back to index

Steven Pinker: AI in the Age of Reason | Lex Fridman Podcast #3


small model | large model

link |
00:00:00.000
You've studied the human mind, cognition, language, vision, evolution, psychology, from child to adult,
link |
00:00:07.360
from the level of individual to the level of our entire civilization,
link |
00:00:11.680
so I feel like I can start with a simple multiple choice question.
link |
00:00:16.240
What is the meaning of life? Is it A, to attain knowledge, as Plato said,
link |
00:00:22.400
B, to attain power, as Nietzsche said, C, to escape death, as Ernest Becker said,
link |
00:00:28.000
D, to propagate our genes, as Darwin and others have said, E, there is no meaning,
link |
00:00:35.040
as the nihilists have said, F, knowing the meaning of life is beyond our cognitive capabilities,
link |
00:00:41.440
as Stephen Pinker said, based on my interpretation 20 years ago, and G, none of the above.
link |
00:00:48.160
I'd say A comes closest, but I would amend that to attaining not only knowledge, but fulfillment
link |
00:00:54.720
more generally. That is, life, health, stimulation, access to the living cultural and social world.
link |
00:01:06.000
Now, this is our meaning of life. It's not the meaning of life, if you were to ask our genes.
link |
00:01:12.160
Their meaning is to propagate copies of themselves, but that is distinct from the
link |
00:01:17.600
meaning that the brain that they lead to sets for itself. So, to you, knowledge is a small subset
link |
00:01:26.640
or a large subset? It's a large subset, but it's not the entirety of human striving, because we
link |
00:01:33.280
also want to interact with people. We want to experience beauty. We want to experience the
link |
00:01:39.120
richness of the natural world, but understanding what makes the universe tick is way up there.
link |
00:01:47.840
For some of us more than others, certainly for me, that's one of the top five.
link |
00:01:54.560
So, is that a fundamental aspect? Are you just describing your own preference, or is this a
link |
00:02:00.080
fundamental aspect of human nature, is to seek knowledge? In your latest book, you talk about
link |
00:02:05.920
the power, the usefulness of rationality and reason and so on. Is that a fundamental
link |
00:02:11.760
nature of human beings, or is it something we should just strive for?
link |
00:02:16.960
Both. We're capable of striving for it, because it is one of the things that
link |
00:02:22.640
make us what we are, homo sapiens, wise men. We are unusual among our animals in the degree to
link |
00:02:31.360
which we acquire knowledge and use it to survive. We make tools. We strike agreements via language.
link |
00:02:39.760
We extract poisons. We predict the behavior of animals. We try to get at the workings of plants.
link |
00:02:47.760
And when I say we, I don't just mean we in the modern west, but we as a species everywhere,
link |
00:02:52.640
which is how we've managed to occupy every niche on the planet, how we've managed to drive other
link |
00:02:58.160
animals to extinction. And the refinement of reason in pursuit of human well being, of health,
link |
00:03:06.480
happiness, social richness, cultural richness, is our main challenge in the present. That is,
link |
00:03:14.480
using our intellect, using our knowledge to figure out how the world works, how we work,
link |
00:03:19.280
in order to make discoveries and strike agreements that make us all better off in the long run.
link |
00:03:25.200
Right. And you do that almost undeniably in a data driven way in your recent book,
link |
00:03:31.920
but I'd like to focus on the artificial intelligence aspect of things, and not just
link |
00:03:36.480
artificial intelligence, but natural intelligence too. So 20 years ago in the book, you've written
link |
00:03:41.840
on how the mind works, you conjecture, again, am I right to interpret things? You can correct me
link |
00:03:49.600
if I'm wrong, but you conjecture that human thought in the brain may be a result of a network, a massive
link |
00:03:55.200
network of highly interconnected neurons. So from this interconnectivity emerges thought,
link |
00:04:01.280
compared to artificial neural networks, which we use for machine learning today,
link |
00:04:06.160
is there something fundamentally more complex, mysterious, even magical about the biological
link |
00:04:12.640
neural networks versus the ones we've been starting to use over the past 60 years and
link |
00:04:19.440
become to success in the past 10? There is something a little bit mysterious about
link |
00:04:25.840
the human neural networks, which is that each one of us who is a neural network knows that we
link |
00:04:31.760
ourselves are conscious, conscious not in the sense of registering our surroundings or even
link |
00:04:36.960
registering our internal state, but in having subjective first person, present tense experience.
