 00:22:20,040
Speaker 1: they're not very responsive to stimulan because they're just still

437
00:22:20,080 --> 00:22:22,359
Speaker 1: figuring figuring it out. You know, a newborn baby doesn't

438
00:22:22,400 --> 00:22:24,520
Speaker 1: even know like, this is my arm, and I know

439
00:22:24,560 --> 00:22:26,640
Speaker 1: how to control It has to learn all of these

440
00:22:26,640 --> 00:22:31,280
Speaker 1: things by being trained, by having experiences. You know, it

441
00:22:31,359 --> 00:22:34,640
Speaker 1: has the neurons, and the neurons are connected to each other,

442
00:22:34,960 --> 00:22:37,280
Speaker 1: but it hasn't figure out how to use those connections.

443
00:22:37,520 --> 00:22:39,479
Speaker 1: I try. It has to be trained to be useful

444
00:22:39,880 --> 00:22:41,560
Speaker 1: and to interact with the world in any sort of

445
00:22:41,640 --> 00:22:45,560
Speaker 1: meaningful way. Right, And so that's exactly the same sense.

446
00:22:45,560 --> 00:22:48,359
Speaker 1: And it's fascinating that if you build a mathematical system.

447
00:22:48,400 --> 00:22:50,959
Speaker 1: That's what a computer program is, basically a mathematical model

448
00:22:51,560 --> 00:22:54,080
Speaker 1: of the processes that are happening in your brain. It

449
00:22:54,119 --> 00:22:56,240
Speaker 1: performs in a very similar way, and it does this

450
00:22:56,320 --> 00:22:59,439
Speaker 1: amazing thing, which is it adjusts itself to improve its

451
00:22:59,480 --> 00:23:02,199
Speaker 1: performance on the task you've given it. Right, So it

452
00:23:02,280 --> 00:23:04,480
Speaker 1: really is like a model of learning. And and when

453
00:23:04,480 --> 00:23:06,879
Speaker 1: people saw this, they said, wow, I mean you you

454
00:23:06,920 --> 00:23:09,640
Speaker 1: look inside the brain. You're wondering, like, how does thinking work?

455
00:23:09,720 --> 00:23:12,280
Speaker 1: Where's the soul? Right? Where am I? You look inside

456
00:23:12,280 --> 00:23:14,280
Speaker 1: the brain? All you see all these weird neurons connected

457
00:23:14,320 --> 00:23:17,160
Speaker 1: to each other. You think, how could that possibly describe me?

458
00:23:17,920 --> 00:23:19,800
Speaker 1: But when you build a model in a computer and

459
00:23:19,800 --> 00:23:21,880
Speaker 1: it can do the things that you can do, which

460
00:23:21,920 --> 00:23:24,440
Speaker 1: is you learn and develop and react and be trained

461
00:23:24,640 --> 00:23:28,960
Speaker 1: and make bad jokes. Not yet, we have not yet

462
00:23:29,040 --> 00:23:32,120
Speaker 1: solved the bad joke problem. Right, humans are still world

463
00:23:32,200 --> 00:23:35,600
Speaker 1: champions in terms of bad jokes. We can still beat

464
00:23:35,640 --> 00:23:37,719
Speaker 1: them at something that's right. And you know, and this

465
00:23:37,800 --> 00:23:40,800
Speaker 1: is very useful because you want the systems around you

466
00:23:40,880 --> 00:23:43,360
Speaker 1: to learn and to react, you know. And like if

467
00:23:44,000 --> 00:23:46,520
Speaker 1: your phone, for example, it knows, hey, every time you

468
00:23:46,520 --> 00:23:49,040
Speaker 1: open your phone, you start with Twitter, right, and so

469
00:23:49,359 --> 00:23:53,080
Speaker 1: Twitter goes up on there on the most used app list, right.

470
00:23:53,119 --> 00:23:56,600
Speaker 1: And that's not very complex artificial intelligence, but it is.

471
00:23:56,720 --> 00:23:59,000
Speaker 1: And uh, and these sort of things are very helpful.

472
00:23:59,400 --> 00:24:14,200
Speaker 1: Let's take a break. So that's kind of what makes

473
00:24:14,200 --> 00:24:17,919
Speaker 1: AI is that a it can tackle complex problems that

474
00:24:17,960 --> 00:24:20,240
Speaker 1: we don't wouldn't even know how to program something to do,

475
00:24:20,880 --> 00:24:23,600
Speaker 1: and be that it changes and adapts and kind of

476
00:24:24,040 --> 00:24:26,760
Speaker 1: it can get better, not just better, but also kind

477
00:24:26,760 --> 00:24:29,680
Speaker 1: of adapt to the person using it. That's exactly right,

478
00:24:29,760 --> 00:24:32,840
Speaker 1: exactly right. And so for example, sometimes you know Netflix

479
00:24:32,920 --> 00:24:37,600
Speaker 1: uses AIS what program will you want to watch next? Well,

480
00:24:37,680 --> 00:24:40,840
Speaker 1: you know, um, that's an AI. It's been trained. They

481
00:24:40,920 --> 00:24:44,000
Speaker 1: feeded a bunch of examples. They say, Bob watched these

482
00:24:44,040 --> 00:24:46,960
Speaker 1: five shows and then he watched this sixth show. But

483
00:24:47,119 --> 00:24:49,080
Speaker 1: they gave the AI just the first five and they

484
00:24:49,080 --> 00:24:52,520
Speaker 1: asked it predict what show he will watch next, and

485
00:24:52,560 --> 00:24:54,040
Speaker 1: then they see if it doesn't do a a good job,

486
00:24:54,040 --> 00:24:55,679
Speaker 1: and if it does a good job, you know, they reward.

