September 8, 2025 • 33 mins
Will AI end up building us into stronger, more talented humans? What might this have to do with linguistics, the movie Arrival, self-driving cars, debate, video games, elections, chess, and the ancient game of Go? Are we going to be taken over, or instead exposed to ideas and concepts that stretch the boundaries of our thinking? Join this week to see how AI might just up the human game.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:05):
Will AI make humans better? And what does this have
to do with linguistics or the movie Arrival or self
driving cars or debate and video games and elections and
chess and the ancient game of go. Welcome to Intercosmos
(00:26):
with me David Eagleman. I'm a neuroscientist and an author
at Stanford and in these episodes we seek to understand
why and how our lives look the way they do.
In today's episode is about whether AI will make us
better humans. I often find myself totally flabbergasted by the
(00:58):
change that I've seen just in my lifetime. When I
was a really little kid, personal computers didn't exist, and
then we passed through a door and suddenly they did.
And I saved up my money and I got a
common or VIC twenty computer, which was something my parents
had never seen the likes of. And I felt like
I was living the largest change in human history, because
(01:20):
for the first time, everybody could have a machine that
would do all kinds of things. But that turned out
not to even be the biggest change, because the next
stop was even bigger, and that was the idea that
some people had to build a system where we could
keep information on computers and make computers talk to each
(01:42):
other such that if the Soviets bombed America, the important
information wouldn't just be stored on one computer, and messages
could follow different routes on the network, and that way
you had a very robust system for keeping information. That
was of course Arpanet, which became the Internet, and not
(02:02):
long after, the idea was introduced of a way that
everyone could use this giant network, not just with text,
but with graphics, and.
Speaker 2 (02:10):
That was the birth of the Worldwide Web.
Speaker 1 (02:13):
And as soon as that technology existed, then people started
figuring out what to do with it, and one young
man started to sell books over the Internet.
Speaker 2 (02:22):
And that became Amazon.
Speaker 1 (02:24):
And two graduate students at Stanford asked the question of
how the heck we were going to be able to
find information on this sprawling network, and they created a
way of measuring all the connections between web pages and
that gave the ability to search for it, and.
Speaker 2 (02:40):
That garage project grew into Google.
Speaker 1 (02:44):
And around the same time, a Harvard kid was thinking
about a better way to allow his classmates to get
to know each other, which was traditionally done at the
beginning of the year by printing a booklet with everybody's
picture and name, and that little booklet was called called
a Facebook, and he thought of digitizing that, and on
(03:04):
and on, people increasingly figured out how to take this
new technology and make things that would live on top
of it. And I felt very lucky to have experienced
two truly world changing inventions in my lifetime, and I
knew that had launched us into a world that was
so different from what my great grandparents could have possibly imagined.
(03:30):
But now here we are in the middle of a
third revolution. It's related to the first two computers and
the Internet, but it makes them look like warm up acts.
And that's the fact that we have created a new
intelligent species that we're going to be sharing the planet
with from now on. This is not to say that
(03:51):
AI has exactly the same type of intelligence that human
brains have, but obviously it has absorbed the higher knowledge
sphere of humankind and it can spit that back to
us with all sorts of remixes. Now, the question is,
are we in trouble because of this new invention? Have
we taken things a step too far? Well, let me
(04:15):
give you an example that's on people's minds. A few
months ago, a Swiss research team carried out a secret
experiment in online persuasion. They went to a forum on
the website Reddit, and in this forum, users post opinions
and invite other people to challenge them. And so these
(04:36):
researchers quietly unleashed a set of AI accounts to pretend
that they were people and to try to change other
people's minds. So their large language models LMS, they participated
in the debating just like any other human. These bots
wrote arguments, they engaged with users, They debated with the
(05:00):
hope of changing minds. Now, the researchers measured success by
tallying whenever an original poster publicly admitted that their mind
had been changed. So how did the bots do well?
