00:00:00
They call you the godfather of AI. So what would you be saying to people about their career prospects in a world of
00:00:06
super intelligence? Train to be a plumber. Really? Yeah. Okay. I'm going to become a plumber. Jeffrey Hinton is
00:00:13
the Nobel Prize winning pioneer whose groundbreaking work has shaped AI and the future of humanity. Why do they call
00:00:20
it the godfather of AI? because there weren't many people who believed that we could model AI on the brain so that it
00:00:25
learned to do complicated things like recognize objects and images or even do reasoning. And I pushed that approach
00:00:30
for 50 years and then Google acquired that technology and I worked there for 10 years on something that's now used all the time in AI. And then you left.
00:00:37
Yeah. Why? So that I could talk freely at a conference. What did you want to talk about freely? How dangerous AI
00:00:43
could be. I realized that these things will one day get smarter than us. And we've never
00:00:49
had to deal with that. And if you want to know what life's like when you're not the apex intelligence, ask a chicken. So
00:00:55
there's risks that come from people misusing AI. And then there's risks from AI getting super smart and deciding it
00:01:00
doesn't need us. Is that a real risk? Yes, it is. But they're not going to stop it cuz it's too good for too many things. What about regulations? They
00:01:06
have some, but they're not designed to deal with most of the threats. Like the European regulations have a clause that say none of these apply to military uses
00:01:13
of AI. Really? Yeah. It's crazy. One of your students left OpenAI. Yeah. He was
00:01:19
probably the most important person behind the development of the early versions of church GPT and I think he left because he had safety concerns. We
00:01:25
should recognize that this stuff is an existential threat and we have to face the possibility that unless we do
00:01:31
something soon we're near the end. So let's do the risks. What do we end up doing in such a world?
00:01:39
This has always blown my mind a little bit. 53% of you that listen to the show regularly haven't yet subscribed to the
00:01:46
show. So, could I ask you for a favor before we start? If you like the show and you like what we do here and you want to support us, the free simple way
00:01:52
that you can do just that is by hitting the subscribe button. And my commitment to you is if you do that, then I'll do
00:01:57
everything in my power, me and my team, to make sure that this show is better for you every single week. We'll listen to your feedback. We'll find the guests
00:02:03
that you want me to speak to and we'll continue to do what we do. Thank you so much.
00:02:11
Jeffrey Hinsson, they call you the godfather of AI. Uh yes they do. Why do they call you
00:02:18
that? There weren't that many people who believed that we could make neural networks work, artificial neural
00:02:24
networks. So for a long time in AI from the 1950s onwards, there were kind of
00:02:31
two ideas about how to do AI. One idea was that sort of core of human
00:02:36
intelligence was reasoning. And to do reasoning, you needed to use some form of logic. And so AI had to be based
00:02:43
around logic. And in your head, you must have something like symbolic expressions
00:02:48
that you manipulated with rules. And that's how intelligence worked. And things like learning or reasoning by
00:02:55
analogy, that all come later once we've figured out how basic reasoning works. There was a different approach, which is
00:03:01
to say, let's model AI on the brain because obviously the brain makes us
00:03:06
intelligent. So simulate a network of brain cells on a computer and try and
00:03:13
figure out how you would learn strengths of connections between brain cells so that it learned to do complicated things
00:03:19
like recognize objects in images or recognize speech or even do reasoning. I
00:03:24
pushed that approach for like 50 years because so few people believed in it.
00:03:30
There weren't many good universities that had groups that did that. So if you
00:03:35
did that the best young students who believed in that came and worked with you. So I was very fortunate in getting
00:03:41
a whole lot of really good students some of which have gone on to create and play
00:03:46
an instrumental role in creating platforms like open AI. Yes. So I sus
00:03:52
a nice example a whole bunch of them. Why did you believe that modeling it off the brain was a more effective approach?
00:03:59
It wasn't just me believed it early on. Fonoyman believed it and Cheuring
00:04:04
believed it and if either of those had lived I think AI would have had a very different history but they both died
00:04:10
young. You think AI would have been here sooner? I think neural net the neural
00:04:15
net approach would have been accepted much sooner if either of them had lived
00:04:20
in this season of your life. What mission are you on? My main mission now
00:04:26
is to warn people how dangerous AI could be. Did you know that when you became
00:04:33
the godfather of AI? No, not really. I was quite slow to understand some of the
00:04:38
risks. Some of the risks were always very obvious, like people would use AI to make autonomous lethal weapons. That
00:04:44
is things that go around deciding by themselves who to kill. Other risks, like the idea that they would one day
00:04:50
get smarter than us and maybe would become irrelevant, I was slow to
00:04:56
recognize that. Other people recognized it 20 years ago. I only recognized it a
00:05:01
few years ago that that was a real risk that was come might be coming quite soon. How could you not have foreseen
00:05:07
that if if with everything you know here about cracking the ability for these computers to learn similar to how humans
00:05:14
learn and just you know introducing any rate of improvement? It's a very good question. How could you not have seen
00:05:20
that? But remember neural networks 20 30 years ago were very primitive in what
00:05:27
they could do. They were nowhere near as good as humans, but things like vision and language and speech recognition. The
00:05:33
idea that you have to now worry about it getting smarter than people, that seems silly then. When did that change? It
00:05:39
changed for the general population when chat GPT came out. It changed for me
00:05:46
when I realized that the kinds of digital intelligences we're making have
00:05:51
something that makes them far superior to the kind of biological intelligence we have. If I want to share information
00:05:57
with you, so I go off and I learn something and I'd like to tell you what I learned. So I produce some sentences.
00:06:04
This is a rather simplistic model, but roughly right. Your brain is trying to figure out how can I change the strength of connections between neurons. So I
00:06:10
might have put that word next. And so you'll do a lot of learning when a very surprising word comes and not much
00:06:15
learning when if it's when it's very obvious word. If I say fish and chips, you don't do much learning when I say
00:06:21
chips. But if I say fish and cucumber, you do a lot more learning. You wonder why did I say cucumber? So that's
00:06:27
roughly what's going on in your brain. I'm predicting what's coming next. That's how we think it's working. Nobody
00:06:33
really knows for sure how the brain works. And nobody knows how it gets the information about whether you should
00:06:39
increase the strength of a connection or decrease the strength of a connection. That's the crucial thing. But what we do
00:06:44
know now from AI is that if you could get information
00:06:49
about whether to increase or decrease the connection strength so as to do better at whatever task you're trying to
00:06:55
do, then we could learn incredible things because that's what we're doing now with artificial neuronets.
00:07:01
It's just we don't know for real brains how they get that signal about whether to increase or decrease.
00:07:06
As we sit here today, what are the big concerns you have around safety of AI? if we were to to list the the top couple
00:07:14
that are really front of mind and that we should be thinking about. Um, can I have more than a couple? Go ahead. I'll
00:07:19
write them all down and we'll go through them. Okay. First of all, I want to make a distinction between two completely
00:07:25
different kinds of risk. There's risks that come from people misusing AI. Yeah. And that's most of
00:07:32
the risks and all of the short-term risks. And then there's risks that come
00:07:37
from AI getting super smart and deciding it doesn't need us. Is that a real risk?
00:07:43
And I talk mainly about that second risk because lots of people say, "Is that a real risk?" And yes, it is. Now, we
00:07:51
don't know how much of a risk it is. We've never been in that situation before. We've never had to deal with things smarter than us. So really, the
00:07:58
thing about that existential threat is that we have no idea how to deal with
00:08:04
it. We have no idea what it's going to look like. And anybody who tells you they know just what's going to happen and how to deal with it, they're talking
00:08:10
nonsense. So, we don't know how to estimate the probabil probabilities it'll replace us. Um, some people say
00:08:17
it's like less than 1%. My friend Yan Lar who was a postto with me thinks no no no, we're always going to be we build
00:08:24
these things. We're always going to be in control. We'll build them to be obedient. And other people like Yudkowski say,
00:08:33
"No, no, no. These things are going to wipe us out for sure. If anybody builds it, it's going to wipe us all out." And
00:08:38
he's confident of that. I think both of those positions are extreme. It's very hard to estimate the probabilities in
00:08:45
between. If you had to bet on who was right out of your two friends,
00:08:51
I simply don't know. So, if I had to bet, I'd say the probabilities in between, and I don't know where to
00:08:57
estimate it in between. I often say 10 to 20% chance they'll wipe us out, but that's just gut based on the idea that
00:09:05
we're we're still making them and we're pretty ingenious. And the hope is that
00:09:10
if enough smart people do enough research with enough resources, we'll figure out a way to build them so
00:09:15
they'll never want to harm us. Sometimes I think if we we talk about that second
00:09:21
um path, sometimes I think about nuclear bombs and the the invention of the atomic bomb and how it compares like how
00:09:27
is this different because the atomic bomb came along and I imagine a lot of people at that time thought our days are numbered. Yes, I was there. We did.
00:09:35
Yeah. But but but what's what h we're still here. We're still here. Yes. So
00:09:40
the atomic bomb was really only good for one thing and it was very obvious how it worked. Even if you hadn't had the
00:09:47
pictures of Hiroshima and Nagasaki, it was obvious that it was a very big bomb
00:09:53
that was very dangerous. With AI, it's good for many, many things. It's
00:10:00
going to be magnificent in healthcare and education and more or less any industry that needs to use its data is
00:10:07
going to be able to use it better with AI. So, we're not going to stop the development.
00:10:13
You know, people say, "Well, why don't we just stop it now?" We're not going to stop it because it's too good for too
00:10:19
many things. Also, we're not going to stop it because it's good for battle robots, and none of the countries that
00:10:25
sell weapons are going to want to stop it. Like the European regulations, they
00:10:30
have some regulations about AI, and it's good they have some regulations, but they're not designed to deal with most of the threats. And in particular, the
00:10:38
European regulations have a a clause in them that say none of these regulations apply to military uses of AI.