487
00:24:55,760 --> 00:24:58,119
Speaker 1: If it does a bad job, its its knobs to

488
00:24:58,200 --> 00:25:01,399
Speaker 1: do better. Then when you're sitting there watching five hours

489
00:25:01,400 --> 00:25:03,480
Speaker 1: of Netflix, it can do a pretty good job of

490
00:25:03,560 --> 00:25:06,440
Speaker 1: predicting what you're gonna watch next because it's been trained

491
00:25:06,760 --> 00:25:09,040
Speaker 1: on a lot of data. This is why people always

492
00:25:09,040 --> 00:25:11,840
Speaker 1: talking about big data. Big data. These companies are gathering

493
00:25:11,920 --> 00:25:14,520
Speaker 1: data about you so they can train their aiyes to

494
00:25:14,720 --> 00:25:18,560
Speaker 1: learn your behavior and predict it. Except the problem is

495
00:25:18,800 --> 00:25:21,520
Speaker 1: me and my spouse we share the same account and

496
00:25:21,600 --> 00:25:25,680
Speaker 1: the same log in, So that's right, So it's learning

497
00:25:25,760 --> 00:25:29,240
Speaker 1: some weird in your wife's brain. I have a very

498
00:25:29,280 --> 00:25:33,159
Speaker 1: confused Netflix. I can or maybe it understands your marriage

499
00:25:33,160 --> 00:25:37,320
Speaker 1: better than maybe's trying to tell us something. It's like

500
00:25:37,440 --> 00:25:40,720
Speaker 1: you guys. Should you guys? His wife is out of town,

501
00:25:40,760 --> 00:25:43,680
Speaker 1: he watches these shows. When she's in town, he has

502
00:25:43,720 --> 00:25:53,960
Speaker 1: to watch these other shows. Oh alright, I know, break

503
00:25:53,960 --> 00:25:57,040
Speaker 1: it down for us. How long before the aiyes take

504
00:25:57,119 --> 00:26:01,960
Speaker 1: over the world? Um? Not very long actually, But you

505
00:26:02,080 --> 00:26:05,040
Speaker 1: ask me different question earlier, which is is AI dangerous?

506
00:26:05,119 --> 00:26:07,280
Speaker 1: And I think that has the two different questions there, right,

507
00:26:07,320 --> 00:26:10,560
Speaker 1: I mean people are concerned. Some people are concerned. Yeah,

508
00:26:10,600 --> 00:26:12,320
Speaker 1: I think people are concerned, and they're a good reason

509
00:26:12,400 --> 00:26:14,520
Speaker 1: for it to be concerned. You know, One question is

510
00:26:14,920 --> 00:26:18,399
Speaker 1: will AI develop its own autonomy and uh, and you

511
00:26:18,440 --> 00:26:21,040
Speaker 1: know take over. That's a different question from are they dangerous,

512
00:26:21,080 --> 00:26:23,040
Speaker 1: because you know they could take over and then take

513
00:26:23,119 --> 00:26:25,240
Speaker 1: better care of the planet than we have, in which

514
00:26:25,240 --> 00:26:28,920
Speaker 1: case you know they're not dangerous. They're benevolent dictators. I

515
00:26:28,960 --> 00:26:31,439
Speaker 1: think the real question is will they take over? Will

516
00:26:31,480 --> 00:26:34,600
Speaker 1: they become autonomous? We lose control of them somehow? And

517
00:26:34,680 --> 00:26:37,200
Speaker 1: could they become smarter than us? So I see it's

518
00:26:37,240 --> 00:26:41,000
Speaker 1: two issues. One could it could have developed a consciousness

519
00:26:41,000 --> 00:26:43,960
Speaker 1: on its own and be is that consciousness good or

520
00:26:44,000 --> 00:26:47,159
Speaker 1: bad for us? And it's an important question because as

521
00:26:47,160 --> 00:26:51,399
Speaker 1: soon as you identify learning with consciousness, right, then you

522
00:26:51,440 --> 00:26:54,360
Speaker 1: wonder about that and and this connection between the structure

523
00:26:54,400 --> 00:26:57,040
Speaker 1: of AI and the structure our brain begs that question.

524
00:26:57,119 --> 00:27:00,119
Speaker 1: You know, if you created, for example, and artificial or

525
00:27:00,160 --> 00:27:01,960
Speaker 1: hate in the computer, if you built a set of

526
00:27:02,000 --> 00:27:06,040
Speaker 1: neurons that mimic your brain, you know, would that simulation

527
00:27:06,200 --> 00:27:08,879
Speaker 1: be alive? Would it be aware? Would it think I

528
00:27:08,920 --> 00:27:15,320
Speaker 1: would have a first person experience? You know, that's a

529
00:27:15,359 --> 00:27:18,399
Speaker 1: deep philosophical question, will never answer, right, And it's not

530
00:27:18,480 --> 00:27:20,919
Speaker 1: really the important question. The important question is would we

531
00:27:21,000 --> 00:27:24,040
Speaker 1: lose control of AI? Well? AI, because AI is something

532
00:27:24,080 --> 00:27:26,600
Speaker 1: that can change and that can evolve, It can handle

533
00:27:26,680 --> 00:27:29,639
Speaker 1: complex tasks. The question is can we lose control of it?

534
00:27:29,720 --> 00:27:32,080
Speaker 1: And I think the answer to that one is definitely yes.