They achieved up to an eighteen percent success rate in
changing people's minds. Now, the critical piece you need to
(05:22):
know is that the average human success rate is about
three percent. So the bots had absolutely crushed their human
competition in an environment teeming with intellectuals. The bots not
only survived, but they thrived. In fact, one of the
AI accounts climbed into the ninety ninth percentile of all
(05:45):
users on this subreddit. It racked up ten thousand karma
points along the way. Now, one of the most surprising
aspects of the experiment was how effectively a small team
of researchers, operating with modest academic resources, how they were
able to outperform essentially every human debater on the platform.
Speaker 2 (06:07):
So let that sink in for a second.
Speaker 1 (06:09):
You have a handful of graduate students armed with an
LM and a sneaky deployment strategy, and they quietly demonstrated
what it might look like if influence operations were scaled
by AI. Now, when this data was released, one of
the scariest parts to people was that nobody noticed that
(06:32):
these debaters were actually AI and not real humans. This
reddit channel prides itself as one of the most critically
minded communities on the platform. If any place could sniff
out an imposter, it should have been here. But for
the entire four months of the experiment, the bots played
along undetected by the humans. So that's worrisome. And there's
(06:57):
another issue too, which is the role of personal data.
When the bots were given just a basic profile of
the user, their aged, their location, and their political leaning,
the success rate bumped up by one percentage point. Now,
you might say wait a minute, who cares. One percentage
point is very small, but in political terms, one percent
(07:17):
could be enough to swing a lot of national elections.
The difference between nudging public sentiment one way or another
can come down to.
Speaker 2 (07:27):
Very tiny tweaks.
Speaker 1 (07:28):
So the lesson that surface here is that even minimal
personalization can significantly sharpen the edge of an AI's persuasive power.
So when this story surfaced recently about these debaters not
being real human beings, the reactions were very grim because
everyone realized that this small academic experiment illustrated what stealth
(07:55):
influence operations could look like in the near future. If
a handful of real searchers could do this undetected, what
happens when you have state actors or corporations or political
campaigns with real resources doing the same thing at a
scale thousands of times larger. Also, what's worse is that
(08:16):
the failure of everyone to detect that these were bots
that raises serious questions about how we can protect public
discourse as we enter this new future. If smart reditor
debaters couldn't spot the difference between a human and a bot,
what hope is there for broader audiences. So even though
(08:37):
the major AI companies have all pledged to avoid building
models with dangerous capabilities like manipulating public opinion on mass.
This Reddit experiment suggests the thresholds may already be easier
to cross than any of us had anticipated. What this
(08:57):
all means is that with major elections in future years,
were genuinely gonna have to worry about this. The halcyon
days are gone when we could assume that the replies
to our online messages came from a fellow human, and
the risks are real and the potential for abuse is obvious.
(09:19):
But for today's episode, we're gonna look at all this
from a very different angle. I think it might be
worth asking a different question, which is why did the
bots succeed? After all, they didn't hack people's brains with
neural interfaces. They didn't spread fear or disinformation. They didn't
(09:39):
manipulate emotions or overwhelm users with noise. They simply made
better arguments. They didn't insult people while arguing with them.
They didn't do little jabs and digs. They just made
good arguments empathetically. The bots present their points calmly and
(10:02):
rationally and persuasively, and when users changed their minds.
Speaker 2 (10:08):
It wasn't because they had been tricked.
Speaker 1 (10:10):
It was because they recognized that another perspective was worth considering.
(10:32):
Humans often change their minds when faced with sound reasoning
that can be backed up, and mind changing is a
great thing. It means you're willing to reconsider some closely
held opinion when the facts or logic warranted. It's a
mark of intellectual strength, not a sign that you've been tricked.
(10:54):
In that light, the success of AI debaters isn't necessarily
a story about manipulation. It could also be a story
about raising the bar of debating. Now, as I followed
the outcome of this AI debater study, it struck me
that there may be a helpful precedent for thinking about
(11:15):
this moment, for thinking about how AI could.
