1
00:00:02,520 --> 00:00:07,040
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:08,039 --> 00:00:10,639
Speaker 2: Let's stick with tech. The impact of generative AI under

3
00:00:10,680 --> 00:00:14,400
Speaker 2: scrutiny as some investors question massive capex spending in the space.

4
00:00:14,680 --> 00:00:17,840
Speaker 2: Read Halfman, the co founder of Malice AI and author

5
00:00:17,840 --> 00:00:21,040
Speaker 2: of Superagency What could possibly go Right with our AI Future?

6
00:00:21,320 --> 00:00:23,400
Speaker 2: Joined us now read It's good to see you. I

7
00:00:23,440 --> 00:00:25,160
Speaker 2: said as you walked into the room. There's a book

8
00:00:25,200 --> 00:00:26,880
Speaker 2: I need to read What Could Go Right? Because I'm

9
00:00:26,960 --> 00:00:29,680
Speaker 2: very worried about what things could go wrong. Let's put

10
00:00:29,680 --> 00:00:32,160
Speaker 2: it that way. You in the book approach something really

11
00:00:32,200 --> 00:00:36,040
Speaker 2: important as we make these technological revolutions, these advancements over

12
00:00:36,240 --> 00:00:40,440
Speaker 2: human history. Actually it benefits human agency, and I'm worried

13
00:00:40,440 --> 00:00:42,839
Speaker 2: that maybe with AI it guys in a different direction.

14
00:00:43,120 --> 00:00:44,839
Speaker 2: Help me out and tell me why it could go right.

15
00:00:45,120 --> 00:00:47,479
Speaker 3: Well, the shorting answer is it gives you superpowers, right,

16
00:00:47,520 --> 00:00:49,560
Speaker 3: and it gives you superpowers across a large number of things.

17
00:00:49,600 --> 00:00:54,120
Speaker 3: Everyone should go try Chat, GPT, Inflections, Pie and Trapicsquad.

18
00:00:54,720 --> 00:00:56,800
Speaker 3: Give it a try. You can find what should I

19
00:00:56,840 --> 00:00:59,120
Speaker 3: make for dinner? How do I get medical advice that's

20
00:00:59,160 --> 00:01:01,720
Speaker 3: available to me? How I am I to have a

21
00:01:01,760 --> 00:01:04,520
Speaker 3: difficult conversation with a colleague, you know, how might I

22
00:01:05,120 --> 00:01:07,840
Speaker 3: actually do things in parenting that might be different? You know,

23
00:01:07,920 --> 00:01:11,640
Speaker 3: all of those things are are our superpowers that we get.

24
00:01:12,120 --> 00:01:14,760
Speaker 2: And each run. Lisa under the bus, I am, do

25
00:01:14,760 --> 00:01:16,760
Speaker 2: you want to share with the audience what you do?

26
00:01:16,319 --> 00:01:18,840
Speaker 4: You can manipulate what your kids look at if you

27
00:01:18,880 --> 00:01:22,080
Speaker 4: send them links to things that then change their algorithms

28
00:01:22,080 --> 00:01:25,840
Speaker 4: that then give them more news feeds in their in

29
00:01:25,880 --> 00:01:27,280
Speaker 4: their feed. I mean there are ways that you can

30
00:01:27,360 --> 00:01:28,920
Speaker 4: kind of play the system a little bit.

31
00:01:28,840 --> 00:01:29,840
Speaker 1: And I think that that's proper.

32
00:01:29,920 --> 00:01:32,000
Speaker 3: Yeah, exactly. And by the way it helps them learn,

33
00:01:32,160 --> 00:01:34,640
Speaker 3: it gets some news about the environment around them. I mean,

34
00:01:34,840 --> 00:01:36,039
Speaker 3: this is one of the things that is if you're

35
00:01:36,080 --> 00:01:40,000
Speaker 3: not actually in fact using these AI agents to learn

36
00:01:40,080 --> 00:01:42,000
Speaker 3: right now, you should. And like, for example, one of

37
00:01:42,040 --> 00:01:44,360
Speaker 3: the ways that I do when I approach a difficult subject,

38
00:01:44,400 --> 00:01:47,520
Speaker 3: say for example, quantum computing, I stick a technical paper

39
00:01:47,520 --> 00:01:49,440
Speaker 3: in and I say, explain this to me like I'm

40
00:01:49,440 --> 00:01:52,360
Speaker 3: twelve superpowers superpowers.