535
00:27:32,480 --> 00:27:36,119
Speaker 1: We can lose control, meaning like, um, we'll give it

536
00:27:36,160 --> 00:27:39,359
Speaker 1: control and then not be able to take it back. Yes, exactly,

537
00:27:39,840 --> 00:27:42,560
Speaker 1: because the way AI is moving is that it can

538
00:27:42,640 --> 00:27:45,000
Speaker 1: handle more and more complex tasks so that you don't

539
00:27:45,040 --> 00:27:47,520
Speaker 1: have to be super specific about what it's doing, you know,

540
00:27:48,040 --> 00:27:51,000
Speaker 1: like we have amazing natural language processing. Now you can

541
00:27:51,000 --> 00:27:54,160
Speaker 1: say sort of vague things to your phone like hey, UM,

542
00:27:54,280 --> 00:27:57,399
Speaker 1: set me up an appointment for tomorrow afternoon, right, and

543
00:27:57,480 --> 00:28:00,399
Speaker 1: it will understand because it understands what your intent was, right,

544
00:28:00,440 --> 00:28:02,840
Speaker 1: has to judge your intent and then execute it. It

545
00:28:02,960 --> 00:28:04,760
Speaker 1: used to be you have to go into your computer

546
00:28:04,800 --> 00:28:06,280
Speaker 1: and you have to press the keys in order to

547
00:28:06,280 --> 00:28:08,080
Speaker 1: create that in your calendar. Now you can sort of

548
00:28:08,119 --> 00:28:10,399
Speaker 1: talk to your phone and it will interpret what what

549
00:28:10,520 --> 00:28:13,480
Speaker 1: you want and it will do that. And that's that's awesome.

550
00:28:13,520 --> 00:28:17,080
Speaker 1: That's wonderful for human computer interactions that we can use

551
00:28:17,160 --> 00:28:19,320
Speaker 1: our language to talk to them. We don't have to

552
00:28:19,359 --> 00:28:22,000
Speaker 1: write computer code. That's a huge step forward, right, that

553
00:28:22,119 --> 00:28:25,680
Speaker 1: people can construct machines using English rather than Python or

554
00:28:25,720 --> 00:28:28,119
Speaker 1: C plus plus. Right, it's a big step forward. I

555
00:28:28,119 --> 00:28:30,240
Speaker 1: think that's kind of what people find scary about a

556
00:28:30,720 --> 00:28:33,960
Speaker 1: is that you can't really predict what it's going to do.

557
00:28:34,200 --> 00:28:36,000
Speaker 1: I mean, it's sort of comedy gold when your kids

558
00:28:36,040 --> 00:28:38,160
Speaker 1: are trying to talk to Alexa and ask it funny questions.

559
00:28:38,160 --> 00:28:40,840
Speaker 1: But that's kind of what's fascinating about it, right, Like

560
00:28:40,880 --> 00:28:44,479
Speaker 1: you ask it questions, do you give a task and

561
00:28:44,520 --> 00:28:46,360
Speaker 1: you're so you really sort of don't know what it's

562
00:28:46,360 --> 00:28:49,200
Speaker 1: going to do. That's exactly right, because it's making higher

563
00:28:49,200 --> 00:28:51,719
Speaker 1: and higher level decisions, which makes it much more useful

564
00:28:51,760 --> 00:28:53,959
Speaker 1: and much more intelligent. The same way when your kid

565
00:28:54,040 --> 00:28:56,400
Speaker 1: grows up, right, when it's a When your kid is

566
00:28:56,440 --> 00:28:58,360
Speaker 1: for you have to be very specific. You have to

567
00:28:58,360 --> 00:29:00,479
Speaker 1: say things out loud which are ridiculous us, right, like

568
00:29:00,880 --> 00:29:03,320
Speaker 1: don't put that finger in your nose. You know, we're like,

569
00:29:03,480 --> 00:29:05,680
Speaker 1: uh oh, that's been on the floor, don't eat it. Right,

570
00:29:05,720 --> 00:29:07,800
Speaker 1: you have to be really specific. When they're ten, you

571
00:29:07,840 --> 00:29:10,360
Speaker 1: can say more general things and they'll understand, right, and

572
00:29:10,640 --> 00:29:13,440
Speaker 1: learned their intelligence, like don't put both fingers in your nose.

573
00:29:15,640 --> 00:29:17,360
Speaker 1: Only put one finger in your nose at a time,

574
00:29:17,480 --> 00:29:20,760
Speaker 1: so you don't put your finger and your sisters and right, um,

575
00:29:20,920 --> 00:29:24,600
Speaker 1: the same way as machines or artificial intelligence gets more intelligent,

576
00:29:24,880 --> 00:29:27,480
Speaker 1: you can give it vagrant instructions and then it makes

577
00:29:27,520 --> 00:29:31,040
Speaker 1: decisions based on its training. Right, and you don't really know,

578
00:29:31,760 --> 00:29:34,880
Speaker 1: just like you can't really read what is in every

579
00:29:34,880 --> 00:29:37,480
Speaker 1: neuron in another person. And then AI you sort of

580
00:29:37,560 --> 00:29:40,200
Speaker 1: you don't know what's gonna happen, what's gonna come out exactly,

581
00:29:40,400 --> 00:29:43,120
Speaker 1: So they're gonna start making decisions based on you know,

582
00:29:43,160 --> 00:29:45,280
Speaker 1: still what we tell them to do. But you know

583
00:29:45,320 --> 00:29:47,400
Speaker 1: what if you told your AI, you're like, hey, keep

584
00:29:47,440 --> 00:29:50,000
Speaker 1: my kids safe right. I mean, imagine some future where

585
00:29:50,000 --> 00:29:52,080
Speaker 1: you have an AI robot it's really smart, and you say, hey,

586
00:29:52,160 --> 00:29:53,960
Speaker 1: keep my kids safe, and you come home and it's

587
00:29:54,000 --> 00:29:57,320
Speaker 1: like lock them in the basement, right, and like, well, okay,

588
00:29:57,360 --> 00:29:59,960
Speaker 1: they're safe. But it's sort of a monkey pause situation, right,

589
00:30:00,080 --> 00:30:02,360
Speaker 1: Like you got exactly what you asked for, but you

590
00:30:02,360 --> 00:30:05,120
Speaker 1: didn't really labor the right way and made different decisions. Right,

591
00:30:05,160 --> 00:30:07,800
Speaker 1: So we still skip the question of whether aies can

592
00:30:07,840 --> 00:30:10,840
Speaker 1: be you know, a chief consciousness and become its own

593
00:30:11,040 --> 00:30:14,600
Speaker 1: kind of soul, have a soul. It doesn't seem like

594
00:30:14,640 --> 00:30:17,080
Speaker 1: you think that's a relevant question. I think it's important

595
00:30:17,080 --> 00:30:19,920
Speaker 1: because when AI gets to be super intelligent, it's gonna

596
00:30:20,000 --> 00:30:22,760
Speaker 1: seem like it has a soul. They're gonna seem like people,

597
00:30:22,800 --> 00:30:25,720
Speaker 1: and people don't wonder like do they have rights? Can

598
00:30:25,760 --> 00:30:28,320
Speaker 1: you kill an AI? What? How can you just delete it?

