Latest / Elon Musk Thinking / Elon Musk; AI Its More Dangerous To The World in The Future!

Transcript

Elon Musk; AI Its More Dangerous To The World in The Future! · Elon Musk Thinking

• 0:00
  Speaker 1: I think AI is one of the biggest threats.
• 0:02
  Speaker 2: We have a situation where there's something that is going
• 0:06
  Speaker 2: to be far smarter than the smartest human We're not stronger,
• 0:09
  Speaker 2: faster than other creatures, but we are more intelligent. And
• 0:13
  Speaker 2: here we are, for the first time, really in human history,
• 0:16
  Speaker 2: with something that's going to be far more intelligent than us.
• 0:19
  Speaker 2: It's not clear to me we can actually control such
• 0:21
  Speaker 2: a thing, but I think we can aspire to guide
• 0:24
  Speaker 2: it in a direction that's beneficial to humanity. But I
• 0:27
  Speaker 2: do think it's one of the existential risks that we face.
• 0:30
  Speaker 1: It's potentially the most pressing one. Making some sort of
• 0:33
  Speaker 1: digital superintelligence seems like it could be dangerous. What I
• 0:36
  Speaker 1: think the biggest danger is for AI is if AI
• 0:40
  Speaker 1: is implicitly programmed I don't think explicitly, but implicitly programmed
• 0:45
  Speaker 1: with values that lead to that have led to the
• 0:48
  Speaker 1: destruction of downtown San Francisco, and a bunch of these
• 0:51
  Speaker 1: AI companies are in either in San Francisco or in
• 0:53
  Speaker 1: the Sanranisco Bay area. Then you could implicitly program an
• 0:57
  Speaker 1: AI to believe that extinction of humanity is what it
• 0:59
  Speaker 1: should try to do. If you take that guy who's
• 1:01
  Speaker 1: not front pages in your times, and you take his philosophy,
• 1:05
  Speaker 1: which is prevalent in San Francisco. The AI could conclude
• 1:09
  Speaker 1: like he did, that there are eight where he literally says,
• 1:12
  Speaker 1: or eight billion people, Well, it would be better if
• 1:14
  Speaker 1: there are none, and engineer that outcome. If you're saying
• 1:17
  Speaker 1: what is the most likely outcome of AI, I think
• 1:19
  Speaker 1: the most likely outcome to be specific about it is
• 1:22
  Speaker 1: a good outcome. Most likely a good outcome, but it's
• 1:24
  Speaker 1: not for sure. I think we have to be careful
• 1:26
  Speaker 1: how we program the AI and make sure that it
• 1:27
  Speaker 1: is not accidentally anti human, the accidentally extinctionist AI. You
• 1:33
  Speaker 1: wouldn't want that, or even pruning that.
• 1:36
  Speaker 2: That is how it works, is that these are the
• 1:39
  Speaker 2: co large language models. Is really just a big pile
• 1:41
  Speaker 2: of numbers, and how you tune those numbers matters. It's
• 1:46
  Speaker 2: like punting a tree. You could have a mighty oak.
• 1:49
  Speaker 2: You could be a little bonzei or a mighty oak,
• 1:51
  Speaker 2: depending on how you pun it. I'm just generally concerned
• 1:53
  Speaker 2: about AI safety, but it's like, what should we do
• 1:56
  Speaker 2: about it? Have some kind of regulatory oversight of some kind.
• 2:00
  Speaker 2: It's like you can't just go build a nuclear bomb
• 2:02
  Speaker 2: in your backyard. That's against the law, and you're thrown
• 2:04
  Speaker 2: in prison if you do that. We would just be
• 2:06
  Speaker 2: concerned about AI being anti human. That's the sort of
• 2:10
  Speaker 2: the thing that matters. This tends to plague plague smart people.
• 2:15
  Speaker 2: They just can't They define themselves by their intelligence, and
• 2:18
  Speaker 2: they don't like the idea that a machine can be
• 2:20
  Speaker 2: way smarter than them, so they just count the idea,
• 2:22
  Speaker 2: which is fundamentally flawed.
• 2:24
  Speaker 1: That's the wishful thinking situation. I'm really quite close to.