41
00:01:52,480 --> 00:01:54,480
Speaker 2: Yes, okay, So my next question is going to be

42
00:01:54,680 --> 00:01:57,800
Speaker 2: how should we participate and how we shake this intelligence?

43
00:01:57,800 --> 00:01:59,520
Speaker 2: And Lisa has given us one idea, could you give

44
00:01:59,560 --> 00:02:00,000
Speaker 2: us another time?

45
00:02:00,640 --> 00:02:03,960
Speaker 3: Well, I think part of the key thing is my

46
00:02:04,080 --> 00:02:07,360
Speaker 3: very first chapter is about chat GBT, where humanity enters

47
00:02:07,360 --> 00:02:09,639
Speaker 3: the chat. You got hundreds of millions of people who

48
00:02:09,639 --> 00:02:13,200
Speaker 3: are interacting with it, and the company watches what works

49
00:02:13,200 --> 00:02:15,280
Speaker 3: and what doesn't work and then improves it. So by

50
00:02:15,480 --> 00:02:18,720
Speaker 3: simply going and experimenting and getting which kind of things

51
00:02:18,760 --> 00:02:22,600
Speaker 3: are the things that help you elevate your agency, then

52
00:02:23,040 --> 00:02:27,119
Speaker 3: actually in fact helps the company know, oh, these things

53
00:02:27,160 --> 00:02:29,160
Speaker 3: really work, Oh these things need to be improved.

54
00:02:29,520 --> 00:02:32,480
Speaker 1: At the same time, you talk about technical papers, and

55
00:02:32,560 --> 00:02:35,280
Speaker 1: one thing that AI does really well is work with

56
00:02:35,320 --> 00:02:38,919
Speaker 1: technical details, not so much, the social not so much.

57
00:02:38,919 --> 00:02:41,560
Speaker 1: When you have to have certain biases that are baked

58
00:02:41,639 --> 00:02:45,000
Speaker 1: into some of these algorithms, what are the limitations to

59
00:02:45,080 --> 00:02:47,520
Speaker 1: this in sort of broad raised application at a time

60
00:02:47,560 --> 00:02:51,040
Speaker 1: where people are saying this could do qualitative jobs and

61
00:02:51,080 --> 00:02:54,240
Speaker 1: there could potentially be a lot of biases baked in.

62
00:02:54,800 --> 00:02:59,920
Speaker 3: So every major group is working intensely on biases. And actually,

63
00:03:00,000 --> 00:03:02,600
Speaker 3: one of the things that we did with inflection and

64
00:03:02,800 --> 00:03:06,320
Speaker 3: PIE personal intelligence PUN intended was to teach it to

65
00:03:06,360 --> 00:03:10,600
Speaker 3: be kind and empathetic, to have EQ as important as IQ.

66
00:03:11,200 --> 00:03:13,440
Speaker 3: And actually you see that now in anthropics Claude and

67
00:03:13,440 --> 00:03:15,760
Speaker 3: other things is it's kind of spreading, and so you

68
00:03:15,800 --> 00:03:19,120
Speaker 3: can actually have these kinds of interactions be trained in them,

69
00:03:19,200 --> 00:03:21,760
Speaker 3: and this is kind of reinforcement learning by human feedback

70
00:03:22,040 --> 00:03:24,600
Speaker 3: to actually be much better than your average human.

71
00:03:24,760 --> 00:03:27,040
Speaker 1: The reason why I find this so fascinating is because

72
00:03:27,600 --> 00:03:29,960
Speaker 1: I think about my children and what careers they potentially

73
00:03:30,000 --> 00:03:32,480
Speaker 1: could have, and I'm like, no, not programmer, because that's

74
00:03:32,480 --> 00:03:34,200
Speaker 1: going to be dead. You know, all of these other ones.