Speaker 2 (11:18):
Actually improve us.
Speaker 1 (11:20):
So think about the events that unfolded in the hyper
competitive world of chess. If you can remember back to
nineteen ninety seven, IBM had built an AI system called
Deep Blue, and it defeated the world champion, Gary Kasprov.
This was seen as a seismic moment. The game of
(11:41):
chess had been quote unquote solved, and everyone worried that
human mastery was obsolete and then about two decades later,
Alpha Go beat the world's top Go player, and the
choir of concerned voices grew louder. But then something unexpected happened.
(12:02):
In May of twenty seventeen, the world's number one Go player,
Could G, faced off against his toughest opponent. G was
the reigning champion in Go, and you know this is
the game where two players use smooth black rocks or
white rocks to surround more territory than their opponent. In
G's case, he was playing against AI. He was playing
(12:25):
against Alpha Go, which had been trained on many millions
of games, and it had deeply absorbed the statistics of
possible plays.
Speaker 2 (12:35):
So G lost the first game.
Speaker 1 (12:38):
Alpha Go had pulled moves that none of G's human
opponents had ever thought of, and then G lost the
second game. The fact is he didn't stand a chance.
The AI had won over a human in a game
that was way more complex than chess, and subsequent versions
of this AI will without doubt continue to win ever more.
(13:01):
But that's not the interesting part of the story. The
interesting part is what happened next. G got over his
embarrassment and he became mesmerized by what the heck had
just transpired. He studied the games that he lost. Now,
before he had played Alpha Go, G had won most
(13:23):
of the games against his human opponents, But after he
played Alpha Go, he was able to beat his human
opponents even more easily. In other words, after his species
shaming defeat in twenty seventeen, G went on to play
twelve straight matches against fellow humans, and he won.
Speaker 2 (13:42):
Them all in a row. Now, what had happened.
Speaker 1 (13:46):
G had been exposed to new kinds of moves and
strategies that AlphaGo was pulling off that lay outside the
traditional ideas. All these moves were legal and pop pole,
but they were different from what had been played over
the previous twenty five hundred years. For go eficionados, this
(14:08):
included novelties like playing a stone directly diagonal to your
opponent's loan stone, or commonly playing six space extensions while
humans tend to prefer five space. So G reported that
playing against the AI was like opening a door to
another world. Some people worry that AI might make games
(14:31):
of Chess and Go irrelevant, but amazingly, that is not
what's happened. When AI Trump's Chess champions and Go champions
It does so with moves that seem inhumanly creative, but
all the moves are allowed by the rules. Humans simply
never thought to go there before. And the key is
(14:53):
that once the moves are seen by humans, then they're
easily incorporated into our models. GE's experience with Alpha Go
illuminated new nooks and crannies in his landscape. It exposed
pathways that have never been lit up before. So AI
(15:14):
immediately became a tool for steep improvement. And nowadays all
chests and Go players above a certain level they all
train with AI. They study the surprising and sometimes counterintuitive
and alien strategies of the artificial mind, and just like Koje,
today's grand masters play a deeper and more creative game
(15:38):
than ever before. In other words, many commentators are worried
that AI is going to leave humans far behind, and
in some respects that's true.
Speaker 2 (15:48):
But as computation improves, so will we.
Speaker 1 (15:53):
AI will illuminate dark parts of our maps, allowing us
to see new roads we didn't even suspect. The key
point I want to make here is that instead of
dampening human excellence, AI sparked a renaissance. In these games,
human minds are elevated by learning from our artificial cousins.
(16:14):
And if it were just chests and go, that's one thing.
But I think we can detect this pattern happening all
over the place. For example, the same thing has happened
in poker. Professional poker players master the art of bluffing.
They project strength when they're weak or weakness when they're strong.