link |
00:04:42.720
That is, when I see red, it's not just different from green, but there's a redness to it that I
link |
00:04:49.840
feel. Whether an artificial system would experience that or not, I don't know and I don't think I
link |
00:04:54.720
can know. That's why it's mysterious. If we had a perfectly lifelike robot that was behaviorally
link |
00:05:00.480
indistinguishable from a human, would we attribute consciousness to it or ought we to attribute
link |
00:05:06.800
consciousness to it? And that's something that it's very hard to know. But putting that aside,
link |
00:05:12.160
putting aside that largely philosophical question, the question is, is there some difference between
link |
00:05:19.040
the human neural network and the ones that we're building in artificial intelligence will mean that
link |
00:05:23.920
we're on the current trajectory not going to reach the point where we've got a lifelike robot
link |
00:05:30.400
indistinguishable from a human because the way their so called neural networks are organized
link |
00:05:35.120
are different from the way ours are organized. I think there's overlap, but I think there are some
link |
00:05:40.560
big differences that their current neural networks, current so called deep learning systems are in
link |
00:05:48.720
reality not all that deep. That is, they are very good at extracting high order statistical
link |
00:05:53.840
regularities. But most of the systems don't have a semantic level, a level of actual understanding
link |
00:06:00.640
of who did what to whom, why, where, how things work, what causes, what else.
link |
00:06:06.400
Do you think that kind of thing can emerge as it does so artificial neural networks are much
link |
00:06:10.960
smaller the number of connections and so on than the current human biological networks? But do you
link |
00:06:16.480
think sort of go to consciousness or to go to this higher level semantic reasoning about things?
link |
00:06:22.640
Do you think that can emerge with just a larger network with a more richly, weirdly interconnected
link |
00:06:28.960
network? Separate it in consciousness because consciousness is even a matter of complexity.
link |
00:06:33.280
A really weird one. Yeah, you could sensibly ask the question of whether shrimp are conscious,
link |
00:06:37.920
for example. They're not terribly complex, but maybe they feel pain. So let's just put that
link |
00:06:43.200
part of it aside. But I think sheer size of a neural network is not enough to give it
link |
00:06:50.960
structure and knowledge. But if it's suitably engineered, then why not? That is, we're neural
link |
00:06:57.360
networks. Natural selection did a kind of equivalent of engineering of our brains. So I don't think
link |
00:07:03.680
there's anything mysterious in the sense that no systemated of silicon could ever do what a human
link |
00:07:10.880
brain can do. I think it's possible in principle. Whether it'll ever happen depends not only on
link |
00:07:16.080
how clever we are in engineering these systems, but whether even we even want to, whether that's
link |
00:07:21.040
even a sensible goal. That is, you can ask the question, is there any locomotion system that is
link |
00:07:28.320
as good as a human? Well, we kind of want to do better than a human ultimately in terms of
link |
00:07:32.960
legged locomotion. There's no reason that humans should be our benchmark. They're tools that might
link |
00:07:39.360
be better in some ways. It may be that we can't duplicate a natural system because at some point,
link |
00:07:49.280
it's so much cheaper to use a natural system that we're not going to invest more brain power
link |
00:07:53.840
and resources. So for example, we don't really have an exact substitute for wood. We still build
link |
00:08:00.000
houses out of wood. We still build furniture out of wood. We like the look. We like the feel. It's
link |
00:08:04.400
wood has certain properties that synthetics don't. There's not that there's anything magical or
link |
00:08:09.280
mysterious about wood. It's just that the extra steps of duplicating everything about wood is
link |
00:08:16.400
something we just haven't bothered because we have wood. Likewise, cotton. I'm wearing cotton
link |
00:08:20.480
clothing now. It feels much better than polyester. It's not that cotton has something magic in it,
link |
00:08:27.600
and it's not that we couldn't ever synthesize something exactly like cotton,
link |
00:08:33.120
but at some point, it's just not worth it. We've got cotton. Likewise, in the case of human
link |
00:08:37.760
intelligence, the goal of making an artificial system that is exactly like the human brain
link |
00:08:43.520
is a goal that we probably know is going to pursue to the bitter end, I suspect, because
link |
00:08:50.080
if you want tools that do things better than humans, you're not going to care whether it
link |
00:08:53.600
does something like humans. So for example, diagnosing cancer or predicting the weather,
link |
00:08:58.720
why set humans as your benchmark? But in general, I suspect you also believe that even if the human
link |
00:09:07.360
should not be a benchmark and we don't want to imitate humans in their system, there's a lot
link |
00:09:11.440
to be learned about how to create an artificial intelligence system by studying the humans.
link |
00:09:16.800
Yeah, I think that's right. In the same way that to build flying machines, we want to understand
link |
00:09:23.440
the laws of aerodynamics, including birds, but not mimic the birds, but they're the same laws.