75
00:03:34,200 --> 00:03:37,320
Speaker 1: Probably if you can be empathetic human, those human skills

76
00:03:37,360 --> 00:03:39,360
Speaker 1: are probably going to be more needed. That's what people say.

77
00:03:39,480 --> 00:03:40,720
Speaker 1: Are you saying those two are going to go out

78
00:03:40,720 --> 00:03:41,000
Speaker 1: the waydo?

79
00:03:41,000 --> 00:03:41,200
Speaker 2: I mean?

80
00:03:41,200 --> 00:03:42,280
Speaker 1: What is this due to employment?

81
00:03:42,520 --> 00:03:46,720
Speaker 3: So first, I actually think programming skill jobs aren't going

82
00:03:46,720 --> 00:03:48,840
Speaker 3: out the door. I think there's human application, but I

83
00:03:48,840 --> 00:03:52,040
Speaker 3: think it's also human application of like EQ as well.

84
00:03:52,080 --> 00:03:55,440
Speaker 3: So the fact that you have a for example, say

85
00:03:55,440 --> 00:03:57,960
Speaker 3: a medical assistant that can help you with something, or

86
00:03:58,000 --> 00:03:59,720
Speaker 3: a coach that might be able to talk with something,

87
00:04:00,000 --> 00:04:02,320
Speaker 3: doesn't mean you don't want a human coach. It's more

88
00:04:02,360 --> 00:04:05,920
Speaker 3: coaches the better, right, and being able to make it work.

89
00:04:06,160 --> 00:04:08,480
Speaker 5: You're talking about AI for the good, but what about

90
00:04:08,480 --> 00:04:10,080
Speaker 5: AI in the hands of adversaries?

91
00:04:11,400 --> 00:04:16,160
Speaker 3: Well, so superpowers superpowers for terrorists, superpowers for criminals, superpowers

92
00:04:16,200 --> 00:04:18,000
Speaker 3: for rogues states. That's one of the things that we

93
00:04:18,080 --> 00:04:20,760
Speaker 3: need to be careful about how we navigate. Once again,

94
00:04:21,120 --> 00:04:23,720
Speaker 3: all of the major groups actually have security groups that

95
00:04:23,800 --> 00:04:25,520
Speaker 3: are working on this to try to make sure and

96
00:04:25,839 --> 00:04:28,400
Speaker 3: that's I think part of the having call so called

97
00:04:28,480 --> 00:04:30,960
Speaker 3: red teaming plans and kind of trying to make sure

98
00:04:30,960 --> 00:04:33,080
Speaker 3: that that's as minimized as much as possible.

99
00:04:33,160 --> 00:04:35,040
Speaker 5: How do think the US government should be approaching this?

100
00:04:35,320 --> 00:04:38,160
Speaker 5: The market's freaked out because of deep seek and how

101
00:04:38,279 --> 00:04:41,640
Speaker 5: much potentially the CCP is putting money into their version

102
00:04:41,760 --> 00:04:44,360
Speaker 5: of AI. But when you're in China and you want

103
00:04:44,360 --> 00:04:46,560
Speaker 5: to look at things on deep seek that are particularly

104
00:04:46,600 --> 00:04:49,240
Speaker 5: critical to the CCP, it will not show you. So

105
00:04:49,279 --> 00:04:51,280
Speaker 5: how does the US government go about this?

106
00:04:51,880 --> 00:04:54,440
Speaker 3: Well, So I think it's important that the US government

107
00:04:55,440 --> 00:04:57,880
Speaker 3: kind of continues some of the threads that the previous

108
00:04:57,920 --> 00:05:02,039
Speaker 3: executive order did, which is you have red teaming things

109
00:05:02,080 --> 00:05:04,919
Speaker 3: that you actually do world leadership about what kind of

110
00:05:04,920 --> 00:05:07,480
Speaker 3: things are happening. I mean, there's connections in the UK

111
00:05:08,480 --> 00:05:10,359
Speaker 3: AI Safety and Suit and the US one and the

112
00:05:10,360 --> 00:05:12,479
Speaker 3: French one, and I think these are important to do.