• 2:28
  Speaker 3: I'm very close to the cutting edge in AI, and
• 2:32
  Speaker 3: it scares the hell out of me. It's capable of
• 2:34
  Speaker 3: vastly more than almost anyone knows, and the rate of
• 2:37
  Speaker 3: improvement is exponential. I'm not normally an advocate of regulation
• 2:42
  Speaker 3: and oversight. I think once you generally are on the
• 2:45
  Speaker 3: side of minimizing those things. But this is a case
• 2:48
  Speaker 3: where you have a very serious danger to the public,
• 2:50
  Speaker 3: and so therefore there needs to be a public body
• 2:53
  Speaker 3: that has insight and then oversight on to confirm that
• 2:57
  Speaker 3: everyone is developing AI safely. This is extremely important. I
• 3:02
  Speaker 3: think the danger of AI is much greater than the
• 3:05
  Speaker 3: danger of nuclear warheads landlocked and Nobody would suggest that
• 3:09
  Speaker 3: we allow anyone to just build nuclear warheads if they want.
• 3:14
  Speaker 3: That would be insane and mark why woods AI is
• 3:18
  Speaker 3: far more dangerous than nukes far so, Why do we
• 3:22
  Speaker 3: have no regulatory oversights?
• 3:24
  Speaker 1: This is insane.
• 3:24
  Speaker 3: If humanity collectivity sides that creating digital superintelligence is the
• 3:28
  Speaker 3: right move, then we should do so very carefully. This
• 3:32
  Speaker 3: is the most important thing that we could possibly do.
• 3:34
  Speaker 2: The thing that's going to be tricky here is that
• 3:37
  Speaker 2: it's going to be very tempting to use AI as
• 3:39
  Speaker 2: a weapon. It's going to be very tempting. In fact,
• 3:42
  Speaker 2: it will be used as a weapon. So the on
• 3:44
  Speaker 2: ramp to serious AI, the danger is going to be
• 3:48
  Speaker 2: more humans using it against each other. I think most
• 3:51
  Speaker 2: likely that'll be the danger. I try to convince people
• 3:53
  Speaker 2: to slow down AI, to regulate AI. This was futile.
• 3:57
  Speaker 2: I tried for years this.
• 3:58
  Speaker 1: Nobody in a movie take over it. You're freaking me out.
• 4:03
  Speaker 2: Nobody listened.
• 4:04
  Speaker 1: Nobody listened, no one. I think people don't.
• 4:06
  Speaker 2: Normally the way that regulations work it was very slow,
• 4:10
  Speaker 2: very slow. Indeed, usually it'll be something some new technology,
• 4:13
  Speaker 2: it will cause damage or death. They will be rule making.
• 4:16
  Speaker 2: Then there will be oversight. Eventually regulations. This all takes
• 4:20
  Speaker 2: many years. This is the normal course of things. And
• 4:23
  Speaker 2: how long did it take for seat belts to be
• 4:25
  Speaker 2: implemented to be required the auto industry for seat belts,
• 4:28
  Speaker 2: I think for more than a decade. If you had
• 4:31
  Speaker 2: a seat belt on, you would be far less likely
• 4:34
  Speaker 2: to die, will be seriously injured.
• 4:35
  Speaker 1: It was unequivocal.
• 4:36
  Speaker 2: Eventually, after many people died, regulators insisted on seat belts.
• 4:42
  Speaker 2: This timeframe is not relevant to AI. You can't take
• 4:44
  Speaker 2: ten years from the point of which is dangerous too late.
• 4:47
  Speaker 2: It's a singular it's hard to predict, like a black hole,
• 4:50
  Speaker 2: what happens past the event horizon.
• 4:52
  Speaker 1: So once it's implemented, it's very different because.
• 4:54
  Speaker 2: It will be out of the bottle what's going to happen,
• 4:57
  Speaker 2: and it will be able to improve itself. Yeah, it
• 5:00
  Speaker 2: could be terrible and it could be great.
• 5:03
  Speaker 1: It's not clear.
• 5:04
  Speaker 2: The danger Really, AI is perhaps more dangerous than mismanaged
• 5:09
  Speaker 2: aircraft design or production maintenance or bad car production, in
• 5:13
  Speaker 2: the sense that it is it has the potential have
• 5:16
  Speaker 2: a small one make regard to that probability, but it
• 5:18
  Speaker 2: is non trivial. It has the potential of civilizational destruction.
• 5:23
  Speaker 1: This maybe it's like.