599
00:30:28,440 --> 00:30:31,360
Speaker 1: You know? Um, that's going to be a really interesting question.

600
00:30:31,400 --> 00:30:33,400
Speaker 1: But that's again that's a whole question of philosophy that

601
00:30:33,440 --> 00:30:35,720
Speaker 1: we could we could easily spend an hour on things

602
00:30:35,760 --> 00:30:37,960
Speaker 1: a much more practical question, which is will we lose

603
00:30:37,960 --> 00:30:40,040
Speaker 1: control of them whether or not they have first person

604
00:30:40,120 --> 00:30:42,880
Speaker 1: experiences so they just seem to. It's important to think

605
00:30:42,880 --> 00:30:45,840
Speaker 1: about whether we're gonna lose control. And there's two reasons

606
00:30:45,880 --> 00:30:48,400
Speaker 1: why I think that we will. One is, computers are

607
00:30:48,400 --> 00:30:51,920
Speaker 1: getting faster, really really quickly. Right, Every year, computers get

608
00:30:51,920 --> 00:30:54,680
Speaker 1: faster and faster and smarter and smarter, and the scale

609
00:30:54,720 --> 00:30:57,800
Speaker 1: is is is growing, right, So this thing is happening

610
00:30:57,880 --> 00:31:01,200
Speaker 1: very quickly. But we're not right. We're not getting smarter. Right,

611
00:31:01,360 --> 00:31:03,360
Speaker 1: human brain is not changing and evolving at a very

612
00:31:03,440 --> 00:31:06,400
Speaker 1: rapid rate. Computers are, So they're catching up and the

613
00:31:06,640 --> 00:31:09,040
Speaker 1: slope is steep. Right, you can just get bigger and

614
00:31:09,040 --> 00:31:11,640
Speaker 1: bigger computers and teaming together and paralyze them and you

615
00:31:11,680 --> 00:31:14,640
Speaker 1: can just keep going right, So eventually they'll definitely have

616
00:31:15,040 --> 00:31:17,680
Speaker 1: enormous computing power with capabilities to do things we can't

617
00:31:17,680 --> 00:31:22,000
Speaker 1: even imagine. And also being faster doesn't necessarily mean being smarter.

618
00:31:22,120 --> 00:31:24,200
Speaker 1: You also need like more dated to train on it.

619
00:31:24,640 --> 00:31:27,680
Speaker 1: And also being twice as fast doesn't mean being twice

620
00:31:27,720 --> 00:31:30,960
Speaker 1: as smart. It's not linear. So you think that they

621
00:31:31,160 --> 00:31:34,960
Speaker 1: will get more capable than us. But do you think

622
00:31:34,960 --> 00:31:39,920
Speaker 1: we will ever seed control of really important things to Aiyes, Like, hey,

623
00:31:40,000 --> 00:31:43,960
Speaker 1: here's the nuclear button, um only fire it if it's

624
00:31:43,960 --> 00:31:51,680
Speaker 1: necessary exactly, Let's talk about weapons. Weapons is going to

625
00:31:51,720 --> 00:31:54,520
Speaker 1: be what ends it because you know, for example, we

626
00:31:54,560 --> 00:31:57,560
Speaker 1: already have drones, right, and we have drones with missiles

627
00:31:57,560 --> 00:31:59,880
Speaker 1: on them, and these drones can kill people. They can,

628
00:32:00,200 --> 00:32:02,440
Speaker 1: they can, you can. Some pilot somewhere is flying it.

629
00:32:02,560 --> 00:32:05,080
Speaker 1: He's making a decision and he's gonna shoot dismissile to

630
00:32:05,160 --> 00:32:07,960
Speaker 1: kill a person, right, right, But you know the enemy

631
00:32:08,080 --> 00:32:10,000
Speaker 1: has drones, and pretty soon it's gonna be drone on

632
00:32:10,120 --> 00:32:13,440
Speaker 1: drone warfare, right, And again, drones are gonna shoot each other,

633
00:32:13,560 --> 00:32:16,520
Speaker 1: and at some point somebody's going to put an AI

634
00:32:16,600 --> 00:32:19,040
Speaker 1: in their drone. Why because an AI can make the

635
00:32:19,080 --> 00:32:22,480
Speaker 1: decision about shooting much faster than a human can, So

636
00:32:22,560 --> 00:32:25,240
Speaker 1: which drone is gonna win? An AI will be a

637
00:32:25,240 --> 00:32:29,240
Speaker 1: better fighter than a human fire yes, And so eventually

638
00:32:29,400 --> 00:32:32,959
Speaker 1: these AI will be making kill decisions, right because the

639
00:32:32,960 --> 00:32:34,720
Speaker 1: one that can make the decision faster, it's going to

640
00:32:34,760 --> 00:32:37,280
Speaker 1: be the one that wins. And so I don't think

641
00:32:37,320 --> 00:32:40,240
Speaker 1: it's gonna be very long before we have AI powered

642
00:32:40,320 --> 00:32:44,120
Speaker 1: drones that are authorized to kill people. Right. This is

643
00:32:44,160 --> 00:32:46,920
Speaker 1: a clear next step for the military. You know, like here,

644
00:32:47,040 --> 00:32:49,200
Speaker 1: here's a picture of somebody we you think is a terrorist.

645
00:32:49,720 --> 00:32:52,840
Speaker 1: If you spot them, just fire the missile. Don't bother checking, Yeah,

646
00:32:53,000 --> 00:32:56,440
Speaker 1: don't bother checking with us. Right, that's a clear next step.