link |
00:09:30.480
You have a view on AI, artificial intelligence and safety, that from my perspective,
link |
00:09:38.400
is refreshingly rational, or perhaps more importantly, has elements of positivity to it,
link |
00:09:49.360
which I think can be inspiring and empowering as opposed to paralyzing. For many people,
link |
00:09:55.440
including AI researchers, the eventual existential threat of AI is obvious, not only possible but
link |
00:10:02.320
obvious. And for many others, including AI researchers, the threat is not obvious. So
link |
00:10:09.520
Elon Musk is famously in the highly concerned about AI camp, saying things like AI is far
link |
00:10:16.480
more dangerous than nuclear weapons, and that AI will likely destroy human civilization.
link |
00:10:22.960
So in February, you said that if Elon was really serious about AI, the threat of AI,
link |
00:10:30.400
he would stop building self driving cars that he's doing very successfully as part of Tesla.
link |
00:10:35.840
Then he said, wow, if even Pinker doesn't understand the difference between narrow AI
link |
00:10:40.880
like a car and general AI, when the latter literally has a million times more compute power
link |
00:10:47.280
and an open ended utility function, humanity is in deep trouble. So first, what did you mean by
link |
00:10:54.240
the statement about Elon Musk should stop building self driving cars if he's deeply concerned?
link |
00:10:59.200
Well, not the last time that Elon Musk has fired off an intemperate tweet.
link |
00:11:04.320
Well, we live in a world where Twitter has power.
link |
00:11:07.600
Yes. Yeah, I think there are two kinds of existential threat that have been discussed
link |
00:11:16.640
in connection with artificial intelligence, and I think that they're both incoherent.
link |
00:11:20.480
One of them is a vague fear of AI takeover, that just as we subjugated animals and less
link |
00:11:28.800
technologically advanced peoples, so if we build something that's more advanced than us,
link |
00:11:33.360
it will inevitably turn us into pets or slaves or domesticated animal equivalents.
link |
00:11:40.240
I think this confuses intelligence with a will to power that it so happens that in the
link |
00:11:46.720
intelligence system we are most familiar with, namely Homo sapiens, we are products of natural
link |
00:11:52.240
selection, which is a competitive process. And so bundled together with our problem solving
link |
00:11:56.800
capacity are a number of nasty traits like dominance and exploitation and maximization of
link |
00:12:05.200
power and glory and resources and influence. There's no reason to think that sheer problem
link |
00:12:11.040
solving capability will set that as one of its goals. Its goals will be whatever we set its goals
link |
00:12:16.640
as, and as long as someone isn't building a megalomaniacal artificial intelligence,
link |
00:12:22.560
then there's no reason to think that it would naturally evolve in that direction.
link |
00:12:25.360
Now you might say, well, what if we gave it the goal of maximizing its own power source?
link |
00:12:31.600
That's a pretty stupid goal to give an autonomous system. You don't give it that goal.
link |
00:12:36.000
I mean, that's just self evidently idiotic. So if you look at the history of the world,
link |
00:12:41.360
there's been a lot of opportunities where engineers could instill in a system destructive
link |
00:12:45.680
power and they choose not to because that's the natural process of engineering.
link |
00:12:49.520
Well, except for weapons. I mean, if you're building a weapon, its goal is to destroy
link |
00:12:52.880
people. And so I think there are good reasons to not build certain kinds of weapons. I think
link |
00:12:58.400
building nuclear weapons was a massive mistake. You do. So maybe pause on that because that is
link |
00:13:06.240
one of the serious threats. Do you think that it was a mistake in a sense that it should have been
link |
00:13:12.880
stopped early on? Or do you think it's just an unfortunate event of invention that this was
link |
00:13:19.200
invented? Do you think it's possible to stop, I guess, is the question on that? Yeah, it's hard to
link |
00:13:22.800
rewind the clock because, of course, it was invented in the context of World War II and the
link |
00:13:27.440
fear that the Nazis might develop one first. Then once it was initiated for that reason,
link |
00:13:33.120
it was hard to turn off, especially since winning the war against the Japanese and the Nazis was
link |
00:13:40.800
such an overwhelming goal of every responsible person that they were just nothing that people
link |
00:13:46.160
wouldn't have done then to ensure victory. It's quite possible if World War II hadn't happened
link |
00:13:51.440
that nuclear weapons wouldn't have been invented. We can't know. But I don't think it was, by any
link |
00:13:56.560
means, a necessity any more than some of the other weapon systems that were envisioned but never
link |
00:14:01.760
implemented, like planes that would disperse poison gas over cities like crop dusters or systems to
link |
00:14:09.040