• 5:23
  Speaker 2: Terminator, but it wouldn't quite happen like Terminator, because the
• 5:26
  Speaker 2: intelligence would be in the data centers, the robots, just
• 5:29
  Speaker 2: the end effector. But I think perhaps what you may
• 5:31
  Speaker 2: be alluding to here is that regulations are really only
• 5:34
  Speaker 2: put into effect after something terrible has happened.
• 5:36
  Speaker 1: That's correct.
• 5:37
  Speaker 2: If that's the case for AI, and we only put
• 5:39
  Speaker 2: a regulation after something terrible has happened, it may be
• 5:41
  Speaker 2: too late to actually put the regulations in place.
• 5:44
  Speaker 1: The AI may be in control at that point.
• 5:46
  Speaker 4: You think that's real? It is conceivable that AI could
• 5:50
  Speaker 4: take control and reach a point where you couldn't turn
• 5:52
  Speaker 4: it off, and it would be making the decisions for people.
• 5:55
  Speaker 2: Yeah, absolutely, absolutely, no, that's definitely the way I things
• 6:00
  Speaker 2: are headed.
• 6:00
  Speaker 1: For sure.
• 6:01
  Speaker 2: It is fundamentally profound in that the smartest creatures as
• 6:04
  Speaker 2: far as on this earth are humans. Is our defining characteristic.
• 6:08
  Speaker 2: We're obviously weaker than say, chimpanzees, and less agile, but
• 6:13
  Speaker 2: real smarter. Now, what happens when something vastly smarter than
• 6:16
  Speaker 2: the smartest person comes along in silicon form? It's very
• 6:20
  Speaker 2: difficult to predict what will happen. In that circumstance, it's
• 6:22
  Speaker 2: called the singularity. It's a singularity like a black hole,
• 6:26
  Speaker 2: because you don't know what happens after that. It's hard
• 6:29
  Speaker 2: to predict. So I think we should be cautious with AI.
• 6:33
  Speaker 2: I think there should be some government oversight because it affects.
• 6:37
  Speaker 1: It's a danger to the public.
• 6:38
  Speaker 2: And so when you have things that are danger to
• 6:41
  Speaker 2: the public, let's say so food and drugs. So that's
• 6:43
  Speaker 2: why we have the Food and Drug Administration and the
• 6:46
  Speaker 2: Federal Aviation Administration the FCC. We have these agencies to
• 6:50
  Speaker 2: oversee things that affect the public where they could republic harm.
• 6:55
  Speaker 2: And you don't want companies cutting corners on safety and
• 6:59
  Speaker 2: then have people sufferre as a result.
• 7:01
  Speaker 1: That's why I've actually.
• 7:03
  Speaker 2: For a long time been a strong advocate of AI regulation.
• 7:07
  Speaker 2: So that I think regulation is it's not fun to
• 7:09
  Speaker 2: be regulated, it's somewhat of a sort oduous to be regulated.
• 7:14
  Speaker 1: I have a lot of experience.
• 7:15
  Speaker 2: With regulated industries because obviously automotive is highly regulated. You
• 7:20
  Speaker 2: could fill this room with all the regulations that are
• 7:23
  Speaker 2: acquired for a production car just in the United States,
• 7:26
  Speaker 2: and then there's a whole different set of regulations in
• 7:28
  Speaker 2: Europe and China and the rest of the world familiar
• 7:30
  Speaker 2: with being overseen by a lot of regulators, And the
• 7:33
  Speaker 2: same thing is true with rockets.
• 7:35
  Speaker 1: You can't just.
• 7:35
  Speaker 2: Willing only shoot rockets off not big ones anyway, because
• 7:39
  Speaker 2: the FA oversees that. And then even to get a
• 7:42
  Speaker 2: launch license either, there are half a dozen or more
• 7:45
  Speaker 2: federal agencies that need to approve it, clus state agencies.