113
00:05:13,560 --> 00:05:15,120
Speaker 3: But I think that this is one of the things

114
00:05:15,120 --> 00:05:17,080
Speaker 3: where there will be multiple AIS and part of the

115
00:05:17,120 --> 00:05:20,080
Speaker 3: reason why I want AI not to just be artificial

116
00:05:20,080 --> 00:05:24,159
Speaker 3: intelligence or amplification intelligence, but also American intelligence as part

117
00:05:24,160 --> 00:05:24,760
Speaker 3: of what we're doing.

118
00:05:24,960 --> 00:05:28,000
Speaker 5: What do you make then, of the tech industry being

119
00:05:28,160 --> 00:05:30,800
Speaker 5: very close to this administration. I'm not just talking about

120
00:05:30,800 --> 00:05:32,320
Speaker 5: your former front Elon Musk. I was at the White

121
00:05:32,360 --> 00:05:34,640
Speaker 5: House yesterday and Bill Gates was walking into the West

122
00:05:34,640 --> 00:05:37,600
Speaker 5: wing for meetings. Do you think they can help the

123
00:05:37,640 --> 00:05:42,400
Speaker 5: administration basically get to there faster when it comes to

124
00:05:42,440 --> 00:05:43,080
Speaker 5: safe AI?

125
00:05:44,080 --> 00:05:47,039
Speaker 3: In the short absolutely, I think it's if you said,

126
00:05:47,200 --> 00:05:50,760
Speaker 3: of any kind of Western democracy government the world, most especially,

127
00:05:50,760 --> 00:05:54,040
Speaker 3: of course, you know, American presidency, American government, should they

128
00:05:54,040 --> 00:05:56,200
Speaker 3: be connected with the tech industry talking about how tech

129
00:05:56,320 --> 00:05:57,960
Speaker 3: creating the future? What are the things that we can

130
00:05:58,000 --> 00:06:00,920
Speaker 3: do to make better American industry, better American society, better

131
00:06:00,960 --> 00:06:03,839
Speaker 3: lives for American citizens? The short answer is absolutely.

132
00:06:04,160 --> 00:06:06,080
Speaker 2: What you make in the back lash against Dale Musk

133
00:06:06,160 --> 00:06:09,320
Speaker 2: more recently about he's in film from the government, well.

134
00:06:09,680 --> 00:06:13,159
Speaker 3: In government, the move fast and break things maybe should

135
00:06:13,160 --> 00:06:16,000
Speaker 3: be you know, a little bit more compassionate and judicious.

136
00:06:16,360 --> 00:06:17,600
Speaker 2: Is that what you think the apprice should be.

137
00:06:18,360 --> 00:06:21,480
Speaker 3: I think that the question should be is trying to say, hey, look,

138
00:06:21,800 --> 00:06:24,440
Speaker 3: there's a reason government doesn't work like companies, even though

139
00:06:24,480 --> 00:06:26,880
Speaker 3: I think there's efficiency that's really good and we need

140
00:06:26,920 --> 00:06:28,200
Speaker 3: to respect those reasons too.

141
00:06:28,600 --> 00:06:32,360
Speaker 5: Do you feel that you're going to basically get some

142
00:06:32,360 --> 00:06:35,840
Speaker 5: sort of repudiation because of where your politics were ahead

143
00:06:35,880 --> 00:06:39,120
Speaker 5: of this election and given how Trump has basically cosied

144
00:06:39,240 --> 00:06:40,960
Speaker 5: up to a lot of individuals in the tech world.

145
00:06:41,000 --> 00:06:42,800
Speaker 5: I have not seen you yet at mar Lago in

146
00:06:42,880 --> 00:06:44,880
Speaker 5: inauguration or walking into the West wag.

147
00:06:46,480 --> 00:06:49,560
Speaker 3: Look, I think the thing that we most want as

148
00:06:49,560 --> 00:06:52,839
Speaker 3: American citizens is for our government to succeed, to succeed

149
00:06:52,880 --> 00:06:56,440
Speaker 3: for our industry, succeed for our citizens, And so that's

150
00:06:56,480 --> 00:06:59,039
Speaker 3: what I most want, and I think we all do.