try to create earthquakes and tsunamis in enemy countries, to weaponize the weather,
link |
00:14:16.080
weaponize solar flares, all kinds of crazy schemes that we thought the better of. I think
link |
00:14:21.520
analogies between nuclear weapons and artificial intelligence are fundamentally misguided because
link |
00:14:26.800
the whole point of nuclear weapons is to destroy things. The point of artificial intelligence
link |
00:14:31.520
is not to destroy things. So the analogy is misleading. So there's two artificial
link |
00:14:37.360
intelligence you mentioned. The first one was the highly intelligent or power hungry. Yeah,
link |
00:14:42.320
an assistant that we design ourselves where we give it the goals. Goals are external to the
link |
00:14:48.320
means to attain the goals. If we don't design an artificially intelligent system to maximize
link |
00:14:56.560
dominance, then it won't maximize dominance. It's just that we're so familiar with homo sapiens
link |
00:15:02.400
where these two traits come bundled together, particularly in men, that we are apt to confuse
link |
00:15:08.800
high intelligence with a will to power. But that's just an error. The other fear is that
link |
00:15:16.720
we'll be collateral damage that will give artificial intelligence a goal like make paper clips
link |
00:15:23.040
and it will pursue that goal so brilliantly that before we can stop it, it turns us into paper
link |
00:15:28.320
clips. We'll give it the goal of curing cancer and it will turn us into guinea pigs for lethal
link |
00:15:34.400
experiments or give it the goal of world peace and its conception of world peace is no people,
link |
00:15:40.000
therefore no fighting and so it will kill us all. Now, I think these are utterly fanciful. In fact,
link |
00:15:44.480
I think they're actually self defeating. They first of all assume that we're going to be so
link |
00:15:49.600
brilliant that we can design an artificial intelligence that can cure cancer. But so stupid
link |
00:15:54.880
that we don't specify what we mean by curing cancer in enough detail that it won't kill us in the
link |
00:16:00.160
process. And it assumes that the system will be so smart that it can cure cancer. But so
link |
00:16:06.720
idiotic that it doesn't can't figure out that what we mean by curing cancer is not killing
link |
00:16:11.520
everyone. So I think that the collateral damage scenario, the value alignment problem is also
link |
00:16:17.920
based on a misconception. So one of the challenges, of course, we don't know how to build either system
link |
00:16:23.200
currently, or are we even close to knowing? Of course, those things can change overnight,
link |
00:16:27.440
but at this time, theorizing about is very challenging in either direction. So that's
link |
00:16:33.840
probably at the core of the problem is without that ability to reason about the real engineering
link |
00:16:39.600
things here at hand is your imagination runs away with things. Exactly. But let me sort of ask,
link |
00:16:45.920
what do you think was the motivation, the thought process of Elon Musk? I build autonomous vehicles,
link |
00:16:52.320
I study autonomous vehicles, I study Tesla autopilot. I think it is one of the greatest
link |
00:16:58.000
currently application, large scale application of artificial intelligence in the world.
link |
00:17:02.880
It has a potentially very positive impact on society. So how does a person who's creating this
link |
00:17:09.120
very good, quote unquote, narrow AI system also seem to be so concerned about this other
link |
00:17:17.680
general AI? What do you think is the motivation there? What do you think is the thing?
link |
00:17:21.120
Well, you probably have to ask him, but there and he is notoriously flamboyant, impulsive to the,
link |
00:17:30.640
as we have just seen, to the detriment of his own goals of the health of a company.
link |
00:17:36.000
So I don't know what's going on in his mind. You probably have to ask him. But I don't think the,
link |
00:17:41.600
and I don't think the distinction between special purpose AI and so called general AI is relevant
link |
00:17:48.160
that in the same way that special purpose AI is not going to do anything conceivable in order to
link |
00:17:54.400
attain a goal, all engineering systems have to are designed to trade off across multiple goals.
link |
00:18:00.560
When we build cars in the first place, we didn't forget to install brakes because the goal of a
link |
00:18:05.920
car is to go fast. It occurred to people, yes, you want to go fast, but not always. So you build
link |
00:18:12.320
and brakes too. Likewise, if a car is going to be autonomous, that doesn't program it to take the
link |
00:18:18.960
shortest route to the airport. It's not going to take the diagonal and mow down people and trees
link |
00:18:23.440
and fences because that's the shortest route. That's not what we mean by the shortest route when we
link |
00:18:28.000
program it. And that's just what an intelligence system is by definition. It takes into account
link |
00:18:34.720
multiple constraints. The same is true, in fact, even more true of so called general intelligence.