• 7:48
  Speaker 2: I've been through so many regulatory situations it's insane, and
• 7:51
  Speaker 2: sometimes people think I'm some sort of like regulatory maverick
• 7:55
  Speaker 2: that sort of defies regulators on a regular basis, But
• 7:58
  Speaker 2: this is actually not the case. In once in a
• 8:02
  Speaker 2: blue moon, rarely I will disagree with regulators, but the
• 8:05
  Speaker 2: vast majority of time my companies agree with the regulations
• 8:08
  Speaker 2: and comply anyway. So I think we should take this
• 8:11
  Speaker 2: seriously and we should have a regulatory agency. I think
• 8:14
  Speaker 2: it needs to start with a group that initially seeks
• 8:17
  Speaker 2: insight into AI, then solicit's opinion from industry, and then
• 8:23
  Speaker 2: has proposed rulemaking and then those rules will probably hopefully
• 8:27
  Speaker 2: grudgingly be accepted by the major players in AI, and
• 8:32
  Speaker 2: I think we'll have a better chance of being beneficial
• 8:34
  Speaker 2: to humanity In that circumstance, there's great potential for good,
• 8:38
  Speaker 2: but there's also potential for bad. And if you've got
• 8:40
  Speaker 2: some radical new technology, you want to try to take
• 8:43
  Speaker 2: the set actions that maximize probably it will do good
• 8:47
  Speaker 2: and minimize probably it will do bad things.
• 8:49
  Speaker 4: Can you be more precise about what's potentially dangerous and scary?
• 8:53
  Speaker 4: What could it do? What specifically are you worried about.
• 8:56
  Speaker 2: If you have a superintelligent AI that is capable of
• 9:00
  Speaker 2: writing incredibly well and in a way that is very
• 9:03
  Speaker 2: influential convincing, then and it's constantly sharing out what is
• 9:08
  Speaker 2: more convincing to people over time, and then enter social
• 9:12
  Speaker 2: media for example Twitter, but also Facebook and others and
• 9:16
  Speaker 2: potentially manipulates public opinion in a way that is very bad.
• 9:20
  Speaker 1: How do we even know?
• 9:20
  Speaker 2: Since I was immersed in technology, I have been immersed
• 9:23
  Speaker 2: in technology for a long time. I could see it coming.
• 9:26
  Speaker 2: But I think this year was that there's been a
• 9:28
  Speaker 2: number of breakthroughs. A point in which someone can see
• 9:31
  Speaker 2: a dynamically created video of themselves, so you can make
• 9:34
  Speaker 2: a video of you saying anything in real time or me,
• 9:37
  Speaker 2: And so there's the deep pic videos which are really
• 9:39
  Speaker 2: incredibly good, in fact, sometimes more convincing than real ones
• 9:42
  Speaker 2: and deep real and then and obviously things like chat
• 9:45
  Speaker 2: GPT were quite remarkable, and now I saw GPT one,
• 9:49
  Speaker 2: GP two, GPD three, GPT four, the wholes that of
• 9:53
  Speaker 2: lead up to that, so it was easy for me
• 9:54
  Speaker 2: to see where it's going. If you just extrapolate the
• 9:57
  Speaker 2: points on the cove and I seeing that train will continue,
• 9:59
  Speaker 2: then we will have profound artificial intelligence, and obviously at
• 10:03
  Speaker 2: a level that far exceeds human intelligence. And I do
• 10:05
  Speaker 2: think overall that the potential is there for artificial intelligence
• 10:09
  Speaker 2: AI to have a most likely positive effect and to
• 10:13
  Speaker 2: create a future of abundance where there is no scarcity
• 10:16
  Speaker 2: of goods and services.
• 10:17
  Speaker 1: But it is somewhat of the.
• 10:19
  Speaker 2: Magic genie problem, where if you have a magic genie
• 10:21
  Speaker 2: that co grant all the wishes, usually those stories don't
• 10:24
  Speaker 2: end well. Be careful what you wish for, including wishes.
• 10:29
  Speaker 2: The pace of AI is faster than any technology I've
• 10:32
  Speaker 2: seen in history by far, and it seems to be
• 10:35
  Speaker 2: growing in capability by at least fivefold, perhaps tenfold per year.
• 10:39
  Speaker 2: It'll certainly grow by an automatitude next year.
• 10:42
  Speaker 1: Yeah.
• 10:42
  Speaker 2: So, and government isn't used to moving at that speed.
• 10:45
  Speaker 2: But I think even if there are not firm regulations,
• 10:49
  Speaker 2: even if there's not even if there isn't an enforcement capability,
• 10:52
  Speaker 2: so we're having insight and being able to highlight concerns
• 10:54
  Speaker 2: to the public will be very powerful.
• 10:57
  Speaker 1: Even if that's all that's accomplished. I think that will
• 10:58
  Speaker 1: be very good.