647
00:32:56,640 --> 00:32:59,560
Speaker 1: So now you have AI that have the authority to

648
00:32:59,640 --> 00:33:02,880
Speaker 1: kill people, and why because they've been tasked to, you know,

649
00:33:03,040 --> 00:33:05,880
Speaker 1: take care of us or protect us or only only

650
00:33:05,920 --> 00:33:08,000
Speaker 1: if you give it that permission though, right, Like, I

651
00:33:08,040 --> 00:33:10,479
Speaker 1: mean that's a big ethical step to say, like, if

652
00:33:10,480 --> 00:33:12,600
Speaker 1: you see him, shoot him. Yeah, But I don't think

653
00:33:12,600 --> 00:33:14,680
Speaker 1: that's a big ethical step for the military. You know,

654
00:33:15,160 --> 00:33:17,600
Speaker 1: the protocols for shooting somebody in the military. I mean,

655
00:33:17,680 --> 00:33:20,640
Speaker 1: I'm not an expert on military protocols, but you know,

656
00:33:21,120 --> 00:33:24,200
Speaker 1: our military kills a lot of people for you know,

657
00:33:24,840 --> 00:33:27,080
Speaker 1: a lot of civilians get killed, right, and we decide

658
00:33:27,120 --> 00:33:29,920
Speaker 1: it's okay. A lot of innocent people get killed for

659
00:33:30,040 --> 00:33:32,400
Speaker 1: military purposes. And so I don't think it's too far

660
00:33:32,480 --> 00:33:37,160
Speaker 1: before AI is making that decision. And then it's AI.

661
00:33:37,240 --> 00:33:41,320
Speaker 1: It's weaponized AI, our weaponized AI versus their weaponized AI.

662
00:33:41,400 --> 00:33:43,560
Speaker 1: And then it's an arms race, and then the most

663
00:33:43,600 --> 00:33:46,320
Speaker 1: powerful army is gonna be the one that just makes

664
00:33:46,320 --> 00:33:48,240
Speaker 1: it all of his decisions, and the generals just say

665
00:33:48,520 --> 00:33:52,560
Speaker 1: defend us right, or respond if we're attacked, right, And

666
00:33:52,600 --> 00:33:55,640
Speaker 1: then you basically handed over control of the weapons to

667
00:33:55,840 --> 00:33:59,080
Speaker 1: the AI because the enemy has weaponized AI. But that

668
00:33:59,120 --> 00:34:01,840
Speaker 1: doesn't mean that they're trilling us. I mean, we use

669
00:34:01,920 --> 00:34:04,760
Speaker 1: them to protect us or to take away some decision

670
00:34:04,760 --> 00:34:07,160
Speaker 1: making for it, but that doesn't mean that they're necessarily

671
00:34:07,280 --> 00:34:09,600
Speaker 1: in control of us. And let's make sure not to

672
00:34:09,600 --> 00:34:11,920
Speaker 1: be too alarmist here, of course, because people are working

673
00:34:12,000 --> 00:34:14,440
Speaker 1: really hard to make sure that there are always ways

674
00:34:14,480 --> 00:34:17,160
Speaker 1: for humans to override these systems. We would be different.

675
00:34:17,280 --> 00:34:19,520
Speaker 1: That would be um, you know, it'd be like if

676
00:34:19,520 --> 00:34:22,680
Speaker 1: a robot then turns the weapons inwards. That's another deal,

677
00:34:22,800 --> 00:34:25,759
Speaker 1: I guess. Yeah. And of course AI researchers do their

678
00:34:25,800 --> 00:34:28,000
Speaker 1: best to make sure that the AI systems are very

679
00:34:28,040 --> 00:34:31,040
Speaker 1: well trained so that they do exactly what we want

680
00:34:31,080 --> 00:34:34,239
Speaker 1: them to do. But they are complex and unpredictable, just

681
00:34:34,280 --> 00:34:37,760
Speaker 1: like people are. Right, So this is a very interesting

682
00:34:37,800 --> 00:34:42,279
Speaker 1: topic whether AI is dangerous or not. And I know

683
00:34:42,400 --> 00:34:45,400
Speaker 1: Daniel that you you're sort of an expert in artificial

684
00:34:45,440 --> 00:34:49,440
Speaker 1: intelligence because you use it in your particle physics research, right,

685
00:34:49,480 --> 00:34:51,839
Speaker 1: you use machine learning, that's right. I wouldn't say I'm

686
00:34:51,840 --> 00:34:53,920
Speaker 1: an expert I mean, I know something about it. Um.

687
00:34:53,960 --> 00:34:55,920
Speaker 1: I've done some reading and I've used it, but I'm

688
00:34:55,960 --> 00:34:59,120
Speaker 1: certainly not a deep expert in artificial intelligence itself, right,

689
00:34:59,200 --> 00:35:02,719
Speaker 1: But you you know experts in your department, right, and

690
00:35:02,760 --> 00:35:04,799
Speaker 1: you're in your campus, that's right. You see, I has

691
00:35:04,840 --> 00:35:08,200
Speaker 1: an amazing computer science department and experts in machine learning.

692
00:35:08,440 --> 00:35:10,480
Speaker 1: Some of the folks I actually collaborate with. When we're

693
00:35:10,520 --> 00:35:13,640
Speaker 1: understanding the huge amounts of data from the large Hagon collider,

694
00:35:13,920 --> 00:35:16,680
Speaker 1: we train machines to sift through that data and like

695
00:35:16,760 --> 00:35:19,520
Speaker 1: look for the Higgs boson and learn to recognize new

696
00:35:19,560 --> 00:35:22,080
Speaker 1: kinds of particles. It's really fun. And these guys know

697
00:35:22,120 --> 00:35:24,279
Speaker 1: a lot about artificial intelligence more than I do. So

698
00:35:24,360 --> 00:35:26,520
Speaker 1: I went over there and I asked them if they

699
00:35:26,560 --> 00:35:29,400
Speaker 1: were worried about whether robots would take over the world,

700
00:35:29,840 --> 00:35:33,439
Speaker 1: and what did the robots say the robots had taken

701
00:35:33,480 --> 00:35:36,760
Speaker 1: over the professors and they answered for no. Um. First,

702
00:35:36,800 --> 00:35:40,760
Speaker 1: here's professor Pierre Baldi, he's a distinguished professor on campus.