151
00:06:59,520 --> 00:07:02,279
Speaker 1: You have been an incredible investor. You are an early

152
00:07:02,320 --> 00:07:06,320
Speaker 1: investor in open Ai, You invested in LinkedIn, just so

153
00:07:06,440 --> 00:07:09,479
Speaker 1: many of the behemoths currently. Do you see a lot

154
00:07:09,480 --> 00:07:12,240
Speaker 1: of upstarts right now that look like they could become

155
00:07:12,800 --> 00:07:16,680
Speaker 1: the next big thing in some of this machine learning technology?

156
00:07:17,800 --> 00:07:21,400
Speaker 3: So the short answer is yes, although picking them is

157
00:07:21,400 --> 00:07:23,440
Speaker 3: one of the difficult things about being a venture capitalist,

158
00:07:23,480 --> 00:07:27,240
Speaker 3: and there's a lot of different options. Back a few

159
00:07:27,280 --> 00:07:29,800
Speaker 3: years ago, I was arguing that we were five large

160
00:07:29,800 --> 00:07:32,720
Speaker 3: tech companies heading to ten. But I wouldn't necessarily have

161
00:07:32,720 --> 00:07:35,840
Speaker 3: predicted that Nvidio would have so quickly become one of those,

162
00:07:36,040 --> 00:07:38,360
Speaker 3: you know, into the seven and then going to ten.

163
00:07:38,400 --> 00:07:40,520
Speaker 3: And so I think that's we are in that progress.

164
00:07:40,560 --> 00:07:42,480
Speaker 3: And I think there are a number of startups that

165
00:07:42,560 --> 00:07:45,480
Speaker 3: are creating amazing technology. And then the question is product

166
00:07:45,480 --> 00:07:46,440
Speaker 3: market fit at scale?

167
00:07:46,600 --> 00:07:49,200
Speaker 1: What do you think the next technological evolution is going

168
00:07:49,240 --> 00:07:52,160
Speaker 1: to be? They could potentially be a sort of chat

169
00:07:52,200 --> 00:07:55,240
Speaker 1: GPT moment, or a deep seek moment, or one of

170
00:07:55,280 --> 00:07:57,640
Speaker 1: these sort of breakthrough ahaz.

171
00:07:58,600 --> 00:08:00,160
Speaker 3: One of the things I think we're going to to

172
00:08:00,200 --> 00:08:04,560
Speaker 3: see this year, all of the major AI companies are

173
00:08:04,600 --> 00:08:07,760
Speaker 3: working on amplifying coding. And one of the things like,

174
00:08:07,800 --> 00:08:10,320
Speaker 3: for example, you know for your kids, other things will

175
00:08:10,360 --> 00:08:13,920
Speaker 3: be yes, you have to bring up the UH will

176
00:08:13,960 --> 00:08:17,400
Speaker 3: be actually all of us will have a software engineering

177
00:08:17,440 --> 00:08:19,640
Speaker 3: assistant and that will make all of our work a

178
00:08:19,680 --> 00:08:21,680
Speaker 3: lot better. UH. And that's one of the things I

179
00:08:21,680 --> 00:08:23,800
Speaker 3: think hasn't yet entered the consciousness. So I think that's

180
00:08:23,840 --> 00:08:26,200
Speaker 3: one of the things that's coming. UH. And then of course,

181
00:08:26,320 --> 00:08:28,520
Speaker 3: you know, with what I'm doing with manus I'm hopeful

182
00:08:28,560 --> 00:08:31,840
Speaker 3: that we will have some really great early results in

183
00:08:32,480 --> 00:08:33,520
Speaker 3: hearing cancer.

184
00:08:33,320 --> 00:08:36,000
Speaker 2: And how we can accelerate the scientific process. I say,

185
00:08:36,040 --> 00:08:38,640
Speaker 2: we're all up with you on that page. You're very successful.

186
00:08:38,679 --> 00:08:40,720
Speaker 2: Thank you, sir, read to appreciate your time. Thank you

187
00:08:40,760 --> 00:08:42,839
Speaker 2: so good to see you. Ride half and there manus

188
00:08:42,920 --> 00:08:44,319
Speaker 2: Ai and a whole lot more