00:10:45
So governments are willing to regulate regulate companies and people, but they're not willing to regulate
00:10:51
themselves. It seems pretty crazy to me that they I go back and forward, but if Europe has a
00:10:58
regulation, but the rest of the world doesn't competitive disadvantage. Yeah, we're
00:11:03
seeing this already. I don't think people realize that when OpenAI release a new model or a new piece of software in America, they can't release it to
00:11:11
Europe yet because of regulations here. So Sam Alman tweeted saying, "Our new AI agent thing is available to everybody,
00:11:16
but it can't come to Europe yet because there's regulations." Yes. What does that gives us a
00:11:22
productive disadvantage? Productivity disadvantage. What we need is I mean at
00:11:27
this point in history when we're about to produce things more intelligent than ourselves, what we really need is a kind
00:11:34
of world government that works run by intelligent, thoughtful people. And that's not what we got.
00:11:40
So free-for-all. Well, that what we've got is sort of
00:11:47
we've got capitalism which is done very nicely by us. is produce lots of goods goods and services for us. But these big
00:11:55
companies, they're legally required to try and maximize profits and that's not
00:12:01
what you want from the people developing this stuff. So let's do the risks then. You talked
00:12:06
about there's human risks and then there's So I've distinguished these two kinds of risk. Let's talk about all the
00:12:11
risks from bad human actors using AI. There's cyber attacks.
00:12:18
So between 2023 and 2024, they increased by about a factor of
00:12:24
12,200%. And that's probably because these large
00:12:29
language models make it much easier to do fishing attacks. And a fishing attack for anyone that doesn't know is it's
00:12:36
they send you something saying, uh, hi, I'm your friend John and I'm stuck in El
00:12:41
Salvador. Could you just wire this money? That's one kind of attack. But the fishing attacks are really trying to
00:12:47
get your loon credentials. And now with AI, they can clone my voice, my image. They can do all that. I'm struggling at
00:12:53
the moment because there's a bunch of AI scams on X and also Meta. And there's one in particular on Meta, so Instagram,
00:12:59
Facebook at the moment, which is a paid advert where they've taken my voice from the podcast. They've taken the my
00:13:04
mannerisms and they've made a new video of me encouraging people to go and take part in this crypto Ponzi scam or
00:13:11
whatever. And we've been, you know, we spent weeks and weeks and weeks and weeks and end emailing Meta telling, "Please take this down." They take it
00:13:17
down, another one pops up. They take that one down, another one pops up. So, it's like whack-a-ole. And then it's very annoying. The the heartbreaking
00:13:23
part is you get the messages from people that have fallen for the scam and they've lost £500 or $500 and they cross
00:13:28
with you cuz you recommended it and I'm I'm like I'm sad for them. It's very annoying. Yeah. I have a a smaller
00:13:34
version of that which is PE some people now publish papers with me as one of the authors. Mhm. And it looks like it's in
00:13:41
order that they can get lots of citations to themselves. Ah, so cyber attacks a very real threat. There's been
00:13:48
an explosion of those. And these already obviously AI is very patient. So they
00:13:53
can go through 100 million lines of code looking for known ways of attacking them. That's easy to do. But they're
00:14:00
going to get more creative and they may some people believe and I some people
00:14:06
who know a lot believe that maybe by 2030 they'll be creating new kinds of
00:14:12
cyber attacks which no person ever thought of. So that's very worrisome
00:14:18
because they can think for themselves and discover they can think for themselves. They can draw new conclusions from much more data than a
00:14:24
person ever saw. Is there anything you're doing to protect yourself from cyber attacks at all? Yes. It's one of
00:14:32
the few places where I changed what I do radically because I'm scared of cyber attacks. Canadian banks are extremely
00:14:39
safe. In 2008, no Canadian banks came anywhere near going bust. So, they're
00:14:45
very safe banks because they're well regulated, fairly well regulated. Nevertheless, I think a cyber attack
00:14:51
might be able to bring down a bank. Now, if you have all my savings are in shares
00:14:57
in banks held by banks, so if the bank gets attacked and it holds your shares,
00:15:04
they're still your shares. And so, I think you'd be okay unless the attacker
00:15:11
sells the shares because the bank can sell the shares. If the attacker sells your shares, I think you're screwed. I
00:15:18
don't know. I mean, maybe the bank would have to try and reimburse you, but the bank's bust by now, right? So,
00:15:24
So I'm worried about a Canadian bank being taken down by a cyber attack and
00:15:29
the attacker selling selling shares that it holds. So I spread my money and my
00:15:34
children's money between three banks in the belief that if a cyber attack takes
00:15:39
down one Canadian bank, the other Canadian banks will very quickly get very careful. And do you have a phone
00:15:47
that's not connected to the internet? Do you have any like, you know, I'm thinking about storing data and stuff like that. Do you think it's wise to
00:15:53
consider having cold storage? I have a little disc drive and I back up my
00:15:58
laptop on this hard drive. So I actually have everything on my laptop on a hard
00:16:03
drive. At least you know if the whole internet went down I had the sense I still got it on my laptop and I still
00:16:09
got my information. Okay. Then the next thing is using AI to create nasty
00:16:16
viruses. Okay. And the problem with that is that
00:16:22
just requires one crazy guy with the grudge. One guy who knows a little bit
00:16:27
of molecular biology, knows a lot about AI, and just wants to destroy the world.
00:16:33
You can now create new viruses relatively cheaply using AI.
00:16:39
And you don't have to be a very skilled molecular biologist to do it. And that's very scary. So you could have a small
00:16:44
cult, for example. a small cult might be able to raise a few million dollars. For a few million
00:16:51
dollars, they might be able to design a whole bunch of viruses. Well, I'm thinking about some of our foreign adversaries doing government funded
00:16:58
programs. I mean, there was lots of talk around COVID and Woo the Wuhan laboratory and what they were doing and gain a function research, but I'm
00:17:04
wondering if in, you know, a China or a Russia or an Iran or something, the government could fund a program for a
00:17:10
small group of scientists to make a virus that they could, you know, I think they could. Yes. Now, they'd be worried
00:17:16
about retaliation. They'd be worried about other governments doing the same to them. Hopefully, that would help keep it under control. They might also be
00:17:22
worried about the virus spreading to their country. Okay? Then there's um
00:17:28
corrupting elections. So, if you wanted to use AI to corrupt
00:17:33
elections, a very effective thing is to be able to do targeted political advertisements
00:17:40
where you know a lot about the person. So anybody who wanted to use AI for
00:17:47
corrupting elections would try and get as much data as they could about
00:17:52
everybody in the electorate. With that in mind, it's a bit worrying what Musk is doing at present in the States, going
00:17:59
in and insisting on getting access to all these things that were very carefully siloed. The claim is it's to
00:18:05
make things more efficient, but it's exactly what you would want if you intended to corrupt the next election.
00:18:10
How do you mean? Because you get all this data on the people. You get all this data on people. You know how much they make where they you know everything
00:18:16
about them. Once you know that, it's very easy to manipulate them because you can make an AI that you can send
00:18:23
messages um that they'll find very convincing telling them not to vote, for example.
00:18:29
So, I have no no reason other than common sense to think this, but I
00:18:35
wouldn't be surprised if part of the motivation of getting all this data from American government sources is to
00:18:42
corrupt elections. Another part might be that it's very nice training data for a big model, but he would have to be
00:18:49
taking that data from the government and feeding it into his Yes. And what they've done is turned off lots of the
00:18:55
security controls, got rid of the some of the organization to protect against that. Um, so that's corrupting
00:19:02
elections. Okay. Then there's creating these two echo chambers
00:19:09
by organizations like YouTube and Facebook showing people things that
00:19:16
will make them indignant. People love to be indignant. Indignant as in angry or
00:19:22
what does indignant mean? Feeling I'm sort of angry but feeling righteous.
00:19:27
Okay. So, for example, if you were to show me something that said Trump did
00:19:33
this crazy thing, here's a video of Trump doing this completely crazy thing. I would immediately click on it.
00:19:40
Okay. So, putting us in echo chambers and dividing us. Yes. And that's um the
00:19:45
policy that YouTube and Facebook and others use for deciding what to show you
00:19:51
next is causing that. If they had a policy of showing you balanced things,
00:19:58
they wouldn't get so many clicks and they wouldn't be able to sell so many advertisements. And so it's basically the profit motive
00:20:04
is saying show them whatever will make them click. And what'll make them click is things that are more and more
00:20:11
extreme. And that confirmed my existing bias. That confirm my existing bias. So you're getting your biases confirmed all
00:20:17
the time further and further and further and further, which means you're you're driving away, which is now there's in the states there's two communities that
00:20:23
don't hardly talk to each other. I'm not sure people realize that this is actually happening every time they open an app. But if you go on a Tik Tok or a
00:20:29
YouTube or one of these big social networks, the algorithm, as you you said, is designed to show you more of
00:20:35
the things that you had interest in last time. So, if you just play that out over 10 years, it's going to drive you
00:20:41
further and further and further into whatever ideology or belief you have and further away from nuance and common
00:20:46
sense and um parity, which is a pretty remarkable thing. I I like people don't
00:20:52
know it's happening. They just open their phones and experience something and think this is the news or the
00:20:59
experience everyone else is having. Right. So, basically, if you have a newspaper and everybody gets the same
00:21:04
newspaper, Yeah. you get to see all sorts of things you weren't looking for and you get a sense that if it's in the
00:21:10
newspaper it's an important thing or significant thing but if you have your own news feed my news feed on my iPhone
00:21:18
3/arters of the stories are about AI and I find it very hard to know if the whole
00:21:23
world's talking about AI all the time or if it's just my newsfeed
00:21:28
okay so driving me into my echo chambers um which is going to continue to divide us further and further I'm actually
00:21:34
noticing that the algorithm are becoming even more, what's the word?
00:21:40
Tailored. And people might go, "Oh, that's great." But what it means is they're becoming even more personalized, which is means that my reality is
00:21:47
becoming even further from your reality. Yeah. It's crazy. We don't have a shared reality anymore. I share reality with
00:21:54
other people who watch the BBC and other BBC news and other people who read the Guardian and other people who read the
00:22:00
New York Times. I have almost no shared reality with people who watch Fox News.