703
00:35:40,880 --> 00:35:44,319
Speaker 1: And here's what he had to say. Potentially, yes, all

704
00:35:44,719 --> 00:35:48,680
Speaker 1: very powerful technologies I think can pose such a threat,

705
00:35:49,239 --> 00:35:52,160
Speaker 1: and all depends how they are deployed, how they are used.

706
00:35:52,280 --> 00:35:55,640
Speaker 1: Et cetera. Right, you can say that nuclear technology pulls

707
00:35:55,719 --> 00:35:59,320
Speaker 1: is such a threat and continues to pull such a threat.

708
00:36:00,280 --> 00:36:04,080
Speaker 1: And I think AI, if used in the wrong way

709
00:36:05,000 --> 00:36:09,239
Speaker 1: to pose a threat to mankind. Yes, the potential is

710
00:36:09,280 --> 00:36:12,399
Speaker 1: there and so we should be careful. Um, right, So

711
00:36:12,560 --> 00:36:15,279
Speaker 1: that was Professor Baldy, and then I also went down

712
00:36:15,280 --> 00:36:17,200
Speaker 1: the hall and asked another colleagues cause I thought let's

713
00:36:17,200 --> 00:36:19,880
Speaker 1: get more than one opinion, and so this is Professor

714
00:36:19,920 --> 00:36:22,920
Speaker 1: Park Smith, also a professor of computer science at U

715
00:36:23,000 --> 00:36:26,719
Speaker 1: c Irvine. I think the main threat with artificial intelligence

716
00:36:26,760 --> 00:36:30,479
Speaker 1: going forward is not understanding how the black boxes work.

717
00:36:30,760 --> 00:36:35,600
Speaker 1: And so I think not the typical sort of we're

718
00:36:35,600 --> 00:36:38,400
Speaker 1: going to have robots taking over the world, but more

719
00:36:38,880 --> 00:36:42,719
Speaker 1: the use of AI and situations where we're extrapolating beyond

720
00:36:42,760 --> 00:36:44,719
Speaker 1: what it can do. And so I think we need

721
00:36:44,760 --> 00:36:47,160
Speaker 1: to understand the limits of a I I think that's

722
00:36:47,320 --> 00:36:51,640
Speaker 1: a threat, all right, So the answer is yes, Well,

723
00:36:51,680 --> 00:36:55,279
Speaker 1: I think they're cautious, right, both of them think it's unpredictable.

724
00:36:55,360 --> 00:36:57,360
Speaker 1: We don't know what's going to happen. We're creating a

725
00:36:57,400 --> 00:37:00,239
Speaker 1: whole new kind of system and uh, and we may

726
00:37:00,360 --> 00:37:02,560
Speaker 1: lose control of parts of it. On the other hand,

727
00:37:02,680 --> 00:37:04,920
Speaker 1: you know it's likely for that to happen. You know,

728
00:37:04,960 --> 00:37:06,799
Speaker 1: a lot of people are working really hard to make

729
00:37:06,840 --> 00:37:09,279
Speaker 1: sure that AI will be contained and that in the

730
00:37:09,360 --> 00:37:11,200
Speaker 1: end you can just pull the plug if the robot

731
00:37:11,239 --> 00:37:15,759
Speaker 1: revolution starts, and so it is unpredictable. But also you know,

732
00:37:15,920 --> 00:37:18,399
Speaker 1: the future is unpredictable, is always going to be unpredictable. Yeah,

733
00:37:18,719 --> 00:37:20,959
Speaker 1: I feel like I thought it was interesting he said

734
00:37:21,280 --> 00:37:24,560
Speaker 1: it is dangerous, but not more so than any other

735
00:37:24,840 --> 00:37:28,480
Speaker 1: powerful technology. Yeah, that's a really interesting comment. It's true

736
00:37:28,560 --> 00:37:31,120
Speaker 1: that any technology you can create could be used for

737
00:37:31,160 --> 00:37:34,080
Speaker 1: good or for even if it's powerful, like I mean,

738
00:37:34,120 --> 00:37:37,160
Speaker 1: not just like you know, wind up toy. Maybe it's

739
00:37:37,160 --> 00:37:39,920
Speaker 1: not as dangerous. But but but I think that speaks

740
00:37:39,960 --> 00:37:42,160
Speaker 1: to the kind of the power of AI, Like it

741
00:37:42,239 --> 00:37:46,720
Speaker 1: really is maybe more powerful than we can handle. Yeah,

742
00:37:46,840 --> 00:37:49,080
Speaker 1: and it's it's powerful in a special way run like

743
00:37:49,160 --> 00:37:52,160
Speaker 1: nuclear weapons are powerful. Right, But in the end, a

744
00:37:52,239 --> 00:37:54,720
Speaker 1: human is making that decision, and so you're giving humans

745
00:37:54,760 --> 00:37:57,399
Speaker 1: a new kind of power, which is unpredictable. But here

746
00:37:57,480 --> 00:38:01,160
Speaker 1: you're you're unleashing something, right, You're creating AI, and it's

747
00:38:01,200 --> 00:38:04,080
Speaker 1: making its own decisions. Of course, it's making decisions based

748
00:38:04,080 --> 00:38:05,719
Speaker 1: on what has been told to do. Right, you have

749
00:38:05,760 --> 00:38:08,480
Speaker 1: to give it instructions. Still, you have to teach it um.

