Latest / INSIGHT TODAY / 'The turbulent AI era is here'_ Bill Gates gives his three AI risks

Transcript

'The turbulent AI era is here'_ Bill Gates gives his three AI risks · INSIGHT TODAY

• 0:00
  Speaker 1: Warning from Bill Gates about artificial intelligence. The Microsoft co-founder
• 0:03
  Speaker 1: is sounding the alarm in a new essay he's put out,
• 0:06
  Speaker 1: arguing that the AI industry is not admitting the real
• 0:09
  Speaker 1: risk of the technology. And he says it should be, quote,
• 0:11
  Speaker 1: the world's top priority to deal with those risks. He writes,
• 0:15
  Speaker 1: the turbulent AI era is here. The choices we make
• 0:18
  Speaker 1: now are critical. He says AI will either be the
• 0:21
  Speaker 1: greatest equalizer ever invented or the worst source of injustice. Gates,
• 0:26
  Speaker 1: since the world is not preparing for what's to come,
• 0:28
  Speaker 1: I spoke to Gates earlier in an exclusive TV interview. Bill,
• 0:33
  Speaker 1: you said that there's no plan to ease the entry
• 0:36
  Speaker 1: into the AI era and that we underestimate the impact
• 0:40
  Speaker 1: of AI. In what way?
• 0:43
  Speaker 2: Over the last year, these AIs have gotten dramatically more powerful,
• 0:47
  Speaker 2: even faster than I expected. And they're now capable of causing...
• 0:54
  Speaker 2: cyber attack risk, bioterrorism risk, psychosocial risk. And I have
• 0:59
  Speaker 2: to say I'm kind of shocked that the exact criteria
• 1:03
  Speaker 2: that we review these models with and the actions we
• 1:06
  Speaker 2: take to minimize the harms are really completely missing. And
• 1:11
  Speaker 2: the broad dialogue we get all of society talking about
• 1:15
  Speaker 2: the trade-offs here, that's just not taking place. And so,
• 1:19
  Speaker 2: you know, we've got to minimize these harms Through careful
• 1:24
  Speaker 2: review and criteria. And we have to accelerate the good
• 1:28
  Speaker 2: stuff or else I know we could have a complete
• 1:31
  Speaker 2: backlash that AI would suffer from.
• 1:36
  Speaker 1: I talked to Dario Amadei last May. He was estimating
• 1:41
  Speaker 1: 50% of all entry-level white-collar jobs could be eliminated in
• 1:45
  Speaker 1: the next one to five years, 10% to 20% unemployment
• 1:48
  Speaker 1: possibly in the next five years. You write about what
• 1:53
  Speaker 1: you think are the three big risks that come with
• 1:56
  Speaker 1: the transition to AI. Can you just go through them?
• 1:59
  Speaker 2: Yeah, so there's a big risk of people with ill intent,
• 2:05
  Speaker 2: either defrauding people, cyber attacks, or in the extreme case, bioterrorism.
• 2:13
  Speaker 2: So these negative uses, we need to monitor things like
• 2:21
  Speaker 2: creating a dangerous virus to see if that's happening. Another
• 2:27
  Speaker 2: problem will be the job market change, which has not
• 2:31
  Speaker 2: happened yet, but because the AIs are getting so reliable now,
• 2:37
  Speaker 2: they can be used for a variety of well-defined jobs.
• 2:40
  Speaker 2: And that would be a tough thing to adjust to.
• 2:46
  Speaker 2: And finally, just the psychosocial effect that a kid might
• 2:51
  Speaker 2: have all their friends being AIs and not be pushed
• 2:55
  Speaker 2: to deal with real humans and all the learnings and
• 3:00
  Speaker 2: joy you get by being engaged socially. And so every
• 3:05
  Speaker 2: one of those things I'm very concerned and I wanna
• 3:10
  Speaker 2: slow down or eliminate those negative effects.
• 3:14
  Speaker 1: Well, I mean, is that possible? I mean, on the
• 3:16
  Speaker 1: job loss, You know, the counter argument that some people
• 3:20
  Speaker 1: make is, look, you know, at every technological change, there's
• 3:24
  Speaker 1: been job losses, you know, with the electric light bulb,
• 3:28
  Speaker 1: lamp lighters lost their jobs. You in this essay make
• 3:31
  Speaker 1: the argument that a lot of past technological changes, the
• 3:35
  Speaker 1: job losses have have taken place over a longer period
• 3:38
  Speaker 1: of time, even several generations. In some cases, these job
• 3:42
  Speaker 1: losses will come much quicker and all at once.
• 3:46
  Speaker 2: Yes, it's across every sector of the economy, initially with
• 3:52
  Speaker 2: white collar, but then as the humanoid robots come along,
• 3:57
  Speaker 2: blue collar as well, you will have for the first
• 4:00
  Speaker 2: time the equivalent of human cognition. So that past history
• 4:05
  Speaker 2: misleads people because there we were richer and so jobs
• 4:10
  Speaker 2: like designing rides at Disneyland, having more psychiatrists, we were
• 4:14
  Speaker 2: richer enough to create those additional jobs. Here, as the
• 4:19
  Speaker 2: machine gets better than human for well-defined jobs and works
• 4:24
  Speaker 2: 24 hours a day, it's going to be a bigger deal.
• 4:27
  Speaker 2: And so the bar for somebody coming along and saying
• 4:30
  Speaker 2: this time is different is very high. But I'm staking
• 4:34