This has always been seen like a deep test of
(16:35):
human psychology. The players read faces, they watch for micro expressions,
they guess intentions, So in this way, poker is a
really human game. But then Carnegie Mellon cooked up an
AI called Librotus. Obviously, it doesn't have a face, it
doesn't sweat, it doesn't blink, so it wasn't really like
playing another human. So Lebronis started playing poker, and it
(16:59):
did things that baffled the human players. The AI began
making betting choices that looked bizarre. Sometimes it over bet
the pot by huge margins, a move that the human
players considered reckless. Other times it made really tiny bets
in situations where no human would bother and the human
(17:19):
players dismissed this all as network nonsense. But as the
games went on, they realized these inhuman strategies were working.
The AI was winning, and it was doing so by
reinventing the language of poker. And now, just like with
chess and go, professional poker players study these strategies. You
(17:43):
have human players adopting patterns of play that just didn't
exist before the machines taught them. AI made humans better
poker players. And let me give you another example. There's
a strategy video game called StarCraft two where players build
armies and manage resources and try to outwit their opponent.
(18:04):
So the company DeepMind set some AI agents on it,
and those agents showed humans an entirely different way of
going about running the strategy. At first, people thought the
AI play seemed unfair and robotic, but then human players
began to adapt. They started rethinking their own strategies. They
(18:26):
started redesigning their build orders. They discovered things like sometimes
sacrificing entire groups of units early in the game could
produce long term advantages. These were not moves that humans
had come up with, or possibly ever would have come
up with, but once they were revealed, they became part
(18:47):
of the human playbook. Or take a different video game
called Dota two. You've got professional players. But then open
AI built a system that used strategies that looked to
human professionals like they were clumsy. The AI would push
aggressively when humans would retreat or take risks that seemed absurd,
(19:08):
but more often than not, the machines won, and just
like with Chests and Go and StarCraft, the human professionals
were forced to ask themselves had they been playing with
blinders on all along? After all, the machines were uncovering
landscapes that we just never knew were there. In other words,
(19:29):
whenever AI discovers new pathways inside the universe of a game,
it illuminates something for us about the fence lines of
our own imagination. Humans had decades to explore video games,
and centuries to explore poker, and millennia to explore the
games of Chess and Go. But within weeks AI uncovered
(19:53):
strategies that had never even struck us. And this is
the point of today's episode. AI can expand the possibility
space for us. What we've seen in the past few
years is the limits of our imagination. Even in domains
we thought we had mastered, our internal models might be
(20:14):
a lot more narrow than we had ever realized. But
by playing with machines, we're learning how they think, and
more importantly, how we might think differently. So beyond chests
or go or video games, the bigger game is how
we can expand our own internal models. A lot of
people cast this as man versus machine, but I think
(20:36):
a more productive lens is seeing it as man learning
from machine. The prize is a new way of seeing
the game, and maybe, by extension, a new way of
seeing the world.
Speaker 2 (20:51):
So let's come back to that.
Speaker 1 (20:53):
Amazing Swiss experiment where they made debate bots on Reddit
that performed way better than humans. I suggest it's possible
and maybe even likely, that a similar dynamic is going
to unfold in the world of persuasive argumentation that happened
in chess and go in video games. If AI agents
(21:14):
can model the best forms of debate, clear and structured
and empathetic and rational, then we humans can learn something
from our artificial cousins. We can try out new moves,
We can sharpen our skills on digital grinding stones. Imagine
(21:36):
a future where students practice crafting arguments by debating highly
skilled AI tutors. Imagine online discussions becoming more useful because
users have gotten used to high quality exchanges. Imagine politicians
and journalists and everyday citizens pushed to improve their things
(22:00):
and better articulate their positions. So, rather than dumbing down conversation,
the rise of high performing debate bots could nudge public
discourse toward a new level of reasoned discussion. If that
turns out to be the case, then we may come
to see AI not as an enemy but as a
(22:22):
sparring partner, just like in Chest and Go, and that
has very different implications. Now, some people will take just
(22:45):
the opposite position. If AI is so much better at
debating than we are, won't that cause us to lose
the skill entirely because there doesn't seem to be a
point in being a good debater if the computer can
do it better. But I think this is not an
issue because even when we extrapolate this era of AI,
(23:07):
people will still be talking with each other most of
the time. You'll still be with your family over dinner,
or with your friends over a coffee, or arguing with
your neighbor about where the fence goes, or debating with
a stranger at a town hall or whatever. It's not
like we are plugging into the matrix through a port
in the back of our neck, and we're only going
(23:27):
to be communicating with machines. We are intensely social creatures,
and the success of our species has been in part
because of our massive sociability. So I think we will
debate other humans all the time, and AI is just
going to train us to be a little bit better
(23:49):
at it. Now, I don't want to minimize the risks
we're facing. We need audit tools and authentication systems that
can verify whether content was written by humans or AIA.