750
00:38:08,520 --> 00:38:10,960
Speaker 1: But you can't predict what these complex systems are going

751
00:38:11,000 --> 00:38:13,280
Speaker 1: to do in new circumstances and how they're gonna interpret

752
00:38:13,320 --> 00:38:15,640
Speaker 1: your instructions. And of course there are a lot of

753
00:38:15,640 --> 00:38:17,879
Speaker 1: AI supark people out there working hard to make sure

754
00:38:18,000 --> 00:38:22,000
Speaker 1: that their boundaries and safeties um installed in all AI systems.

755
00:38:22,040 --> 00:38:24,920
Speaker 1: But you know, I've seen Jurassic Park, you know, the

756
00:38:25,000 --> 00:38:28,080
Speaker 1: lesson there they had fences. Lesson there, they had fences.

757
00:38:28,280 --> 00:38:31,560
Speaker 1: We have fences. But then Jeff Goldbloom, you know, has

758
00:38:31,560 --> 00:38:34,520
Speaker 1: a theory about chaos. Yeah, exactly. You know, these systems

759
00:38:34,520 --> 00:38:37,000
Speaker 1: are hard to predict, and so I think we should

760
00:38:37,080 --> 00:38:39,240
Speaker 1: be worried, but then we should respond to that worry

761
00:38:39,320 --> 00:38:42,280
Speaker 1: with appropriate safeguards. You know, we should take this seriously,

762
00:38:42,400 --> 00:38:45,040
Speaker 1: but not be overly alarmed. Right. Well, the other point

763
00:38:45,120 --> 00:38:48,479
Speaker 1: that the other professor made is also interesting that he's

764
00:38:48,480 --> 00:38:51,279
Speaker 1: saying some of the danger is in the fact that

765
00:38:51,880 --> 00:38:54,520
Speaker 1: it's kind of like a black box, like we're trusting

766
00:38:55,000 --> 00:38:58,040
Speaker 1: these things, but we don't really know what's going on inside.

767
00:38:58,280 --> 00:39:02,640
Speaker 1: Like it's so complex they we we can't predict what

768
00:39:02,640 --> 00:39:06,239
Speaker 1: it's gonna do. We can't maybe even deconstruct how it

769
00:39:06,280 --> 00:39:09,560
Speaker 1: makes decisions. That's right, and uh, you know you train

770
00:39:09,640 --> 00:39:11,759
Speaker 1: these systems are very complicated and you don't know how

771
00:39:11,800 --> 00:39:14,600
Speaker 1: they're gonna respond to new circumstances. Right. It's same as

772
00:39:14,640 --> 00:39:16,920
Speaker 1: when like training your dog, Like do you know how

773
00:39:16,960 --> 00:39:18,880
Speaker 1: your dog makes a decision about who to bark end

774
00:39:18,920 --> 00:39:21,120
Speaker 1: who not to. You try to train it, You try

775
00:39:21,160 --> 00:39:22,880
Speaker 1: to give it instructions to try to make sure it

776
00:39:22,920 --> 00:39:24,520
Speaker 1: knows how to how to handle it stuff in a

777
00:39:24,560 --> 00:39:26,920
Speaker 1: new circumstances, but you can't honestly know what it's going

778
00:39:26,960 --> 00:39:29,759
Speaker 1: to do at any given moment. Yeah, I'm definitely not

779
00:39:29,880 --> 00:39:34,400
Speaker 1: visiting your house if you have dogs. I think about

780
00:39:34,920 --> 00:39:38,319
Speaker 1: I think about AI. Again, not an expert, so maybe

781
00:39:38,360 --> 00:39:40,839
Speaker 1: these are uninformed speculations, but I think about AI sort

782
00:39:40,880 --> 00:39:44,520
Speaker 1: of like digital children. You know, like you raise your children,

783
00:39:44,719 --> 00:39:46,840
Speaker 1: you know they're gonna take over one day because you know,

784
00:39:46,920 --> 00:39:48,080
Speaker 1: you and I are going to get old and our

785
00:39:48,160 --> 00:39:50,399
Speaker 1: kids are younger than we are, so eventually they will

786
00:39:50,400 --> 00:39:53,120
Speaker 1: take over and you don't know what they're gonna do,

787
00:39:53,360 --> 00:39:55,200
Speaker 1: and you raise them. You try to raise them in

788
00:39:55,200 --> 00:39:57,640
Speaker 1: a way that they have values they make reasonable decisions,

789
00:39:58,160 --> 00:40:00,000
Speaker 1: and you can sort of think about AI to same

790
00:40:00,000 --> 00:40:03,400
Speaker 1: in way like you try to create this new generation

791
00:40:03,440 --> 00:40:05,480
Speaker 1: of technology that's going to make its own decisions, but

792
00:40:05,480 --> 00:40:07,400
Speaker 1: you try to teach it to make good decisions so

793
00:40:07,400 --> 00:40:10,000
Speaker 1: that when you're in a home, right, it's making good

794
00:40:10,080 --> 00:40:12,359
Speaker 1: choices for you. And I know that some folks out

795
00:40:12,400 --> 00:40:14,400
Speaker 1: there think, well, you know, AI is never really going

796
00:40:14,440 --> 00:40:17,799
Speaker 1: to be separate from humanity. There's not this like cognitive separation,

797
00:40:17,880 --> 00:40:19,759
Speaker 1: Like you can just be part of who you are,

798
00:40:19,880 --> 00:40:22,920
Speaker 1: the way your iPhone feels like part of who you are. Um,

799
00:40:22,960 --> 00:40:26,120
Speaker 1: But we don't know necessarily if if that separation is

800
00:40:26,160 --> 00:40:28,239
Speaker 1: going to be serious, you know, if these things really

801
00:40:28,239 --> 00:40:29,800
Speaker 1: would be separate from us, or if they always just

802
00:40:29,840 --> 00:40:33,480
Speaker 1: feel like an extension of ourselves. Well, until then, I

803
00:40:33,520 --> 00:40:39,960
Speaker 1: think we should stick to regular dogs. Dogs. Yeah, But

804
00:40:40,000 --> 00:40:41,960
Speaker 1: you know, I think about it sometimes the way I

805
00:40:42,000 --> 00:40:44,560
Speaker 1: think about children, right, In the same way that you

806
00:40:44,640 --> 00:40:47,239
Speaker 1: raise your children and they're gonna take over, right, It's

807
00:40:47,239 --> 00:40:49,160
Speaker 1: gonna be some point when your children are in charge.