00:22:07
It's pretty It's pretty um I I It's worrisome. Yeah. Behind all this is the
00:22:13
idea that these companies just want to make profit and they'll do whatever it takes to make more profit because they
00:22:19
have to. They're legally obliged to do that. So, we almost can't blame the
00:22:24
company, can we? If they're if Well, capitalism's done very well for us. It's
00:22:30
produced lots of goodies. Yeah. But you need to have it very well regulated. So what you really want is to have rules
00:22:37
so that when some company is trying to make as much profit as possible,
00:22:43
in order to make that profit, they have to do things that are good for people in general, not things that are bad for
00:22:48
people in general. So once you get to a situation where in order to make more profit the company starts doing things
00:22:54
that are very bad for society like showing you things that are more and more extreme that's what regulations are
00:23:00
for. So you need regulations with capitalism. Now companies will always
00:23:06
say regulations get in the way make us less efficient and that's true. The
00:23:11
whole point of regulations is to stop them doing things to make profit that hurt society. And we need strong
00:23:16
regulation. who's going to decide whether it hurts society or not because you know that's the job of politicians
00:23:23
unfortunately if the politicians are owned by the companies that's not so good and also the politicians might not understand the technology we you've
00:23:29
probably seen the Senate hearings where they wheel out you know Mark Zuckerberg and these big tech CEOs and it is quite embarrassing because they're asking the
00:23:36
wrong questions well I've seen the video of the US education secretary talking
00:23:42
about how they're going to get AI in the classrooms except she thought it was called A1
00:23:48
She's actually there saying we're going to have all the kids interacting with A1. There is a school system that's
00:23:54
going to start um making sure that first graders or even preks have A1 teaching,
00:24:01
you know, every year starting, you know, that far down in the grades. And that's just a that's a wonderful thing.
00:24:07
[Laughter] And these are what these are the people
00:24:12
that these are the people in charge. Ultimately the tech companies are in charge because they will outsmart the
00:24:18
tech companies in the states now at least a few weeks ago when I was there
00:24:23
they were running an advertisement about how it was very important not to regulate AI because it would hurt us in
00:24:30
the competition with China. Yeah. And that's a that's a plausible argument there. Yes it will. But you have to
00:24:36
decide, do you want to compete with China by doing things that will do a lot
00:24:44
of harm to your society? And you probably don't.
00:24:49
I guess they would say that it's not just China, it's Denmark and Australia and Canada and the UK. They're not so
00:24:56
worried about and Germany. But if they kneecap themselves with regulation, if they slow themselves down, then the
00:25:01
founders, the entrepreneurs, the investors are going to go. I think calling it kneecapping is taking a particular point of view is take taking
00:25:08
the point of view that regulations are sort of very harmful. What you need to do is just constrain the big companies
00:25:13
so that in order to make profit, they have to do things that are socially useful. Like Google search is a great
00:25:20
example that didn't need regulation because it just made information available to people. It was great. But
00:25:27
then if you take YouTube which starts showing you adverts and showing you more and more extreme things that needs
00:25:33
regulation but we don't have the people to regulate it as we've identified. I think people know pretty well um that
00:25:41
particular problem of showing you more and more extreme things. That's a well-known problem that the politicians understand. They just um need to get on
00:25:49
and regulate it. So that was the the next point which was that the algorithms are going to drive us further into our
00:25:54
echo chambers, right? What's next? Lethal autonomous weapons.
00:26:00
Lethal autonomous weapons. That means things that can kill you and
00:26:05
make their own decision about whether to kill you, which is the great dream, I guess, of the military-industrial
00:26:12
complex being able to create such weapons. So, the worst thing about them is big powerful countries always have
00:26:20
the ability to invade smaller poorer countries. they're just more powerful.
00:26:26
But if you do that using actual soldiers, you get bodies coming back in
00:26:31
bags and the relatives of the soldiers who were killed don't like it. So you
00:26:37
get something like Vietnam. Mhm. In the end, there's a lot of protest at home. If instead of bodies coming back in
00:26:44
bags, it was dead robots, there'd be much less protest and the
00:26:49
military-industrial complex would like it much more because robots are expensive. And suppose you had something
00:26:56
that could get killed and was expensive to replace. That would be just great.
00:27:01
Big countries can invade small countries much more easily because they don't have their soldiers being killed. And the
00:27:07
risk here is that these robots will malfunction or they'll just be more No,
00:27:14
no, that's even if the robots do exactly what the people who built the robots want them to do, the risk is that it's
00:27:20
going to make big countries invade small countries more often. More often because they can Yeah. And it's not a nice thing to do. So it brings down the friction of
00:27:26
war. It brings down the cost of doing an invasion. And these machines will be smarter at
00:27:32
warfare as well. So they'll be well even when the machines aren't smarter. So the lethal autonomous weapons, they can make
00:27:39
them now. And they I think all the big defense models are busy making them.
00:27:44
Even if they're not smarter than people, are still very nasty, scary things. Cuz I'm thinking that, you know, they could
00:27:50
show just a picture. Go get this guy. Yeah. And go take out anyone he's been
00:27:55
texting and this little wasp. So, two days ago, I was visiting a friend of mine in Sussex who had a drone that cost
00:28:03
less than £200 and the drone went up. It took a good look
00:28:09
at me and then it could follow me through the woods and it follow It was
00:28:14
very spooky having this drone. It was about 2 meters behind me. It was looking at me and if I moved over there, it
00:28:20
moved over there. It could just track me. Mhm. For 200 pounds, but it was already quite spooky. Yeah. And I
00:28:26
imagine there's as you say a race going on as we speak to who can build the most complex autonomous autonomous weapons.
00:28:33
There is a a risk I often hear that some of these things will combine and the cyber attack will release weapons.
00:28:41
Sure. Um you can you can get combinatorily many risks by combining these other risks. Mhm. So, I mean, for
00:28:49
example, you could get a super intelligent AI that decides to get rid of people, and the obvious way to do
00:28:56
that is just to make one of these nasty viruses. If you made a virus that was
00:29:01
very contagious, very lethal, and very slow, everybody would have it before they
00:29:07
realized what was happening. I mean, I think if a super intelligence wanted to get rid of us, it will probably go for
00:29:13
something biological like that that wouldn't affect it. Do you not think it could just very quickly turn us against each other? For example, it could send a
00:29:19
warning on the nuclear systems in America that there's a nuclear bomb coming from Russia or vice versa and one
00:29:26
retaliates. Yeah. I mean, my basic view is there's so many ways in which the super intelligence could get rid of us.
00:29:32
It's not worth speculating about. What What is What you have to do is
00:29:38
prevent it ever wanting to. That's what we should be doing research on. There's no way we're going to prevent it from
00:29:45
it's smarter than us, right? There's no way we're going to prevent it getting rid of us if it wants to. We're not used
00:29:50
to thinking about things smarter than us. If you want to know what life's like when you're not the apex intelligence,
00:29:58
ask a chicken. Yeah. I was thinking about my dog Pablo,
00:30:04
my French bulldog, this morning as I left home. He has no idea where I'm going. He has no idea what I do, right?
00:30:10
Can't even talk to him. Yeah. And the g the intelligence gap will be like that. So you're telling me that if I'm Pablo,
00:30:17
my French bulldog, I need to figure out a way to make my owner not wipe me out.
00:30:24
Yeah. So we have one example of that which is mothers and babies. Evolution
00:30:30
put a lot of work into that. Mothers are smarter than babies, but babies are in control. And they're in control because
00:30:35
the mother just can't bear lots of hormones and things, but the b the mother just can't bear the sound of the
00:30:41
baby crying. Not all mothers. Not all mothers. And then the baby's not in control and then bad things happen. We
00:30:48
somehow need to figure out how to make them not want to take over. The analogy
00:30:53
I often use is forget about intelligence, think about physical strength. Suppose you have a nice little
00:30:59
tiger cup. It's sort of bit bigger than a cat. It's really cute. It's very cuddly, very interesting to
00:31:06
watch. Except that you better be sure that when it grows up, it never wants to kill you. Cuz if it ever wanted to kill
00:31:12
you, you'd be dead in a few seconds. And you're saying the AI we have now is the target cub. Yep. And it's growing up.
00:31:19
Yep. So, we need to train it as it's when it's a baby. Well, now a tiger has lots
00:31:24
of in stuff built in. So, you know, when it grows up, it's not a safe thing to have around. But lions, people that have
00:31:31
lions as pets, yes. Sometimes the lion is affectionate to its creator but not to others. Yes. And we don't know
00:31:37
whether these AIs we we simply don't know whether we can make them not want to take over and not
00:31:44
want to hurt us. Do you think we can? Do you think it's possible to train super intelligence? I don't think it's clear
00:31:49
that we can. So I think it might be hopeless. But I also think we might be
00:31:55
able to. And it'd be sort of crazy if people went extinct cuz we couldn't be bothered to try. If that's even a
00:32:02
possibility, how do you feel about your life's work? Because you were Yeah. Um,
00:32:08
it sort of takes the edge off it, doesn't it? I mean, the idea is going to be wonderful in healthcare and wonderful
00:32:13
in education and wonderful. I mean, it's going to make call centers much more efficient, though one worries a bit
00:32:19
about what the people who are doing that job now do. It makes me sad. I don't feel particularly guilty about
00:32:25
developing AI like 40 years ago because
00:32:30
at that time we had no idea that this stuff was going to happen this fast. We thought we had plenty of time to worry
00:32:36
about things like that. They when you when you can't get the to do much, you want to get it to do a little bit more.
00:32:41
You don't worry about this stupid little thing is going to take over from people. You just want it to be able to do a
00:32:47
little bit more of the things people can do. It's not like I knowingly did something thinking this might wipe us
00:32:54
all out, but I'm going to do it anyway. Mhm. But it is a bit sad that it's not
00:33:00
just going to be something for good. So I feel I have a duty now to talk about the risks.