We need technical solutions like water marking or cryptographic content authentication,
and providence tracking. This is all essential, but beyond these
(24:11):
defensive measures, we should also recognize the opportunity because like
the chests and go engines that reshaped how champion players think,
debate bots will reshape how we reason and argue and
understand one another. If we take this on correctly, AI
(24:32):
might just up our game. It's still early days, but
my impression so far is that AI is already playing
this role in many areas, including for example, the arts.
At least in some ways, it's amplifying human creativity. It's
like a jazz partner who plays a riff you weren't expecting,
(24:52):
or a painting mentor who introduces a color that you
never thought to use. AI forces us out of our
group in the arts. It shows us that the boundaries
of our imagination can be stretched by encountering minds, even
synthetic minds that think differently. These AI systems presumably don't
(25:13):
have esthetic tastes or emotional longing in the way that
we do, but they're awesome at doing remixes and trying
strange new things out, and in this way they teach
us just how much more flexible and expansive our own
creativity can be. The takeaway is that AI shakes us
(25:36):
loose from esthetic grooves that we might never leave on
our own, and AI is doing exactly that in science
as well. Just as one example, in material science and
drug design, AI is proposing molecular structures that humans just
wouldn't think to try already. It's nudging us to rethink
(25:57):
what counts as a reasonable chemical design, and we're seeing
the same thing in AI assisted math proofs. There are
systems like lean and GPT fueled theorem provers. There are
these systems that are suggesting lemmas or strategies that mathematicians
(26:18):
just hadn't thought of. They could have thought of it,
they just never did. These AI math proofs sometimes come
out strange or elegant or messy, but in all cases
they force mathematicians to think differently about structure and possibility.
So AI can serve as a creativity engine in arts
(26:42):
and in science, pushing us outside of our intuition driven
blind spots. And there's something else that AI might be
able to help us with, which is our personal lives
and how AI can uncover blind spots there as well.
If you're a regular listener, you know I've talked on
many previous episodes about the ways in which we fool
(27:03):
ourselves because you're.
Speaker 2 (27:04):
Not one thing.
Speaker 1 (27:06):
Instead, you're built of many different drives, or, as I
wrote my book Incognito, you can think about the brain
as a team of rivals. So think about the way
that many people approach fitness. Someone might sign up for
an expensive year long gym membership because they're determined to
get in shape, and on paper, that looks like a
(27:26):
great decision. It's an investment in their health. But at
the same time, that person keeps telling themselves that they
just don't have the time to exercise, so months pass,
the membership goes unused, the rationalization continues. We've all seen
this sort of thing. An AI personal assistant reviewing their
spending and scheduling could point out the contradiction. It could say, Hey,
(27:51):
you know what, You've paid twelve hundred bucks for a
gym that you rarely visit, but you also spend seven
hours a week watching streaming shows and an hour every
day doom scrolling on social media. You say you don't
have time, but your calendar suggests otherwise. We humans are
so good at fooling ourselves with our little stories. We
(28:14):
smooth over inconsistencies without even realizing it. But a good
AI isn't going to buy the story. It's going to
see the data and it's going to highlight the bias
for us. And when it does, it forces us to
confront something we might have preferred to ignore.