link |
00:18:40.640
That is, if it's genuinely intelligent, it's not going to pursue some goal single mindedly,
link |
00:18:47.040
omitting every other consideration and collateral effect. That's not artificial and
link |
00:18:53.280
general intelligence. That's artificial stupidity. I agree with you, by the way,
link |
00:18:58.560
on the promise of autonomous vehicles for improving human welfare. I think it's spectacular.
link |
00:19:03.280
And I'm surprised at how little press coverage notes that in the United States alone,
link |
00:19:08.080
something like 40,000 people die every year on the highways, vastly more than are killed by
link |
00:19:13.200
terrorists. And we spend a trillion dollars on a war to combat deaths by terrorism,
link |
00:19:19.440
about half a dozen a year, whereas every year and year out, 40,000 people are
link |
00:19:24.080
massacred on the highways, which could be brought down to very close to zero.
link |
00:19:28.560
So I'm with you on the humanitarian benefit.
link |
00:19:31.840
Let me just mention that as a person who's building these cars, it is a little bit offensive to me
link |
00:19:36.240
to say that engineers would be clueless enough not to engineer safety into systems. I often
link |
00:19:41.680
stay up at night thinking about those 40,000 people that are dying. And everything I try to
link |
00:19:46.400
engineer is to save those people's lives. So every new invention that I'm super excited about,
link |
00:19:52.000
every new, all the deep learning literature and CVPR conferences and NIPS, everything I'm super
link |
00:19:59.680
excited about is all grounded in making it safe and help people. So I just don't see how that
link |
00:20:08.320
trajectory can all of a sudden slip into a situation where intelligence will be highly
link |
00:20:13.200
negative. You and I certainly agree on that. And I think that's only the beginning of the
link |
00:20:17.840
potential humanitarian benefits of artificial intelligence. There's been enormous attention
link |
00:20:23.760
to what are we going to do with the people whose jobs are made obsolete by artificial
link |
00:20:27.680
intelligence. But very little attention given to the fact that the jobs that are going to be
link |
00:20:31.600
made obsolete are horrible jobs. The fact that people aren't going to be picking crops and making
link |
00:20:37.600
beds and driving trucks and mining coal, these are soul deadening jobs. And we have a whole
link |
00:20:43.760
literature sympathizing with the people stuck in these menial, mind deadening, dangerous jobs.
link |
00:20:52.080
If we can eliminate them, this is a fantastic boon to humanity. Now, granted,
link |
00:20:56.160
we, you solve one problem and there's another one, namely, how do we get these people a decent
link |
00:21:02.160
income? But if we're smart enough to invent machines that can make beds and put away dishes and
link |
00:21:09.520
handle hospital patients, I think we're smart enough to figure out how to redistribute income
link |
00:21:14.080
to a portion, some of the vast economic savings to the human beings who will no longer be needed to
link |
00:21:20.960
make beds. Okay. Sam Harris says that it's obvious that eventually AI will be an existential risk.
link |
00:21:29.280
He's one of the people who says it's obvious. We don't know when the claim goes, but eventually
link |
00:21:36.240
it's obvious. And because we don't know when we should worry about it now. It's a very interesting
link |
00:21:41.760
argument in my eyes. So how do we think about timescale? How do we think about existential
link |
00:21:49.120
threats when we don't really, we know so little about the threat, unlike nuclear weapons, perhaps,
link |
00:21:55.040
about this particular threat, that it could happen tomorrow, right? So, but very likely it won't.
link |
00:22:03.120
Very likely it'd be 100 years away. So how do, do we ignore it? Do, how do we talk about it?
link |
00:22:08.880
Do we worry about it? What, how do we think about those? What is it?
link |
00:22:13.040
A threat that we can imagine, it's within the limits of our imagination, but not within our
link |
00:22:19.600
limits of understanding to sufficient, to accurately predict it. But what, what is, what is the it
link |
00:22:25.760
that we're referring to? Oh, AI, sorry, AI, AI being the existential threat. AI can always...
link |
00:22:31.280
How? But like enslaving us or turning us into paperclips?
link |
00:22:35.120
I think the most compelling from the Sam Harris perspective would be the paperclip situation.