00:33:07
And if you could play it forward and you could go forward 30, 50 years and you found out that it led to the extinction of humanity and if that does end up
00:33:14
being being the outcome,
00:33:20
well, if you played it forward and it led to the extinction of humanity, I would use that to tell people to tell
00:33:28
their governments that we really have to work on how we're going to keep this stuff under control. I think we need
00:33:34
people to tell governments that governments have to force the companies to use their resources to work on safety
00:33:41
and they're not doing much of that because you don't make profits that way. One of your your students we talked
00:33:46
about earlier um Ilia Yep. Ilia left OpenAI. Yep. And there was lots of
00:33:53
conversation around the fact that he left because he had safety concerns. Yes. And he's gone on to set set up a AI
00:34:01
safety company. Yes. Why do you think he left?
00:34:06
I think he left because he had safety concerns. Really? He um I still have lunch with him from time to time. His
00:34:13
parents live in Toronto. When he comes to Toronto, we have lunch together. He doesn't talk to me about what went on at Open AI, so I have no inside information
00:34:19
about that. But I know I very well and he is genuinely concerned with safety.
00:34:24
So I think that's why he left because he was one of the top people. I mean he was he was probably the most important
00:34:30
person behind the development of um church GPT the the early versions like GPT2 he was very important in the
00:34:37
development of that you know him personally so you know his character yes he has a good moral compass he's not
00:34:43
like someone like Musco has no moral compass does Sam Alman have a good moral compass
00:34:50
we'll see I don't know Sam so I don't want to
00:34:56
comment on that. But from what you've seen, are you concerned about the actions that they've taken? Because if
00:35:02
you know Ilia and Ilia's a good guy and he's left that would give you some insight. Yes.
00:35:08
It would give you some reason to believe that there's a problem there. And if you look at Sam's statements
00:35:15
some years ago, he sort of happily said in one interview
00:35:20
and this stuff will probably kill us all. That's not exactly what he said, but that's what it amounted to. Now he's
00:35:25
saying you don't need to worry too much about it. And I suspect that's not driven by
00:35:32
seeking after the truth. That's driven by seeking after money. Is it money or
00:35:38
is it power? Yeah. I shouldn't have said money. It's some some combination of those. Yes. Okay. I guess money is a
00:35:44
proxy for power. But I am I've got a friend who's a billionaire and he is in
00:35:50
those circles. And when I went to his house and had uh lunch with him one day, he knows lots of people in AI, building
00:35:55
the biggest AI companies in the world. And he gave me a cautionary warning across the across his kitchen table in
00:36:01
London where he gave me an insight into the private conversations these people have, not the media interviews they do
00:36:06
where they talk about safety and all these things, but actually what some of these individuals think is going to happen and what do they think is going
00:36:13
to happen. It's not what they say publicly. You know, one one person who I
00:36:19
shouldn't name who is the who is leading one of the biggest AI companies in the world. He told me that he knows this person very well and he privately thinks
00:36:25
that we're heading towards this kind of dystopian world where we have just huge amounts of free time. We don't work
00:36:31
anymore. And this person doesn't really give a [ __ ] about the harm that it's going to have on the world. And this
00:36:36
person who I'm referring to is building one of the biggest AI companies in the world. And I then watch this person's interviews online trying to figure out
00:36:42
which of three people it is. Yeah. Well, it's one of those three people. Okay. And I watch this person's interviews online and I I reflect on a conversation
00:36:48
that my billionaire friend had with me who knows him and I go, "Fucking hell, this guy's lying publicly." Like, he's
00:36:54
not telling the the truth to the world. And that's haunted me a little bit. It's part of the reason I have so many conversations around AR in this podcast
00:36:59
because I'm like, I don't know if they're I think they're a some of them are a little bit sadistic about power. I
00:37:06
think they they like the idea that they will change the world, that they will be the one that fundamentally shifts the
00:37:14
world. I think Musk is clearly like that, right? He's such a complex character that I
00:37:21
don't I don't really know how to place Musk. Um he's done some really good things like um pushing electric cars.
00:37:28
That was a really good thing to do. Yeah. Some of the things he said about self-driving were a bit exaggerated, but
00:37:33
he that was a really useful thing he did. Giving the Ukrainians communication
00:37:38
during the war with Russia. Stling. Um that was a really good thing he did. there's a bunch of things like that. Um,
00:37:45
but he's also done some very bad things. So, coming back to this point of
00:37:53
the possibility of destruction and the motives of these big companies,
00:38:01
are you at all hopeful that anything can be done to slow down the pace and acceleration of AI? Okay, there's two
00:38:07
issues. One is can you slow it down? Yeah. And the other is, can you make it so it will be safe in the end? It won't
00:38:15
wipe us all out. I don't believe we're going to slow it down. Yeah. And the reason I don't believe we're going to
00:38:21
slow it down is because there's competition between countries and competition between companies within a
00:38:26
country and all of that is making it go faster and faster. And if the US slowed it down, China wouldn't slow it down.
00:38:34
Does IA think it's possible to make AI safe?
00:38:40
I think he does. He won't tell me what his secret source is. I I'm not sure how
00:38:45
many people know what his secret source is. I think a lot of the investors don't know what his secret source is, but they've given him billions of dollars
00:38:51
anyway because they have so much faith in Asia, which isn't foolish. I mean, he
00:38:56
was very important in Alexet, which got object recognition working well. He was the main the main force behind the
00:39:05
things like GBC2 which then led to CH GPT.
00:39:10
So I think having a lot of faith in IA is a very reasonable decision. There's something quite haunting about the guy
00:39:16
that made and was the main force behind GPT2 which led rise to this whole revolution left the company because of
00:39:24
safety reasons. He knows something that I don't know about what might happen next. Well, the company had now I don't
00:39:32
know the precise details um but I'm fairly sure the company had indicated that would it would use a significant
00:39:38
fraction of its resources of the compute time for doing safety research and then
00:39:44
it kept then it reduced that fraction. I think that's one of the things that happened. Yeah, that was reported publicly. Yes. Yeah.
00:39:51
We've gotten to the autonomous weapons part of the risk framework. Right. So
00:39:57
the next one is joblessness. Yeah. In the past, new technologies have come in
00:40:03
which didn't lead to joblessness. New jobs were created. So the classic example people use is automatic tele
00:40:09
machines. When automatic tele machines came in, a lot of bank tellers didn't lose their jobs. They just got to do
00:40:16
more interesting things. But here, I think this is more like when they got
00:40:21
machines in the industrial revolution. And you can't have a job digging ditches now
00:40:28
because a machine can dig ditches much better than you can. And I think for mundane intellectual labor, AI is just
00:40:36
going to replace everybody. Now, it will may well be in the form of you have
00:40:43
fewer people using air assistance. So it's a combination of a person and an AI assistant are now doing the work that 10
00:40:50
people could do previously. People say that it will create new jobs though, so we'll be fine. Yes. And that's been the
00:40:57
case for other technologies, but this is a very different kind of technology. If it can do all mundane human intellectual
00:41:03
labor, then what new jobs is it going to create? You'd you'd have to be very
00:41:09
skilled to have a job that it couldn't just do. So I don't I don't think they're right. I think you can try and
00:41:15
generalize from other technologies that have come in like computers or automatic tele machines, but I think this is
00:41:22
different. People use this phrase. They say AI won't take your job. A human using AI will take your job. Yes, I
00:41:28
think that's true. But for many jobs, that'll mean you need far fewer people.
00:41:34
My niece answers letters of complaint to a health service. It used to take her 25
00:41:39
minutes. She'd read the complaint and she'd think how to reply and she'd write a letter. And now she just scans it into
00:41:47
um a chatbot and it writes the letter. She just checks the letter. Occasionally
00:41:53
she tells it to revise it in some ways. The whole process takes her five minutes. That means she can answer five
00:42:00
times as many letters and that means they need five times fewer of her so she
00:42:06
can do the job that five of her used to do. Now, that will mean they need less
00:42:13
people. In other jobs, like in health care, they're much more elastic. So, if
00:42:18
you could make doctors five times as efficient, we could all have five times as much health care for the same price,
00:42:24
and that would be great. There's there's almost no limit to how much health care people can absorb. They always want more
00:42:31
healthare if there's no cost to it. There are jobs where you can make a
00:42:36
person with an AI assistant much more efficient and you won't lead to less people because you'll just have much
00:42:42
more of that being done. But most jobs I think are not like that. Am I right in
00:42:48
thinking the sort of industrial revolution played a role in replacing muscles? Yes.
00:42:53
Exactly. And this revolution in AI replaces intelligence the brain. Yeah. So, so mundane intellectual labor is
00:43:00
like having strong muscles and it's not worth much anymore. So, muscles have
00:43:05
been replaced. Now we intelligence is being replaced. Yeah. So, what remains?
00:43:11
Maybe for a while some kinds of creativity but the whole idea of super intelligence is nothing remains. Um
00:43:18
these things will get to be better than us at everything. So, what what do we end up doing in such a world? Well, if
00:43:23
they work for us, we end up getting lots of goods and services for not much
00:43:29
effort. Okay. But that sounds tempting and nice, but I don't know. There's a
00:43:35
cautionary tale in creating more and more ease for humans in in it going badly. Yes. And we need to figure out if
00:43:43
we can make it go well. So the the nice scenario is imagine a company with a CEO
00:43:50
who is very dumb, probably the son of the former CEO. And he has an executive
00:43:56
assistant who's very smart and he says,
00:44:01
"I think we should do this." And the executive assistant makes it all work. The CEO feels great. He doesn't
00:44:07
understand that he's not really in control. And in in some sense, he is in control. He suggests what the company
00:44:14
should do. She just makes it all work. Everything's great. That's the good scenario. And the bad scenario, the bad
00:44:21
scenario, she thinks, "Why do we need him?" Yeah.