Speaker 2 (28:31):
But in this way it can make us better.
Speaker 1 (28:34):
Finally, I'll just mention something in my life where I'm
noticing that AI is improving me. I have been driving
for decades, and knock on wood, I've never had an accident,
presumably because I'm a perfectly good driver. But for the
past half year, I haven't driven myself around too much
because I have a Tesla with full self driving mode
and I let it drive me everywhere. So you just
(28:57):
tell the Tesla where you want to go, and it
does all the every bit of it.
Speaker 2 (29:01):
And here's the important part.
Speaker 1 (29:03):
I've had to admit that it's a better driver than
I am. There's the obvious stuff, like the fact that
it never blinks or sneezes or gets distracted by something,
but instead it has cameras that take thirty six frames
per second and never ever rest, even for the length
of an eyeblink. But more than that is a deeper,
(29:23):
more subtle issue. It's taught me that there are certain
reactions I have that aren't optimized. For example, when some
car pulls down traffic in front of me, I tend
to slow down, but on full self driving mode, the
Tesla just keeps going about the same speed that it
was going. And while I would have thought that that
(29:44):
seems a little aggressive, it turns out to be just
fine because the other car gets up to speed, so
there was no real need to slow down. It turns
out it's not aggressive, it's just optimized. And I see
lots of examples of this subtle differences in the way
that it drives versus me, and I am learning from it.
(30:05):
And there's also a relationship point to be made here,
because my wife generally thinks that I am a backseat driver,
because I'll often react with my body when she's driving,
and she's a perfectly great driver, but I can't help
myself because I would slow down when someone pulls out
in front of me and she doesn't. But ever since
I've seen how AI drives, I'm no longer an insufferable passenger.
(30:30):
Now I know what optimized driving is, and I admit
that she was closer to it than I was. So
I don't know if I'm going to argue that AI
is going to help marriages, but maybe Okay, so let
me zoom out to the big picture. We have very
limited internal models, and the main thing we're going to
get from AI is an illumination of ideas outside the
(30:54):
borders of our models. And to this end, I was
thinking the other day about the twenty sixteen movie Rival.
If you haven't seen it, this is a wonderful science
fiction drama where a linguist is recruited by the US
military after mysterious alien spacecraft appear around the world, So
(31:14):
rather than focusing on flashy battles, the film centers on communication,
decoding the alien strange language to understand why they have come. Now,
there's good suspense, but the movie is actually quietly building
towards its philosophical core. It's a concept from linguistics known
as the Sapier Wharf hypothesis, which I've talked about on
(31:37):
a couple of previous episodes. This hypothesis suggests that the
language you speak doesn't only allow you to express your thoughts,
but more than that, it shapes the way you think
and even possibly how you perceive reality itself. So in
the movie Arrival, this hypothesis becomes the lens through which
(32:00):
the entire story unfolds. So I won't give away the
spoiler of the movie, but it pivots on this point.
Once the linguist learns the alien's language, she's able to
perceive and experience the world differently. Upon learning their language,
she now has powers that humans don't normally have. So
(32:22):
I think that's a good metaphor for thinking about our
moment with AI landing here like an alien species. Are
we going to be exposed to ideas and concepts that
expand our thinking. When we think about the arrival of AI,
it's tempting to frame it as a contest will the
(32:44):
machine replace the worker and the scientist, and the gamer
and the composer.
Speaker 2 (32:50):
But the more interesting story is not about competition.
Speaker 1 (32:55):
It's about collaboration and how AI is going to stretch
the boundaries of human imagination. Go to eagleman dot com
slash podcast for more information and to find further reading.
Check out my newsletter on substack and be a part
(33:15):
of the online chats there. And you can watch the
videos of Inner Cosmos on YouTube, where you can leave comments.