  Speaker 2: my reputation completely on the fact that in terms of
• 4:39
  Speaker 2: the impact on the job market, this is very different
• 4:43
  Speaker 2: than any technology that I've been involved in or that
• 4:47
  Speaker 2: has come across in human history.
• 4:51
  Speaker 1: How concerned are you about, and do you think society
• 4:54
  Speaker 1: has kind of considered enough, what the impact of just
• 4:58
  Speaker 1: people not having, or machines or computers being able to
• 5:03
  Speaker 1: do jobs much better than humans can? I mean, like
• 5:06
  Speaker 1: what happens to pride in work or human initiative or,
• 5:11
  Speaker 1: I mean, you write about what happens to kids who,
• 5:13
  Speaker 1: and critical thinking, that seems really alarming to me.
• 5:17
  Speaker 2: Yes. You know, the AIs can do unbelievable good, including
• 5:24
  Speaker 2: if you want to learn more, you know, and, you know,
• 5:29
  Speaker 2: have an aid to help make better decisions, analyze data. I'm,
• 5:33
  Speaker 2: you know, a very excited user, but It is such
• 5:37
  Speaker 2: a dramatic change that I really expected this last year,
• 5:42
  Speaker 2: as we crossed the threshold of danger, that society would
• 5:47
  Speaker 2: get involved in this dialogue. Instead, right now, you have
• 5:51
  Speaker 2: people in the industry sort of disagreeing with each other,
• 5:54
  Speaker 2: but to get the right solutions politically, to bring the
• 5:57
  Speaker 2: right expertise to bear, to make it democratically valid, to
• 6:01
  Speaker 2: move as quickly as we need to, it's got to
• 6:04
  Speaker 2: be the academicians, it's got to be more corporate voices,
• 6:10
  Speaker 2: and eventually politicians in both parties listening to these concerns
• 6:16
  Speaker 2: and making sure we don't end up with some sort
• 6:20
  Speaker 2: of just pure black and white stop at all because
• 6:23
  Speaker 2: we don't have the right reviews and thresholds and safeguards.
• 6:30
  Speaker 1: As a parent, it's hard to fully grasp what's going
• 6:33
  Speaker 1: to happen in the next five years, let alone 15.
• 6:36
  Speaker 1: What do you say to parents about, if you had
• 6:40
  Speaker 1: little kids now, how would you be approaching this?
• 6:43
  Speaker 2: Yeah, I spent a few days with the two granddaughters
• 6:47
  Speaker 2: I have, three and one, and it just makes you think,
• 6:53
  Speaker 2: what will education be like? I divide... the challenge here
• 6:58
  Speaker 2: into two periods. We have the next 20 years where
• 7:02
  Speaker 2: the AIs are not able to make all the food
• 7:07
  Speaker 2: and do the manufacturing and maintain all the robots. And
• 7:10
  Speaker 2: so you'll have, once you get to that 20 year period,
• 7:14
  Speaker 2: you'll have deep purpose issues, but you'll have a lot
• 7:18
  Speaker 2: of abundance. I'm more focused in this memo on how
• 7:22
  Speaker 2: do we get through this incredibly turmoil-ridden period we'll have
• 7:28
  Speaker 2: where society's adjusting to this completely unprecedented set of capabilities.
• 7:36
  Speaker 2: And I'm saying, don't think that they're going to stop
• 7:41
  Speaker 2: and not be good at things. There's people who think, oh,
• 7:44
  Speaker 2: they must, you're going to reach some upper bound. For
• 7:49
  Speaker 2: the majority of jobs, that is not the case. And
• 7:52
  Speaker 2: so the silence is what really drove me to speak
• 7:58
  Speaker 2: out at this point, because we have to get everybody's
• 8:02
  Speaker 2: voice in this. You know, these are a set of
• 8:04
  Speaker 2: democratic trade-offs that have to be made. And, you know,
• 8:09
  Speaker 2: people think, okay, I'll block a data center. That really
• 8:11
  Speaker 2: isn't going to slow things down. Those data centers will
• 8:14
  Speaker 2: show up somewhere. We have to take the AI issue
• 8:18
  Speaker 2: and its the bads versus the goods on in a
• 8:21
  Speaker 2: very explicit way.
• 8:23
  Speaker 1: One of the things you write, you said this is
• 8:25
  Speaker 1: either going to be the greatest equalizer ever invented or
• 8:27
  Speaker 1: the worst source of injustice. Which one do you think
• 8:30
  Speaker 1: it is right now, if you had to guess, which
• 8:32
  Speaker 1: one is it going to be?
• 8:33
  Speaker 2: I still believe that if we activate society broadly, both
• 8:39
  Speaker 2: as users, as a lot of experts, what I am
• 8:45
  Speaker 2: saying is that for all of the challenges, there would
• 8:49
  Speaker 2: be a serious review and a criteria under which a
• 8:52
  Speaker 2: model had to meet things like being monitorable for creating
• 8:57
  Speaker 2: bioterrorism tools. And so I still think it's within our
• 9:03
  Speaker 2: grasp to shape this in a net positive way. It's
• 9:08
  Speaker 2: like when nuclear came along, is it nuclear weapons or
• 9:10
  Speaker 2: is it nuclear energy? This is a thousand times bigger
• 9:14
  Speaker 2: and has both the negative aspect and the positive coexisting
• 9:19
  Speaker 2: in the same technology.