00:44:26
I mean, in a world where we have super intelligence, which you don't believe is that far away. Yeah, I think it might
00:44:32
not be that far away. It's very hard to predict, but I think we might get it in like 20 years or even less. I made the
00:44:39
biggest investment I've ever made in a company because of my girlfriend. I came home one night and my lovely girlfriend
00:44:46
was up at 1:00 a.m. in the morning pulling her hair out as she tried to piece together her own online store for
00:44:53
her business. And in that moment, I remembered an email I'd had from a guy called John, the founder of Stanto, our
00:44:59
new sponsor and a company I've invested incredibly heavily in. And Standtore helps creators to sell digital products,
00:45:05
courses, coaching, and memberships all through a simple customizable link in bio system. And it handles everything,
00:45:12
payments, bookings, emails, community engagement, and even links with Shopify. And I believe in it so much that I'm
00:45:18
going to launch a Stan challenge. And as part of this challenge, I'm going to
00:45:23
give away $100,000 to one of you. If you want to take part in this challenge, if you want to monetize the knowledge that
00:45:29
you have, visit stephenbartlet.stan stan.store to sign up. And you'll also
00:45:34
get an extended 30-day free trial of Stan Store if you use that link. Your next move could quite frankly change
00:45:41
everything. Because I talked about ketosis on this podcast and ketones, a brand called Ketone IQ sent me their
00:45:47
little product here and it was on my desk when I got to the office. I picked it up. It sat on my desk for a couple of weeks. Then one day, I tried it and
00:45:55
honestly, I have not looked back ever since. I now have this everywhere I go
00:46:01
when I travel all around the world. It's in my hotel room. My team will put it there. Before I did the podcast recording today that I've just finished,
00:46:07
I had a shot of Ketone IQ. And as is always the case when I fall in love with a product, I called the CEO and asked if
00:46:12
I could invest a couple of million quid into their company. So, I'm now an investor in the company as well as them
00:46:18
being a brand sponsor. I find it so easy to drop into deep focused work when I've had one of these. I would love you to
00:46:24
try one and see the impact it has on you, your focus, your productivity, and your endurance. So, if you want to try
00:46:29
it today, visit ketone.com/stephven for 30% off your subscription. Plus, you'll receive a free gift with your
00:46:35
second shipment. That's ketone.com/stephven. I'm excited for you. I am. So, what's
00:46:42
the difference between what we have now and super intelligence? Because it seems to be really intelligent to me when I use like chatbt3 or Gemini or Okay. So
00:46:50
it's already AI is already better than us at a lot of things in particular areas like chess for example. Yeah. AI
00:46:59
is so much better than us that people will never beat those things again. Maybe the occasional win but basically
00:47:05
they'll never be comparable again. Obviously the same in go in terms of the amount of knowledge they have. Um
00:47:12
something like GBT4 knows thousands of times more than you do. There's a few areas in which your knowledge is better
00:47:19
than its and in almost all areas it just knows more than you do. What areas am I
00:47:25
better than it? Probably in interviewing CEOs. You're probably better at that.
00:47:33
You've got a lot of experience at it. You're a good interviewer. You know a lot about it. If you tried if you got
00:47:39
GPT4 to interview a CEO, probably do a worse job. Okay.
00:47:46
I'm trying to think if that if I agree with that statement. Uh GPT4 I think for sure. Yeah. Um but I but I guess you
00:47:52
could but it may not be long before Yeah. I guess you could train one on this how I ask questions and what I do
00:47:57
and Sure. And if you took a general purpose sort of foundation model and
00:48:02
then you trained it up on not just you but every every interviewer you could find doing interviews like this but
00:48:10
especially you. You'll probably get to be quite good at doing your job but probably not as good as you for a while.
00:48:17
Okay. So, there's a few areas left and then super intelligence becomes when it's better than us at all things. When
00:48:23
it's much smarter than you and almost all things is better than you. Yeah. And you you you say that this might be a
00:48:29
decade away or so. Yeah. It might be. It might be even closer. Some people think
00:48:34
it's even closer and might well be much further. It might be 50 years away. That's still a possibility. It might be
00:48:40
that somehow training on human data limits you to not being much smarter than humans. My guess is between 10 and
00:48:47
20 years we'll have super intelligence. On this point of joblessness, it's something that I've been thinking a lot
00:48:53
about in particular because I started messing around with AI agents and we released an episode on the podcast actually this morning where we had a
00:48:58
debate about AI agents with some a CEO of a big AI agent company and a few other people and it was the first moment
00:49:05
where I had no it was another moment where I had a Eureka moment about what the future might look like when I was
00:49:11
able in the interview to tell this agent to order all of us drinks and then 5 minutes later in the interview you see
00:49:17
the guy show up with the drinks and I didn't touch anything. I just told it to order us drinks to the studio. And you
00:49:23
didn't know about who you normally got your drinks from. It figured that out from the web. Yeah, figured out cuz it went on Uber Eats. It has my my my data,
00:49:30
I guess. And it I we put it on the screen in real time so everyone at home could see the agent going through the internet, picking the drinks, adding a
00:49:36
tip for the driver, putting my address in, putting my credit card details in, and then the next thing you see is the drinks show up. So that was one moment.
00:49:44
And then the other moment was when I used a tool called Replet and I built
00:49:49
software by just telling the agent what I wanted. Yes. It's amazing, right? It's amazing and terrifying at the same time.
00:49:56
Yes. Because and if it can build software like that, right? Yeah. Remember that the AI when it's training
00:50:03
is using code and if it can modify its own code
00:50:08
then it gets quite scary, right? because it can modify. It can change itself in a way we can't change ourselves. We can't
00:50:14
change our innate endowment, right? There's nothing about itself that it couldn't change.
00:50:21
On this point of joblessness, you have kids. I do. And they have kids. No, they don't have kids. No grandkids yet. What
00:50:27
would you be saying to people about their career prospects in a world of super intelligence? What should we we be
00:50:33
thinking about? Um, in the meantime, I'd say it's going to be a long time before it's as good at physical manipulation as
00:50:40
us. Okay. And so, a good bet would be to be a plumber.
00:50:47
until the humanoid robots show up in such a world where there is mass joblessness which is not something that
00:50:53
you just predict but this is something that Sam Alman open AI I've heard him predict and many of the CEOs Elon Musk I
00:50:59
watched an interview which I'll play on screen of him being asked this question and it's very rare that you see Elon Musk silent for 12 seconds or whatever
00:51:05
it was and then he basically says something about he actually is living in suspended disbelief i.e. He's basically
00:51:12
just not thinking about it. When you think about advising your children on a career with so much that is changing,
00:51:18
what do you tell them is going to be of value?
00:51:33
Well, that is a tough question to answer. I would just say, you know, to to sort of
00:51:39
follow their heart in terms of what they they find um interesting to do or fulfilling to do. I mean, if I think
00:51:45
about it too hard, frankly, it can be uh dispariting and uh demotivating. Um
00:51:53
because I mean, I I go through I mean I I I've put a lot of blood, sweat, and
00:51:59
tears into building the companies and then it and then I'm like, wait, should I be doing this? Because if I'm
00:52:06
sacrificing time with friends and family that I would prefer to to to but but then ultimately the AI can do all these
00:52:13
things. Does that make sense? I I don't know. Um to some extent I have to have
00:52:19
deliberate suspension of disbelief in order to to remain motivated. Um so I I
00:52:26
guess I would say just you know
00:52:31
work on things that you find interesting, fulfilling and um and and that contribute uh some good to the rest
00:52:36
of society. Yeah. A lot of these threats it's very hard to intellectually you can
00:52:42
see the threat but it's very hard to come to terms with it emotionally. Yeah. I haven't come to terms with it
00:52:49
emotionally yet. What do you mean by that? I haven't come to terms with what the
00:52:56
development of super intelligence could do to my children's future.
00:53:01
I'm okay. I'm 77. I'm going to be out of here soon. But
00:53:07
for my children and my my younger friends, my nephews and nieces and their
00:53:13
children, um I just don't like to think about what
00:53:19
could happen. Why? Cuz it could be awful.
00:53:29
In In what way? Well, if I ever decided to take over. I
00:53:35
mean, it would need people for a while to run the power stations until it designed better analog machines to run
00:53:42
the power stations. There's so many ways it could get rid of people, all of which
00:53:47
would of course be very nasty. Is that part of the reason you do what you do now? Yeah. I I mean, I think we
00:53:54
should be making a huge effort right now to try and figure out if we can develop it safely. Are you concerned about the
00:54:01
midterm impact potentially on your nephews and your your kids in terms of their jobs as well? Yeah, I'm concerned
00:54:07
about all that. Are there any particular industries that you think are most at risk? People talk about the creative industries a lot and sort of knowledge
00:54:14
work. They talk about lawyers and accountants and stuff like that. Yeah. So, that's why I mentioned plumbers. I think plumbers are less at risk. Okay,
00:54:21
I'm going to become a plumber. Someone like a legal assistant, a parallegal.
00:54:26
Um they're not going to be needed for very long. And is there a wealth inequality issue here that will will
00:54:32
arise from this? Yeah, I think in a society which shared out things fairly,
00:54:37
if you get a big increase in productivity, everybody should be better off.
00:54:43
But if you can replace lots of people by AIS,
00:54:48
then the people who get replaced will be worse off and the company that supplies the AIS
00:54:55
will be much better off and the company that uses the AIS. So
00:55:01
it's going to increase the gap between rich and poor. And we know that if you look at that gap between rich and poor,
00:55:08
that basically tells you how nice the society is. If you have a big gap, you get very nasty societies in which people
00:55:14
live in world communities and put other people in mass jails. It's not good to
00:55:21
increase the gap between rich and poor. The International Monetary Fund has expressed profound concerns that generative AI could cause massive labor
00:55:28
disruptions and rising inequality and has called for policies that prevent this from happening. I read that in the
00:55:34
business insider. So, have they given any of what the policies should look like? No. Yeah, that's the problem. I
00:55:40
mean, if AI can make everything much more efficient and get rid of people for most jobs or have a person assisted by I
00:55:47
doing many many people's work, it's not obvious what to do about it. It's
00:55:53
universal basic income, give everybody money. Yeah, I I I think that's a good start and it stops people
00:56:01
starving. But for a lot of people, their dignity is tied up with their job. I mean, who you think you are is tied up
00:56:08
with you doing this job, right? Yeah. And if we said, "We'll give you the same
00:56:13
money just to sit around," that would impact your dignity. You said something
00:56:18
earlier about it surpassing or being superior to human intelligence. A lot of people, I think, like to believe that AI
00:56:26
is is on a computer and it's something you can just turn off if you don't like it. Well, let me tell you why I think
00:56:31
it's superior. Okay. Um, it's digital. And because it's digital, you can have
00:56:38
you can simulate a neural network on one piece of hardware. Yeah. And you can simulate exactly the same neural network
00:56:44
on a different piece of hardware. So you can have clones of the same intelligence.