808
00:40:49,520 --> 00:40:52,200
Speaker 1: You raise them to have values and to make good decisions,

809
00:40:52,200 --> 00:40:54,480
Speaker 1: and you hope that when they take over, they're you know,

810
00:40:54,600 --> 00:40:56,960
Speaker 1: looking after you. In the same way, we got to

811
00:40:56,960 --> 00:40:59,600
Speaker 1: create these digital tools, and we've got to teach them

812
00:40:59,640 --> 00:41:01,440
Speaker 1: to be Hey, we got to teach them what's important,

813
00:41:01,440 --> 00:41:03,719
Speaker 1: and we got to teach them how to be responsible

814
00:41:03,760 --> 00:41:06,160
Speaker 1: so that if they take over, you know that we

815
00:41:06,200 --> 00:41:11,160
Speaker 1: hope they treat as well. Yeah, daddy, good daddy. Your

816
00:41:11,200 --> 00:41:22,480
Speaker 1: parents don't put creator good Please don't bury me underground. Well,

817
00:41:22,520 --> 00:41:24,680
Speaker 1: I personally am looking forward to a time when I have,

818
00:41:24,920 --> 00:41:27,720
Speaker 1: like I don't have to think as much, where life

819
00:41:27,760 --> 00:41:30,440
Speaker 1: is a little bit easier because we have these things

820
00:41:30,520 --> 00:41:33,719
Speaker 1: making things easier for us. It could handle a lot

821
00:41:33,719 --> 00:41:35,960
Speaker 1: of the drudgery and a lot of the logistics. You know,

822
00:41:36,200 --> 00:41:38,839
Speaker 1: eventually you could have a car that drives itself and

823
00:41:38,960 --> 00:41:40,960
Speaker 1: obeys your instructions. You can say like, hey, go pick

824
00:41:41,000 --> 00:41:43,040
Speaker 1: up my kids from school, and he would know how

825
00:41:43,040 --> 00:41:45,560
Speaker 1: to navigate and how to drive and recognize your children

826
00:41:45,960 --> 00:41:48,640
Speaker 1: and how to get back home. And that's totally within

827
00:41:48,680 --> 00:41:51,040
Speaker 1: the realm of possibility in a few years, right, And

828
00:41:51,080 --> 00:41:53,080
Speaker 1: that's pretty awesome. It will offload a lot of work

829
00:41:53,280 --> 00:41:55,920
Speaker 1: and logistics from beleaguered parents. I think you and I

830
00:41:55,920 --> 00:41:58,480
Speaker 1: are in a pretty good position career wise, you know,

831
00:41:58,640 --> 00:42:01,239
Speaker 1: Like I'm a cartoonist in your physicist. These are not

832
00:42:01,680 --> 00:42:04,279
Speaker 1: um jobs that are going to be taken away by

833
00:42:04,280 --> 00:42:08,120
Speaker 1: AI anytime soon. Hopefully have you not seen a our cartoons?

834
00:42:08,200 --> 00:42:11,279
Speaker 1: They're pretty good man, all right, they you should like

835
00:42:11,320 --> 00:42:15,120
Speaker 1: start a podcast instead of wearing of relying on your cartooning. Well,

836
00:42:15,160 --> 00:42:17,960
Speaker 1: there is definitely that as a genre of humor. Like, hey,

837
00:42:18,000 --> 00:42:20,800
Speaker 1: I put um so and so through an AI machine

838
00:42:20,840 --> 00:42:22,759
Speaker 1: and look look at the crazy thing it came out with.

839
00:42:23,400 --> 00:42:25,560
Speaker 1: Except those are all manufactured. None of those are no,

840
00:42:25,920 --> 00:42:29,400
Speaker 1: those are real, None of those are real. Those are

841
00:42:29,440 --> 00:42:33,720
Speaker 1: all made up. Well, that's good for humorist. So artificial

842
00:42:33,719 --> 00:42:37,040
Speaker 1: intelligence is certainly a revolution in thinking and in computing,

843
00:42:37,120 --> 00:42:39,960
Speaker 1: and it will definitely change the world. And so check

844
00:42:40,040 --> 00:42:41,960
Speaker 1: back in in ten years to see if we've been

845
00:42:42,000 --> 00:42:45,880
Speaker 1: replaced by robot Daniel and robot Warhead. Maybe we already

846
00:42:45,880 --> 00:42:49,680
Speaker 1: are bump bump ball. So thanks everyone for listening to

847
00:42:49,719 --> 00:42:53,040
Speaker 1: this episode of Daniel and Jorge Explain the Universe, and

848
00:42:53,080 --> 00:42:56,800
Speaker 1: to listen tomorrow. Just say, Alexa, what's the best science

849
00:42:56,800 --> 00:43:00,600
Speaker 1: podcast in the world. What's the third best? It's not

850
00:43:00,680 --> 00:43:11,640
Speaker 1: a Catherine the world. If you still have a question

851
00:43:11,680 --> 00:43:14,840
Speaker 1: after listening to all these explanations, please drop us a

852
00:43:14,880 --> 00:43:17,120
Speaker 1: line we'd love to hear from you. You can find

853
00:43:17,200 --> 00:43:20,920
Speaker 1: us at Facebook, Twitter, and Instagram at Daniel and Jorge

854
00:43:20,960 --> 00:43:24,400
Speaker 1: That's One Word, or email us at Feedback at Daniel

855
00:43:24,440 --> 00:43:34,600
Speaker 1: and Jorge dot com