Until next time, I'm David Eagleman, and this is Inner Cosmos.
Will AI make humans better? And what does this have
to do with linguistics or the movie Arrival or self
driving cars or debate and video games and elections and
chess and the ancient game of go. Welcome to Intercosmos
(00:26):
with me David Eagleman. I'm a neuroscientist and an author
at Stanford and in these episodes we seek to understand
why and how our lives look the way they do.
In today's episode is about whether AI will make us
better humans. I often find myself totally flabbergasted by the
(00:58):
change that I've seen just in my lifetime. When I
was a really little kid, personal computers didn't exist, and
then we passed through a door and suddenly they did.
And I saved up my money and I got a
common or VIC twenty computer, which was something my parents
had never seen the likes of. And I felt like
I was living the largest change in human history, because
(01:20):
for the first time, everybody could have a machine that
would do all kinds of things. But that turned out
not to even be the biggest change, because the next
stop was even bigger, and that was the idea that
some people had to build a system where we could
keep information on computers and make computers talk to each
(01:42):
other such that if the Soviets bombed America, the important
information wouldn't just be stored on one computer, and messages
could follow different routes on the network, and that way
you had a very robust system for keeping information. That
was of course Arpanet, which became the Internet, and not
(02:02):
long after, the idea was introduced of a way that
everyone could use this giant network, not just with text,
but with graphics, and.
Speaker 2 (02:10):
That was the birth of the Worldwide Web.
Speaker 1 (02:13):
And as soon as that technology existed, then people started
figuring out what to do with it, and one young
man started to sell books over the Internet.
Speaker 2 (02:22):
And that became Amazon.
Speaker 1 (02:24):
And two graduate students at Stanford asked the question of
how the heck we were going to be able to
find information on this sprawling network, and they created a
way of measuring all the connections between web pages and
that gave the ability to search for it, and.
Speaker 2 (02:40):
That garage project grew into Google.
Speaker 1 (02:44):
And around the same time, a Harvard kid was thinking
about a better way to allow his classmates to get
to know each other, which was traditionally done at the
beginning of the year by printing a booklet with everybody's
picture and name, and that little booklet was called called
a Facebook, and he thought of digitizing that, and on
(03:04):
and on, people increasingly figured out how to take this
new technology and make things that would live on top
of it. And I felt very lucky to have experienced
two truly world changing inventions in my lifetime, and I
knew that had launched us into a world that was
so different from what my great grandparents could have possibly imagined.
(03:30):
But now here we are in the middle of a
third revolution. It's related to the first two computers and
the Internet, but it makes them look like warm up acts.
And that's the fact that we have created a new
intelligent species that we're going to be sharing the planet
with from now on. This is not to say that
(03:51):
AI has exactly the same type of intelligence that human
brains have, but obviously it has absorbed the higher knowledge
sphere of humankind and it can spit that back to
us with all sorts of remixes. Now, the question is,
are we in trouble because of this new invention? Have
we taken things a step too far? Well, let me
(04:15):
give you an example that's on people's minds. A few
months ago, a Swiss research team carried out a secret
experiment in online persuasion. They went to a forum on
the website Reddit, and in this forum, users post opinions
and invite other people to challenge them. And so these
(04:36):
researchers quietly unleashed a set of AI accounts to pretend
that they were people and to try to change other
people's minds. So their large language models LMS, they participated
in the debating just like any other human. These bots
wrote arguments, they engaged with users, They debated with the
(05:00):
hope of changing minds. Now, the researchers measured success by
tallying whenever an original poster publicly admitted that their mind
had been changed. So how did the bots do well?