00:56:50
Now you could get this one to go off and look at one bit of the internet and this other one to look at a different bit of
00:56:56
the internet. And while they're looking at these different bits of the internet, they can be syncing with each other. So
00:57:02
they keep their weights the same, the connection strengths the same. Weights are connection strengths. Mhm. So this one might look at something on the
00:57:08
internet and say, "Oh, I'd like to increase this strength of this connection a bit." And it can convey
00:57:14
that information to this one. So it can increase the strength of that connection a bit based on this one's experience.
00:57:19
And when you say the strength of the connection, you're talking about learning. That's learning. Yes. Learning consists of saying instead of this one
00:57:26
giving 2.4 four votes for whether that one should turn on. We'll have this one give 2.5 votes for whether this one
00:57:31
should turn on. And that will be a little bit of learning. So these two different copies of the same neural net
00:57:39
are getting different experiences. They're looking at different data, but they're sharing what they've learned by
00:57:44
averaging their weights together. Mhm. And they can do that averaging at like a you can average a trillion weights. When
00:57:52
you and I transfer information, we're limited to the amount of information in a sentence. And the amount of
00:57:57
information in a sentence is maybe a 100 bits. It's very little information. We're lucky if we're transferring like
00:58:03
10 bits a second. These things are transferring trillions of bits a second. So, they're billions of times better
00:58:09
than us at sharing information. And that's because they're digital. And you can have two bits of hardware using
00:58:16
the connection strengths in exactly the same way. We're analog and you can't do that. Your brain's different from my
00:58:21
brain. And if I could see the connection strengths between all your neurons, it wouldn't do me any good because my
00:58:27
neurons work slightly differently and they're connected up slightly differently. Mhm. So when you die, all
00:58:33
your knowledge dies with you. When these things die, suppose you take these two digital intelligences that are clones of
00:58:39
each other and you destroy the hardware they run on. As long as you've stored the connection strength somewhere, you
00:58:46
can just build new hardware that executes the same instructions. So, it'll know how to use those connection
00:58:52
strengths and you've recreated that intelligence. So, they're immortal. We've actually solved the problem of
00:58:58
immortality, but it's only for digital things. So, it knows it will essentially
00:59:05
know everything that humans know but more because it will learn new things. It will learn new things. It would also
00:59:11
see all sorts of analogies that people probably never saw. So, for example, at the point when GPT4
00:59:18
couldn't look on the web, I asked it, "Why is a compost heap like an atom
00:59:24
bomb?" Off you go. I have no idea. Exactly. Excellent. Most that's exactly what most
00:59:30
people would say. It said, "Well, the time scales are very different and the energy scales are very different." But
00:59:37
then I went on to talk about how a compost he as it gets hotter generates heat faster and an atom bomb as it
00:59:44
produces more neutrons generates neutrons faster. And so they're both chain reactions but at very different
00:59:50
time in energy scales. And I believe GPT4 had seen that during its training.
00:59:56
It had understood the analogy between a compost heap and an atom bomb. And the reason I believe that is if you've only
01:00:02
got a trillion connections, remember you have 100 trillion. And you need to have thousands of times more knowledge than a
01:00:08
person, you need to compress information into those connections. And to compress
01:00:14
information, you need to see analogies between different things. In other words, it needs to see all the things
01:00:20
that are chain reactions and understand the basic idea of a chain reaction and code that code the ways in which they're
01:00:25
different. And that's just a more efficient way of coding things than coding each of them separately.
01:00:31
So it's seen many many analogies probably many analogies that people have never seen. That's why I also think that
01:00:38
people who say these things will never be creative. They're going to be much more creative than us because they're
01:00:43
going to see all sorts of analogies we never saw. And a lot of creativity is about seeing strange analogies.
01:00:49
People are somewhat romantic about the specialness of what it is to be human. And you hear lots of people saying it's
01:00:54
very very different. It's a it's a computer. We are, you know, we're conscious. We are creatives. We we have
01:00:59
these sort of innate unique abilities that the computers will never have. What do you say to those people? I'd argue a
01:01:06
bit with the innate. Um, so
01:01:11
the first thing I say is we have a long history of believing people were special. And we should have learned by
01:01:17
now. We thought we were at the center of the universe. We thought we were made in the image of God. white people thought
01:01:24
they were very special. We just tend to want to think we're special.
01:01:29
My belief is that more or less everyone has a completely wrong model of what the
01:01:35
mind is. Let's suppose I drink a lot or I drop some acid and not recommended and
01:01:41
I say to you I have the subjective experience of little pink elephants
01:01:47
floating in front of me. Mhm. Most people interpret that as there's some kind of
01:01:54
inner theater called the mind and only I can see what's in my mind and
01:02:01
in this inner theata there's little pink elephants floating around. So in other words, what's happened is my
01:02:08
perceptual systems gone wrong and I'm trying to indicate to you how it's gone
01:02:13
wrong and what it's trying to tell me. And the way I do that is by telling you what would have to be out there in the
01:02:19
real world for it to be telling the truth.
01:02:24
And so these little pink elephants, they're not in some inner theater. These little pink elephants are hypothetical
01:02:31
things in the real world. And that's my way of telling you how my perceptual systems telling me FIPS. So now let's do
01:02:38
that with a chatbot. Yeah. because I believe that current multimodal chatbots have subjective experiences and very few
01:02:46
people believe that. But I'll try and make you believe it. So suppose I have a multimodal chatbot. It's got a robot arm
01:02:53
so it can point and it's got a camera so it can see things and I put an object in front of it and I say point at the
01:03:00
object. It goes like this. No problem. Then I put a prism in front of its lens.
01:03:06
And so then I put an object in front of it and I say point at the object and it goes there.
01:03:11
And I say, "No, that's not where the object is. The object's actually straight in front of you, but I put a
01:03:17
prism in front of your lens." And the chatbot says, "Oh, I see. The prism bent the light rays." So, um, the object's
01:03:24
actually there, but I had the subjective experience that it was there. Now, if the chatbot says that, is using
01:03:31
the word subjective experience exactly the way people use them. It's an alternative view of what's going on.
01:03:37
They're hypothetical states of the world. which if they were true would mean my perceptual system wasn't lying.
01:03:43
And that's the best way I can tell you what my perceptual system is doing when it's lying to me. Now, we need to go
01:03:49
further to deal with sentience and consciousness and feelings and emotions, but I think in the end they're all going to be dealt with in a similar way.
01:03:56
There's no reason machines can't have them all because people say machines can't have feelings. And people are
01:04:02
curiously confident about that. I have no idea why. Suppose I make a battle robot and it's a little battle robot and
01:04:09
it sees a big battle robot that's much more powerful than it. It would be
01:04:14
really useful if it got scared. Now, when I get scared, um, various
01:04:22
physiological things happen that we don't need to go into, and those won't happen with the robot. But all the
01:04:27
cognitive things like I better get the hell out of here and I better sort of change my way of thinking so I focus and
01:04:35
focus and focus and don't get distracted. All of that will happen with robots, too. People will build in things
01:04:42
so that they when the circumstances such they should get the hell out of there, they get scared and run away. They'll
01:04:48
have emotions then. They won't have the physiological aspects, but they will have all the cognitive aspects. And I
01:04:55
think it would be odd to say they're just simulating emotions. No, they're really having those emotions. The little robot got scared and ran away. It's not
01:05:02
running away because of adrenaline. It's running away because of a sequence of sort of neurological in its neural net
01:05:09
processes happened which which have the equivalent effect to adrenaline. So do you do you and it's not just adrenaline,
01:05:15
right? There's a lot of cognitive stuff goes on when you get scared. Yeah. So, do you think that
01:05:21
there is conscious AI? And when I say conscious, I mean that represents the
01:05:26
same properties of consciousness that a human has. There's two issues here. There's a sort of empirical one and a
01:05:31
philosophical one. I don't think there's anything in principle that stops machines from being conscious.
01:05:38
I'll give you a little demonstration of that before we carry on. Suppose I take your brain and I take one brain cell in
01:05:44
your brain and I replace it by this a bit black mirror-l like. I replace it by
01:05:50
a little piece of nanotechnology that's just the same size that behaves in exactly the same way when it gets pings
01:05:56
from other neurons. It sends out pings just as the brain cell would have. So the other neurons don't know anything's
01:06:02
changed. Okay. I've just replaced one of your brain cells with this little piece of nanote technology. Would you still be
01:06:08
conscious? Yeah. Now you can see where this argument is going. Yeah. So if you replaced all of them as I replace them
01:06:16
all, at what point do you stop being conscious? Well, people think of consciousness as this like ethereal
01:06:22
thing that exists maybe beyond the brain cells. Yeah. Well, people have a lot of crazy ideas.