link |
00:22:38.800
Yeah. I mean, I just think it's totally fanciful. I mean, that is, don't build a system. Don't give a,
link |
00:22:44.000
don't... First of all, the code of engineering is you don't implement a system with massive
link |
00:22:50.400
control before testing it. Now, perhaps the culture of engineering will radically change,
link |
00:22:55.040
then I would worry, but I don't see any signs that engineers will suddenly do idiotic things,
link |
00:23:00.320
like put a, an electrical power plant in control of a system that they haven't tested
link |
00:23:05.440
first. Or all of these scenarios not only imagine a almost a magically powered intelligence,
link |
00:23:15.360
you know, including things like cure cancer, which is probably an incoherent goal because
link |
00:23:20.000
there's so many different kinds of cancer or bring about world peace. I mean, how do you even specify
link |
00:23:25.440
that as a goal? But the scenarios also imagine some degree of control of every molecule in the
link |
00:23:31.360
universe, which not only is itself unlikely, but we would not start to connect these systems to
link |
00:23:39.200
infrastructure without, without testing as we would any kind of engineering system. Now,
link |
00:23:45.840
maybe some engineers will be irresponsible and we need legal and regulatory and legal
link |
00:23:53.920
responsibility implemented so that engineers don't do things that are stupid by their own standards.
link |
00:23:59.440
But the, I've never seen enough of a plausible scenario of existential threat to devote large
link |
00:24:08.560
amounts of brain power to, to forestall it. So you believe in the sort of the power en masse of
link |
00:24:14.720
the engineering of reason as you argue in your latest book of reason and science to sort of
link |
00:24:20.400
be the very thing that guides the development of new technology so it's safe and also keeps us
link |
00:24:26.160
safe. Yeah, the same, you know, granted the same culture of safety that currently is part of the
link |
00:24:32.480
engineering mindset for airplanes, for example. So yeah, I don't think that, that that should
link |
00:24:38.960
be thrown out the window and that untested, all powerful systems should be suddenly implemented.
link |
00:24:44.800
But there's no reason to think they are. And in fact, if you look at the
link |
00:24:48.160
progress of artificial intelligence, it's been, you know, it's been impressive, especially in
link |
00:24:51.760
the last 10 years or so. But the idea that suddenly there'll be a step function that all of a sudden
link |
00:24:56.960
before we know it, it will be all powerful, that there'll be some kind of recursive self
link |
00:25:02.160
improvement, some kind of fume is also fanciful. Certainly by the technology that we that we're
link |
00:25:11.200
now impresses us, such as deep learning, where you train something on hundreds of thousands or
link |
00:25:16.720
millions of examples, they're not hundreds of thousands of problems of which curing cancer is
link |
00:25:24.320
typical example. And so the kind of techniques that have allowed AI to increase in the last
link |
00:25:30.560
five years are not the kind that are going to lead to this fantasy of exponential sudden
link |
00:25:37.600
self improvement. So I think it's kind of a magical thinking. It's not based on our understanding
link |
00:25:43.680
of how AI actually works. Now, give me a chance here. So you said fanciful, magical thinking.
link |
00:25:50.240
In his TED Talk, Sam Harris says that thinking about AI killing all human civilization is somehow
link |
00:25:55.280
fun intellectually. Now, I have to say as a scientist engineer, I don't find it fun.
link |
00:26:01.200
But when I'm having beer with my non AI friends, there is indeed something fun and appealing about
link |
00:26:08.560
it. Like talking about an episode of Black Mirror, considering if a large meteor is headed towards
link |
00:26:14.720
Earth, we were just told a large meteor is headed towards Earth, something like this. And can you
link |
00:26:20.640
relate to this sense of fun? And do you understand the psychology of it? Yes, great. Good question.
link |
00:26:26.880
I personally don't find it fun. I find it kind of actually a waste of time, because there are
link |
00:26:33.440
genuine threats that we ought to be thinking about, like pandemics, like cybersecurity
link |
00:26:39.840
vulnerabilities, like the possibility of nuclear war and certainly climate change. This is enough
link |
00:26:47.040
to fill many conversations without. And I think Sam did put his finger on something, namely that
link |
00:26:55.280
there is a community, sometimes called the rationality community, that delights in using its
link |
00:27:03.120
brain power to come up with scenarios that would not occur to mere mortals, to less cerebral people.
link |
00:27:10.160
So there is a kind of intellectual thrill in finding new things to worry about that no one
link |
00:27:15.360
has worried about yet. I actually think, though, that it's not only is it a kind of fun that doesn't
link |
00:27:21.200
give me particular pleasure. But I think there can be a pernicious side to it, namely that you
link |
00:27:27.280
overcome people with such dread, such fatalism, that there's so many ways to die to annihilate
link |
00:27:35.280
our civilization that we may as well enjoy life while we can. There's nothing we can do about it.
link |
00:27:40.160
If climate change doesn't do us in, then runaway robots will. So let's enjoy ourselves now. We
link |
00:27:46.560
got to prioritize. We have to look at threats that are close to certainty, such as climate change,
link |
00:27:55.200
and distinguish those from ones that are merely imaginable, but with infinitesimal probabilities.