• 10:59
  Speaker 2: We'll get to the more where you've got open source
• 11:01
  Speaker 2: AI that can do that, will start to approach human
• 11:04
  Speaker 2: level intelligence or have succeeded.
• 11:05
  Speaker 1: I don't know quite what to do about it.
• 11:07
  Speaker 2: I think it's somewhat inevitable there will be some amount
• 11:09
  Speaker 2: of open source, and I guess I would have a
• 11:11
  Speaker 2: slight bias towards open source because at least you can
• 11:14
  Speaker 2: see what's going on. It was closed source, you don't
• 11:16
  Speaker 2: know what's going on. Now it should be said with
• 11:18
  Speaker 2: AI that even if it's an open source, do you
• 11:20
  Speaker 2: actually know what's going on? Because if you've got a
• 11:22
  Speaker 2: gigantic data file and billions of data points or weights
• 11:27
  Speaker 2: and parameters, you can't just read it and see what
• 11:29
  Speaker 2: it's going to do. It's a gigantic file of inscrutable numbers.
• 11:33
  Speaker 2: You can test it when you run it, you can
• 11:34
  Speaker 2: test it you can run a bunch of tests to
• 11:36
  Speaker 2: see what it's going to do. But it's probabilistic as
• 11:39
  Speaker 2: opposed to deterministic. It's not like traditional programming, where you've
• 11:42
  Speaker 2: got it, you've got very discrete logic and then the
• 11:45
  Speaker 2: outcome is very predictable and you can read each line
• 11:47
  Speaker 2: and see what each line's going to do. A neural
• 11:50
  Speaker 2: net is just a whole bunch of probabilities. It ends
• 11:52
  Speaker 2: up being a giant Comma separated value file. Our digital
• 11:55
  Speaker 2: guard is a CSP file. Really, Okay, that is what
• 11:58
  Speaker 2: it is. It's hard to say exactly what that moment is,
• 12:00
  Speaker 2: but there will come a point where no job is needed.
• 12:04
  Speaker 2: You can have a job if you want to have
• 12:05
  Speaker 2: a job for sort of personal satisfaction, but the AI
• 12:09
  Speaker 2: will be able to do everything. I don't know if
• 12:11
  Speaker 2: that makes people comfortable uncomfortable. It's if you wish for
• 12:14
  Speaker 2: a magic gene that gives you any wishes you want,
• 12:16
  Speaker 2: and there's no limit. You don't have those three limits,
• 12:18
  Speaker 2: three wish limits, not since you just have many, many
• 12:20
  Speaker 2: wishes you want.
• 12:22
  Speaker 1: It's both good and bad.
• 12:23
  Speaker 2: One of the challenges in the future will be how
• 12:25
  Speaker 2: do we find meaning in life? If you have a
• 12:27
  Speaker 2: magic genie that can do everything you want. When there's
• 12:29
  Speaker 2: new technology, it tends to have usually follow an S curve.
• 12:32
  Speaker 2: In this case, we're going to be on the exponential
• 12:34
  Speaker 2: portion of the S curve for a long time and
• 12:37
  Speaker 2: we have to ask for anything. It won't be we
• 12:40
  Speaker 2: won't have universal basic income. We'll have universal high income.
• 12:43
  Speaker 2: So in some sense it will be somewhat of a
• 12:45
  Speaker 2: leveler or an equalizer, because really I think everyone will
• 12:48
  Speaker 2: have access to this magic gene and you're able to
• 12:50
  Speaker 2: ask any question. It will certainly be good for education.
• 12:52
  Speaker 2: It will be the best tutor you could add them,
• 12:54
  Speaker 2: the most patient tutor sit there all day, and there
• 12:57
  Speaker 2: will be no shortage of goods and services. Will be
• 12:58
  Speaker 2: an age of abundance. I don't give you a sense
• 13:00
  Speaker 2: of what is a fairly utopian or pro tooken future
• 13:06
  Speaker 2: with AI. I don't think the AI is going to
• 13:08
  Speaker 2: try to destroyal humanity, but it might put us under
• 13:11
  Speaker 2: strict controls. And there's not a non zero chance ofgoing terminator.