01:06:29
Um, people don't know what consciousness is and they often don't know what they mean by it. And then they fall back on
01:06:35
saying, well, I know it cuz I've got it and I can see that I've got it and they fall back on this theata model of the
01:06:41
mind which I think is nonsense. What do you think of consciousness as if you had to try and define it? Is it because I
01:06:47
think of it as just like the awareness of myself? I don't know. I think it's a term we'll stop using. Suppose you want
01:06:53
to understand how a car works. Well, you know, some cars have a lot of oomph and other cars have a lot less oomph. Like
01:07:00
an Aston Martin's got lots of oomph. And a little Toyota Corolla doesn't have much oomph. But oomph isn't a very good
01:07:08
concept for understanding cars. Um, if you want to understand cars, you need to understand about electric engines or
01:07:14
petrol engines and how they work. And it gives rise to oomph, but oomph isn't a
01:07:19
very useful explanatory concept. It's a kind of essence of a car. It's the essence of an Aston Martin, but it
01:07:25
doesn't explain much. I think consciousness is like that. And I think we'll stop using that term, but I don't
01:07:31
think there's anything any reason why a machine shouldn't have it. If your view
01:07:36
of consciousness is that it intrinsically involves self-awareness, then the machine's got to have
01:07:41
self-awareness. He's got to have cognition about its own cognition and stuff. But
01:07:47
I'm a materialist through and through. And I don't think there's any reason why
01:07:52
a machine shouldn't have consciousness. Do you think they do then have the same consciousness that we think of ourselves
01:07:59
as being uniquely uh given as a gift when we're born? I'm ambivalent about
01:08:05
that at present. So I don't think there's this hard line. I
01:08:10
think as soon as you have a machine that has some self-awareness, it's got some consciousness. Um, I think
01:08:18
it's an emergent property of a complex system. It's not a sort of essence that's
01:08:24
throughout the universe. It's you make this really complicated system that's complicated enough to have a model of
01:08:29
itself and it does perception. And I think then
01:08:34
you're beginning to get a conscious machines. So I don't think there's any sharp distinction between what we've got
01:08:39
now and conscious machines. I don't think it's going to one day we're going to wake up and say, "Hey, if you put
01:08:45
this special chemical in, it becomes conscious." It's not going to be like that. I think we all wonder if these
01:08:50
computers are like thinking like we are on their own when we're not there. And if they're experiencing emotions, if
01:08:56
they're contending with I think we probably, you know, we think about things like love and things that are feel unique to biological species. Um,
01:09:04
are they sat there thinking? Are they do they have concerns? I think they really are thinking and I think as soon as you
01:09:11
make AI agents they will have concerns. If you wanted to make an effective AI agent suppose you let's take a call
01:09:18
center. In a call center you have people at present they have all sorts of
01:09:23
emotions and feelings which are kind of useful. So suppose I call up the call
01:09:28
center and I'm actually lonely and I don't actually want to know the answer
01:09:34
to why my computer isn't working. I just want somebody to talk to. After a while,
01:09:40
the person in the call center will either get bored or get annoyed with me and will terminate it.
01:09:47
Well, you replace them by an AI agent. The AI agent needs to have the same kind of responses. If someone's just called
01:09:54
up because they just want to talk to the AI agent and we're happy to talk for the whole day to the AI agent, that's not
01:10:00
good for business. And you want an AI agent that either gets bored or gets irritated and says, "I'm sorry, but I
01:10:06
don't have time for this." And once it does that, I think it's got emotions.
01:10:12
Now, like I say, emotions have two aspects to them. There's the cognitive aspect and the behavioral aspect, and
01:10:18
then there's a physiological aspect, and those go together with us. And if the AI
01:10:25
agent gets embarrassed, it won't go red. Yeah. Um, so there's no physiological skin won't start sweating. Yeah, but it
01:10:31
might have all the same behavior. And in that case, I'd say yeah, it's having emotion. It's got an emotion. So, it's going to have the same sort of cognitive
01:10:37
thought and then it's going to act upon that cognitive in the same way, but without the physiological responses. And
01:10:43
does that matter that it doesn't go red in the face? And it's just a different I mean, that's a response to the It makes it somewhat different from us. Yeah. For
01:10:51
some things, the physiological aspects are very important like love. They're a long way from having love the same way
01:10:57
we do. But I don't see why they shouldn't have emotions. So I think
01:11:02
what's happened is people have a model of how the mind works and what feelings
01:11:07
are and what emotions are and their model is just wrong. What um what
01:11:13
brought you to Google? You you worked at Google for about a decade, right? Yeah. What brought you there? I have a son who
01:11:21
has learning difficulties and in order to be sure he would never be out on the street, I needed to get
01:11:30
several million dollars and I wasn't going to get that as an academic. I
01:11:35
tried. So, I taught a Corsera course in the hope that I'd make lots of money that way, but there was no money in
01:11:40
that. Mhm. So I figured out well the only way to get millions of dollars is
01:11:46
to sell myself to a big company. And so when I was 65,
01:11:54
fortunately for me, I had two brilliant students who produced something called Alexet, which was neural net that was
01:12:01
very good at recognizing objects in images. And so Ilia and Alex and I set up a little
01:12:10
company and auctioned it. And we actually set up an auction where we had a number of big companies bidding for
01:12:16
us. And that company was called AlexNet. No,
01:12:21
the the the network that recognized objects was called Alexet. The company was called DNN Research, deep neural
01:12:27
network research. And it was doing things like this. I'll put this graph up on the screen. That's that's Alexet.
01:12:33
This picture shows eight images and Alex Net's ability, which is your company's
01:12:39
ability to spot what was in those images. Yeah. So, it could tell the difference between various kinds of
01:12:44
mushroom. And about 12% of imageet is dogs. And to be good at imageet, you
01:12:51
have to tell the difference between very similar kinds of dog. And it would got to be very good at that. And your your
01:12:57
company Alexet won several awards I believe for its ability to out outperform its competitors. And so
01:13:04
Google ultimately ended up acquiring your technology. Google acquired that technology and some other technology.
01:13:12
And you went to work at Google at age what 66. I went at age 65 to work at
01:13:17
Google. 65. And you left at age 76? 75. 75. Okay. I worked there for more or
01:13:23
less exactly 10 years. And what were you doing there? Okay, they were very nice to me. They said they said pretty much
01:13:29
you can do what you like. I worked on something called distillation that did really work well
01:13:35
and that's now used all the time in AI in AI and distillation is a way of
01:13:40
taking what a big model knows a big neural net knows and getting that knowledge into a small neural net. Then
01:13:46
at the end I got very interested in analog computation and whether it would be possible to get these big language
01:13:52
models running in analog hardware. So they used much less energy. And it was
01:13:58
when I was doing that work that I began to really realize how much better digital is for sharing information.
01:14:05
Was there a Eureka moment? There was a Eureka month or two. Um and
01:14:11
it was a sort of coupling of chat beauty coming out although Google had very similar things a year earlier and I'd
01:14:19
seen those and that had a big effect effect on me. The closest I had to a Eureka moment was when a Google system
01:14:26
called Palm was able to say why a joke was funny. And I'd always thought of
01:14:31
that as a kind of landmark. If it can say why a joke's funny, it really does understand and it could say why a joke
01:14:38
was funny. And that coupled with realizing why
01:14:43
digital is so much better than analog for sharing information suddenly made me very interested in AI
01:14:50
safety and that these things were going to get a lot smarter than us. Why did
01:14:56
you leave Google? The main reason I left Google was cuz I was 75 and I wanted to
01:15:01
retire. I've done a very bad job of that. The precise timing of when I left
01:15:07
Google was so that I could talk freely at a conference at MIT, but I left because I was I'm old and I was finding
01:15:15
it harder to program. I was making many more mistakes when I programmed, which is very annoying. You wanted to talk
01:15:20
freely at a conference at MIT. Yes. At MIT, organized by MIT Tech Review. What did you want to talk about freely? AI
01:15:27
safety. And you couldn't do that while you were at Google. Well, I could have done it while I was at Google. And
01:15:32
Google encouraged me to stay and work on AI safety and said I could do whatever I liked on AI safety. You kind of sense to
01:15:38
yourself if you work for a big company. You don't feel right saying things that will damage the big company. Even if you
01:15:45
could get away with it, it just feels wrong to me. I didn't leave because I was cross with anything Google was
01:15:51
doing. I think Google actually behaved very responsibly. When they had these big chat bots, they didn't release them
01:15:57
possibly because they were worried about their reputation. they had a very good reputation and they didn't want to
01:16:02
damage it. So open AI didn't have a reputation and so they could afford to take the gamble. I mean there's also a
01:16:09
big conversation happening around how it will cannibalize their core business in search. There is now. Yes. Yeah. Yeah.
01:16:16
And it's the old innovators dilemas to some degree I guess that contending with bad skin. I've had it and I'm sure many
01:16:23
of you listening have had it too or maybe you have it right now. I know how
01:16:28
draining it can be, especially if you're in a job where you're presenting often like I am. So, let me tell you about something that's helped both my partner
01:16:35
and me and my sister, which is red light therapy. I only got into this a couple of years ago, but I wish I'd known a
01:16:41
little bit sooner. I've been using our show sponsors Boncharg's infrared sauna blanket for a while now, but I just got
01:16:47
hold of their red light therapy mask as well. Red light has been proven to have so many benefits for the body. Like any
01:16:53
area of your skin that's exposed will see a reduction in scarring, wrinkles, and even blemishes. It also helps with
01:16:59
complexion. It boosts collagen, and it does that by targeting the upper layers of your skin. And Boncharge ships
01:17:05
worldwide with easy returns and a year-long warranty on all of their products. So, if you'd like to try it
01:17:10
yourself, head over to bondcharge.com/diary and use code diary for 25% off any
01:17:16
product sitewide. Just make sure you order through this link. bondcharge.com/diary
01:17:22
with code diary. Make sure you keep what I'm about to say to yourself. I'm inviting 10,000 of you to come even
01:17:29
deeper into the diary of a CEO. Welcome to my inner circle. This is a brand new
01:17:34
private community that I'm launching to the world. We have so many incredible things that happen that you are never
01:17:39
shown. We have the briefs that are on my iPad when I'm recording the conversation. We have clips we've never
01:17:44
released. We have behindthe-scenes conversations with the guests. and also the episodes that we've never ever
01:17:50
released and so much more. In the circle, you'll have direct access to me.