link |
00:28:01.360
And we have to take into account people's worry budget. You can't worry about everything. And
link |
00:28:07.120
if you sow dread and fear and terror and and fatalism, it can lead to a kind of numbness. Well,
link |
00:28:13.920
they're just these problems are overwhelming and the engineers are just going to kill us all.
link |
00:28:19.040
So let's either destroy the entire infrastructure of science, technology,
link |
00:28:26.640
or let's just enjoy life while we can. So there's a certain line of worry, which I'm
link |
00:28:32.080
worried about a lot of things engineering. There's a certain line of worry when you cross,
link |
00:28:36.160
you allow it to cross, that it becomes paralyzing fear as opposed to productive fear. And that's
link |
00:28:42.800
kind of what you're highlighting. Exactly right. And we've seen some, we know that human effort is
link |
00:28:49.760
not well calibrated against risk in that because a basic tenet of cognitive psychology is that
link |
00:28:59.440
perception of risk and hence perception of fear is driven by imaginability, not by data.
link |
00:29:05.920
And so we misallocate vast amounts of resources to avoiding terrorism,
link |
00:29:11.200
which kills on average about six Americans a year with a one exception of 9 11. We invade
link |
00:29:16.240
countries, we invent an entire new departments of government with massive, massive expenditure
link |
00:29:23.920
of resources and lives to defend ourselves against a trivial risk. Whereas guaranteed risks,
link |
00:29:30.800
you mentioned as one of them, you mentioned traffic fatalities and even risks that are
link |
00:29:36.720
not here, but are plausible enough to worry about like pandemics, like nuclear war,
link |
00:29:47.120
receive far too little attention. In presidential debates, there's no discussion of
link |
00:29:51.760
how to minimize the risk of nuclear war, lots of discussion of terrorism, for example.
link |
00:29:57.840
And so we, I think it's essential to calibrate our budget of fear, worry, concerned planning
link |
00:30:05.520
to the actual probability of harm. Yep. So let me ask this in this question.
link |
00:30:13.520
So speaking of imaginability, you said it's important to think about reason. And one of my
link |
00:30:18.960
favorite people who likes to dip into the outskirts of reason through fascinating exploration of his
link |
00:30:26.560
imagination is Joe Rogan. Oh, yes. So who has, through reason, used to believe a lot of conspiracies
link |
00:30:34.880
and through reason has stripped away a lot of his beliefs in that way. So it's fascinating actually
link |
00:30:40.000
to watch him through rationality, kind of throw away the ideas of Bigfoot and 911. I'm not sure
link |
00:30:47.920
exactly. Kim Trails. I don't know what he believes in. Yes, okay. But he no longer believed in,
link |
00:30:52.320
that's right. No, he's become a real force for good. So you were on the Joe Rogan podcast in
link |
00:30:57.920
February and had a fascinating conversation, but as far as I remember, didn't talk much about
link |
00:31:02.880
artificial intelligence. I will be on his podcast in a couple of weeks. Joe is very much concerned
link |
00:31:09.280
about existential threat of AI. I'm not sure if you're, this is why I was hoping that you'll get
link |
00:31:14.640
into that topic. And in this way, he represents quite a lot of people who look at the topic of AI
link |
00:31:20.480
from 10,000 foot level. So as an exercise of communication, you said it's important to be
link |
00:31:27.840
rational and reason about these things. Let me ask, if you were to coach me as an AI researcher
link |
00:31:33.280
about how to speak to Joe and the general public about AI, what would you advise?
link |
00:31:38.320
Well, the short answer would be to read the sections that I wrote in Enlightenment.
link |
00:31:44.080
But longer reason would be, I think to emphasize, and I think you're very well positioned as an
link |
00:31:48.880
engineer to remind people about the culture of engineering, that it really is safety oriented,
link |
00:31:54.800
that another discussion in Enlightenment now, I plot rates of accidental death from various
link |
00:32:02.160
causes, plane crashes, car crashes, occupational accidents, even death by lightning strikes,
link |
00:32:09.280
and they all plummet. Because the culture of engineering is how do you squeeze out the lethal
link |
00:32:16.560
risks, death by fire, death by drowning, death by asphyxiation, all of them drastically declined
link |
00:32:23.360
because of advances in engineering, that I got to say, I did not appreciate until I saw those
link |
00:32:28.160
graphs. And it is because exactly people like you who stay up at night thinking, oh my God,
link |
00:32:36.000
what I'm inventing likely to hurt people and to deploy ingenuity to prevent that from happening.
link |
00:32:42.560
Now, I'm not an engineer, although I spent 22 years at MIT, so I know something about the culture
link |
00:32:47.360
of engineering. My understanding is that this is the way you think if you're an engineer.