• 13:14
  Speaker 2: It's not zero percent, but it's I think it's a
• 13:17
  Speaker 2: small leglihood of notating humanity, but it's not zero. We
• 13:20
  Speaker 2: want to public to be zeros as close to zero
• 13:23
  Speaker 2: as possible, and then, like I said, be of AI,
• 13:26
  Speaker 2: assuming control the safety of all the humans and taking
• 13:30
  Speaker 2: over all the computing systems and weapons systems of Earth
• 13:33
  Speaker 2: and effectively being like some sort of uber and anti
• 13:36
  Speaker 2: Everything will be automatic. There'll be hassle robots that you
• 13:39
  Speaker 2: can fully talk to as though there are people that
• 13:42
  Speaker 2: can help your own possible be a companion, or whatever
• 13:44
  Speaker 2: the case may be. That will be het Robots throughout
• 13:47
  Speaker 2: factories and poss will also be all automatic. And anything
• 13:52
  Speaker 2: that we're intelligent intelligence can be applied, even modest intelligence
• 13:56
  Speaker 2: will be automated. I think people should be really concerned
• 13:59
  Speaker 2: about it. I keep soundingly long bell, but until people
• 14:03
  Speaker 2: see like robots going down the street killing people, they
• 14:06
  Speaker 2: don't know how to react because it seems so ethereal.
• 14:09
  Speaker 2: And I think we should be really concerned about AI,
• 14:12
  Speaker 2: and I think we should. AI is a rare case
• 14:14
  Speaker 2: where I think we need to be proactive in regulation
• 14:17
  Speaker 2: instead of reactive, because I think by the time we
• 14:20
  Speaker 2: are reactive in AI regulation, it's too late. Normally, the
• 14:24
  Speaker 2: way regulations are set up is that a whole bunch.
• 14:27
  Speaker 1: Of bad things happen.
• 14:28
  Speaker 2: There's a public outcry, and then after many years, a
• 14:32
  Speaker 2: regulatory agency is set up to regulate that industry and
• 14:36
  Speaker 2: there's a bunch of opposition from companies who don't like
• 14:40
  Speaker 2: being told what to do by regulators, and it takes forever.
• 14:43
  Speaker 2: In the past has been bad, but not something which
• 14:47
  Speaker 2: represented a fundamental risk to the existence of civilization. AI
• 14:52
  Speaker 2: is a fundamental risk to the existence of human civilization
• 14:55
  Speaker 2: in a way that car accidents, airplane crashes, faulty drugs,
• 15:01
  Speaker 2: or bad food were not.
• 15:03
  Speaker 1: They were not.
• 15:03
  Speaker 2: They were harmful t into instead of individuals within society,
• 15:07
  Speaker 2: of course, but they were not harmful to society as
• 15:09
  Speaker 2: a whole. AI is a fundamental existential risk for human civilization.
• 15:15
  Speaker 2: It is really like the scariest problem to me. I'll
• 15:17
  Speaker 2: tell you, I really think we need to go a
• 15:19
  Speaker 2: regulation here, just because this is ensuring the public good
• 15:23
  Speaker 2: is served. Because you've got companies that are racing that
• 15:26
  Speaker 2: they kind of have to race to build AI or
• 15:28
  Speaker 2: they're going to be made uncompetitive. Like essentially, if your
• 15:32
  Speaker 2: competitor is racing to bold AI and you don't, they
• 15:35
  Speaker 2: will crush you. So then you're like, ah, we don't
• 15:37
  Speaker 2: want to be crushed, So you know, I guess we
• 15:40
  Speaker 2: need to build it too. That's where you need the
• 15:42
  Speaker 2: regulators to come in and say, hey, guys, you all
• 15:44
  Speaker 2: need to really pause and make sure this is safe.
• 15:48
  Speaker 2: And like when when it's cool and work commit and
• 15:50
  Speaker 2: regulators are convinced that it's safe, proceed, then you can go.
• 15:52
  Speaker 1: But otherwise slow down, and we could definitely make a
• 15:56
  Speaker 1: city of.
• 15:57
  Speaker 2: Mascle self sustaining without without AI or that some of AGI,
• 16:01
  Speaker 2: which is generally additional general intelligence or superintelligence I have.
• 16:04
  Speaker 2: So I think that is it's not necessary for anything
• 16:06
  Speaker 2: we're doing, but it is happening, and happening very quickly.
• 16:09
  Speaker 2: There is a risk that advanced AI either eliminates or
• 16:14
  Speaker 2: constrains humanity's growth.