01:17:55
You can tell us what you want this show to be, who you want us to interview, and the types of conversations you would love us to have. But remember, for now,
01:18:02
we're only inviting the first 10,000 people that join before it closes. So, if you want to join our private closed
01:18:08
community, head to the link in the description below or go to daccircle.com.
01:18:14
I will speak to you there. I'm continually shocked by the types of individuals that listen to this conversation um because they come up to
01:18:20
me sometimes. So I hear from politicians, I hear from some real people, I hear from entrepreneurs all over the world, whether they are the
01:18:26
entrepreneurs building some of the biggest companies in the world or their, you know, early stage startups. For
01:18:31
those people that are listening to this conversation now that are in positions of power and influence,
01:18:38
world leaders, let's say, what's your message to them? I'd say what you need is highly
01:18:44
regulated capitalism. That's what seems to work best. And what would you say to the average person
01:18:49
not doesn't work in the industry, somewhat concerned about the future, doesn't know if they're helpless or not.
01:18:56
What should they be doing in their own lives? My feeling is there's not much they can
01:19:01
do. This isn't isn't going to be decided by just as climate change isn't going to
01:19:06
be decided by people separating out the plastic bags from the um compostables.
01:19:12
That's not going to have much effect. It's going to be decided by whether the lobbyists for the big energy companies
01:19:18
can be kept under control. I don't think there's much people can do to except for
01:19:26
try and pressure their governments to force the big companies to work on AI
01:19:32
safety that they can do. You've lived a a fascinating fascinating
01:19:39
winding life. I think one of the things most people don't know about you is that your family has a
01:19:45
big history of being involved in tremendous things. You have a family tree which is one of the most impressive
01:19:51
that I've ever seen or read about. Your great greatgrandfather George Bull
01:19:57
founded the Boolean algebra logic which is one of the foundational principles of modern computer science. You have uh
01:20:03
your great great grandmother Mary Everest Bull who was a mathematician and educator who made huge leaps forward in
01:20:11
mathematics from what I was able to ascertain. Um I mean I can the list goes on and on and on. I mean, your great
01:20:16
great uncle George Everest is what Mount Everest is named after.
01:20:22
Is that is that correct? I think he's my great great great uncle. His his niece
01:20:30
married George Bull. So Mary Mary Bull was Mary Everest Bull.
01:20:36
Um she was the niece of Everest. And your first cousin once removed, Joan Hinton, was involved in the a nuclear
01:20:43
physicist who worked on the Manhattan project, which is the World War II development of the first nuclear bomb.
01:20:48
Yeah. She was one of the two female physicists at Los Alamos. And then after they dropped the bomb,
01:20:56
she moved to China. Why? She was very cross with them dropping the bomb. And
01:21:02
her family had a lot of links with China. Her mother was friends with
01:21:08
Chairman Mo. Quite weird. When you look back at your life,
01:21:14
Jeffrey, we have the hindsight you have now and the ret retrospective clarity,
01:21:22
what might you have done differently if you were advising me? I guess I have two pieces of advice. One
01:21:30
is if you have an intuition that people are doing things wrong and there's a
01:21:35
better way to do things, don't give up on that intuition just because people say it's silly. Don't give up on the
01:21:42
intuition until you figured out why it's wrong. Figured out for yourself why that intuition isn't correct. And usually
01:21:50
it's wrong if it disagrees with everybody else and you'll eventually figure out why it's wrong.
01:21:56
But just occasionally you'll have an intuition that's actually right and everybody else is wrong. And I lucked
01:22:02
out that way. Early on I thought neural nets are definitely the way to go to
01:22:08
make AI and almost everybody said that was crazy and I stuck with it because I
01:22:13
couldn't. It seemed to me it was obviously right. Now the idea that you should stick with
01:22:19
your intuitions isn't going to work if you have bad intuitions. But if you have bad intuitions, you're never going to do
01:22:26
anything anyway, so you might as well stick with them. And in your own career journey, is there
01:22:32
anything you look back on and say, "With the hindsight I have now, I should have taken a different approach at that juncture."
01:22:38
I wish I'd spent more time with my wife um
01:22:47
and with my children when they were little. I was kind of obsessed with work.
01:22:55
Your wife passed away. Yeah. From ovarian cancer. No. Or that was another
01:23:00
wife. Okay. Um I had two wives to have cancer. Oh, really? Sorry. The first one died of ovarian cancer and the second
01:23:07
one died of pancreatic cancer. And you wish you'd spent more time with her? With the second wife? Yeah. Who was a
01:23:12
wonderful person? Why did you say that in your 70s? What is it that you've you figured out that I
01:23:19
might not know yet? Oh, just cuz she's gone and I can't spend more time with her now. Mhm.
01:23:26
But you didn't know that at the time. At the time, you think
01:23:33
I mean it was likely I would die before her just cuz she was a woman and I was a man. Um I didn't
01:23:40
I just didn't spend enough time when I could. I I think I I inquire there because I
01:23:46
think there's many of us that are so consumed with what we're doing professionally that we kind of assume immortality with our partners because
01:23:52
they've always been there. So we Yeah. I mean she was very supportive of me spending a lot of time working but and
01:23:59
why did you say your children as well? What's the what's the Well, I didn't spend enough time with them when they were little
01:24:05
and you regret that now. Yeah.
01:24:12
If you um if you had a closing message for for my for my listeners about AI and AI safety, what would that be? Jeffrey,
01:24:20
there's still a chance that we can figure out how to develop AI that won't want to take over from us. And because
01:24:27
there's a chance, we should put enormous resources into trying to figure that out because if we don't, it's going to take
01:24:33
over. And are you hopeful? I just don't know. I'm agnostic.
01:24:40
you must get get bed get in bed at night and when you're thinking to yourself about probabilities of outcomes there
01:24:46
must be a bias in one direction because there certainly is for me I imagine everyone listening now has a
01:24:53
internal prediction that they might not say out loud but of how they think it's going to play out I really don't know I
01:25:00
genuinely don't know I think it's incredibly uncertain when I'm feeling slightly depressed I think people are
01:25:08
toast is going to take over while I'm feeling cheerful. I think we'll figure out a way. Maybe one of the facets of
01:25:14
being a human um is because we've always been here, like we were saying about our loved ones and our relationships, we
01:25:21
assume casually that we will always be here and we'll always figure everything out. But there's a beginning and an end
01:25:27
to everything as we saw from the dinosaurs. I mean, yeah. And
01:25:32
we have to face the possibility that unless we do something soon,
01:25:39
we're near the end. We have a closing tradition on this podcast where the last guest leaves a
01:25:45
question in their diary. And the question that they've left for you is
01:25:54
with everything that you see ahead of us, what is the biggest threat you see to
01:26:00
human happiness? I think the joblessness is a fairly
01:26:07
urgent short-term threat to human happiness. I think if you make lots and lots of people unemployed, even if they
01:26:14
get universal basic income, um they're not going to be happy
01:26:19
because they need purpose. Because they need purpose. Yes. And struggle. They need to feel they're contributing
01:26:25
something. They're useful. And do you think that outcome that there's going to be huge job displacement is more
01:26:31
probable than not? Yes, I do. And what sort of that one I think is definitely
01:26:36
more probable than not. If I worked in a call center, I'd be terrified. And what's the time frame for that in
01:26:42
terms of mass jobs? I think it's beginning to happen already. I read an article in the Atlantic recently that
01:26:49
said it's already getting hard for university graduates to get jobs. And
01:26:54
part of that may be that people are already using AI for the jobs they would have got. I spoke to the CEO of a major
01:27:02
company that everyone will know of, lots of people use, and he said to me in DMs
01:27:07
that they used to have seven just over 7,000 employees. He said uh by last year they were down to I think 5,000. He said
01:27:13
right now they have 3,600. And he said by the end of summer because of AI agents they'll be down to 3,000. So
01:27:19
you've got So it's happening already. Yes. He's halfed his workforce because AI agents can now handle 80% of the
01:27:25
customer service inquiries and other things. So it's it's happening already.
01:27:30
Yeah. So urgent action is needed. Yep. I don't know what that urgent action is.
01:27:36
That's a tricky one because that depends very much on the political system and political systems are all going in the
01:27:42
wrong direction at present. I mean what do we need to do? Save up money? Like do we save money? Do we move to another
01:27:47
part of the world? I don't know. What would you tell your kids to do? They
01:27:53
said, "Dad, like there's going to be loads of job displacement." Because I worked for Google for 10 years. is they
01:27:58
have enough money. Okay. Okay. [ __ ] So, they're not typical. What if they didn't have money? Trained to be a plumber.
01:28:06
Really? Yeah. Jeffrey, thank you so much. You're the
01:28:12
first Nobel Prize winner that I've ever had a conversation with, I think, in my
01:28:18
life. So, that's a tremendous honor. And you you you received that award for a lifetime of exceptional work and pushing
01:28:23
the world forward in so many profound ways that will lead to great and that have led to great advancements and
01:28:29
things that matter so much to us. And now you've turned this season in your life to shining a light on some of your
01:28:34
own work, but also on the the the broader risks of AI and how um and how
01:28:40
it might impact us adversely. And there's very few people that have worked inside the machine of a Google or a big
01:28:47
tech company that have contributed to the field of AI that are now at the very forefront of warning us against the very
01:28:53
thing that they worked upon. There are actually surprising number of us now.
01:28:58
They're not as uh as public and they're actually quite hard to get to have these kinds of conversations because many of
01:29:04
them are still in that industry. So, you know, someone who tries to contact these people often and ask invites them to
01:29:10
have conversations, they often are a little bit hesitant to speak openly. They speak privately, but they're less
01:29:15
willing to openly because maybe maybe they still have something at some sort of incentives at play. I have an
01:29:21
advantage over them, which is I'm older, so I'm unemployed, so I can say what I Well, there you go. So, thank you for
01:29:27
doing what you do. It's a real honor and please do continue to do it. Thank you. Thank you so much.
01:29:32
People think I'm joking when I say that, but I'm not. The plumbing fish. Yeah. Yeah.
01:29:41
And plumbers are pretty well paid. [Music]
01:30:03
[Music]