link |
00:32:51.360
And it's essential that that culture not be suddenly switched off when it comes to artificial
link |
00:32:58.160
intelligence. So I mean, that could be a problem, but is there any reason to think it would be
link |
00:33:02.080
switched off? I don't think so. And one, there's not enough engineers speaking up for this way,
link |
00:33:07.360
for the excitement, for the positive view of human nature, what you're trying to create is
link |
00:33:13.680
the positivity, like everything we try to invent is trying to do good for the world.
link |
00:33:18.240
But let me ask you about the psychology of negativity. It seems just objectively,
link |
00:33:23.600
not considering the topic, it seems that being negative about the future, it makes you sound
link |
00:33:27.680
smarter than being positive about the future, in regard to this topic. Am I correct in this
link |
00:33:32.720
observation? And if so, why do you think that is? Yeah, I think there is that phenomenon,
link |
00:33:39.120
that as Tom Lehrer, the satirist said, always predict the worst and you'll be hailed as a
link |
00:33:43.920
prophet. It may be part of our overall negativity bias. We are as a species more attuned to the
link |
00:33:51.840
negative than the positive. We dread losses more than we enjoy gains. And that might open up a
link |
00:33:59.200
space for prophets to remind us of harms and risks and losses that we may have overlooked.
link |
00:34:06.560
So I think there is that asymmetry. So you've written some of my favorite books
link |
00:34:16.080
all over the place. So starting from Enlightenment now, to the better ranges of our nature,
link |
00:34:21.680
blank slate, how the mind works, the one about language, language instinct. Bill Gates,
link |
00:34:28.560
big fan too, said of your most recent book that it's my new favorite book of all time. So for
link |
00:34:37.840
you as an author, what was the book early on in your life that had a profound impact on the way
link |
00:34:44.000
you saw the world? Certainly this book Enlightenment now is influenced by David Deutch's The Beginning
link |
00:34:50.560
of Infinity. We have a rather deep reflection on knowledge and the power of knowledge to improve
link |
00:34:57.520
the human condition. They end with bits of wisdom such as that problems are inevitable,
link |
00:35:02.960
but problems are solvable given the right knowledge and that solutions create new problems
link |
00:35:07.760
that have to be solved in their turn. That's I think a kind of wisdom about the human condition
link |
00:35:12.480
that influenced the writing of this book. There's some books that are excellent but obscure,
link |
00:35:16.960
some of which I have on my page on my website. I read a book called The History of Force,
link |
00:35:22.080
self published by a political scientist named James Payne on the historical decline of violence and
link |
00:35:27.920
that was one of the inspirations for the better angels of our nature. What about early on if
link |
00:35:35.120
you look back when you were maybe a teenager? I loved a book called One, Two, Three, Infinity.
link |
00:35:40.640
When I was a young adult, I read that book by George Gamov, the physicist, which had very
link |
00:35:45.920
accessible and humorous explanations of relativity, of number theory, of dimensionality, high
link |
00:35:56.080
multiple dimensional spaces in a way that I think is still delightful 70 years after it was published.
link |
00:36:03.120
I like the Time Life Science series. These are books that arrive every month that my mother
link |
00:36:09.280
subscribed to. Each one on a different topic. One would be on electricity, one would be on
link |
00:36:15.600
forests, one would be on evolution, and then one was on the mind. I was just intrigued that there
link |
00:36:21.440
could be a science of mind. That book, I would cite as an influence as well. Then later on.
link |
00:36:27.040
That's when you fell in love with the idea of studying the mind. Was that the thing that grabbed
link |
00:36:30.960
you? It was one of the things, I would say. I read as a college student the book Reflections on
link |
00:36:38.560
Language by Noam Chomsky. He spent most of his career here at MIT. Richard Dawkins,
link |
00:36:44.800
two books, The Blind Watchmaker and the Selfish Gene were enormously influential,
link |
00:36:49.520
partly mainly for the content, but also for the writing style, the ability to explain
link |
00:36:56.640
abstract concepts in lively prose. Stephen Jay Gould's first collection ever since Darwin, also
link |
00:37:05.040
excellent example of lively writing. George Miller, the psychologist that most psychologists
link |
00:37:11.120
are familiar with, came up with the idea that human memory has a capacity of seven plus or minus
link |
00:37:17.440
two chunks. That's probably his biggest claim to fame. He wrote a couple of books on language
link |
00:37:21.920
and communication that I'd read as an undergraduate. Again, beautifully written and intellectually deep.
link |
00:37:28.400
Wonderful. Stephen, thank you so much for taking the time today.
link |
00:37:31.840
My pleasure. Thanks a lot, Lex.