They achieved up to an eighteen percent success rate in
changing people's minds. Now, the critical piece you need to
(05:22):
know is that the average human success rate is about
three percent. So the bots had absolutely crushed their human
competition in an environment teeming with intellectuals. The bots not
only survived, but they thrived. In fact, one of the
AI accounts climbed into the ninety ninth percentile of all
(05:45):
users on this subreddit. It racked up ten thousand karma
points along the way. Now, one of the most surprising
aspects of the experiment was how effectively a small team
of researchers, operating with modest academic resources, how they were
able to outperform essentially every human debater on the platform.
Speaker 2 (06:07):
So let that sink in for a second.
Speaker 1 (06:09):
You have a handful of graduate students armed with an
LM and a sneaky deployment strategy, and they quietly demonstrated
what it might look like if influence operations were scaled
by AI. Now, when this data was released, one of
the scariest parts to people was that nobody noticed that
(06:32):
these debaters were actually AI and not real humans. This
reddit channel prides itself as one of the most critically
minded communities on the platform. If any place could sniff
out an imposter, it should have been here. But for
the entire four months of the experiment, the bots played
along undetected by the humans. So that's worrisome. And there's
(06:57):
another issue too, which is the role of personal data.
When the bots were given just a basic profile of
the user, their aged, their location, and their political leaning,
the success rate bumped up by one percentage point. Now,
you might say wait a minute, who cares. One percentage
point is very small, but in political terms, one percent
(07:17):
could be enough to swing a lot of national elections.
The difference between nudging public sentiment one way or another
can come down to.
Speaker 2 (07:27):
Very tiny tweaks.
Speaker 1 (07:28):
So the lesson that surface here is that even minimal
personalization can significantly sharpen the edge of an AI's persuasive power.
So when this story surfaced recently about these debaters not
being real human beings, the reactions were very grim because
everyone realized that this small academic experiment illustrated what stealth
(07:55):
influence operations could look like in the near future. If
a handful of real searchers could do this undetected, what
happens when you have state actors or corporations or political
campaigns with real resources doing the same thing at a
scale thousands of times larger. Also, what's worse is that
(08:16):
the failure of everyone to detect that these were bots
that raises serious questions about how we can protect public
discourse as we enter this new future. If smart reditor
debaters couldn't spot the difference between a human and a bot,
what hope is there for broader audiences. So even though
(08:37):
the major AI companies have all pledged to avoid building
models with dangerous capabilities like manipulating public opinion on mass.
This Reddit experiment suggests the thresholds may already be easier
to cross than any of us had anticipated. What this
(08:57):
all means is that with major elections in future years,
were genuinely gonna have to worry about this. The halcyon
days are gone when we could assume that the replies
to our online messages came from a fellow human, and
the risks are real and the potential for abuse is obvious.
(09:19):
But for today's episode, we're gonna look at all this
from a very different angle. I think it might be
worth asking a different question, which is why did the
bots succeed? After all, they didn't hack people's brains with
neural interfaces. They didn't spread fear or disinformation. They didn't
(09:39):
manipulate emotions or overwhelm users with noise. They simply made
better arguments. They didn't insult people while arguing with them.
They didn't do little jabs and digs. They just made
good arguments empathetically. The bots present their points calmly and
(10:02):
rationally and persuasively, and when users changed their minds.
Speaker 2 (10:08):
It wasn't because they had been tricked.
Speaker 1 (10:10):
It was because they recognized that another perspective was worth considering.
(10:32):
Humans often change their minds when faced with sound reasoning
that can be backed up, and mind changing is a
great thing. It means you're willing to reconsider some closely
held opinion when the facts or logic warranted. It's a
mark of intellectual strength, not a sign that you've been tricked.
(10:54):
In that light, the success of AI debaters isn't necessarily
a story about manipulation. It could also be a story
about raising the bar of debating. Now, as I followed
the outcome of this AI debater study, it struck me
that there may be a helpful precedent for thinking about
(11:15):
this moment, for thinking about how AI could.
Speaker 2 (11:18):
Actually improve us.
Speaker 1 (11:20):
So think about the events that unfolded in the hyper
competitive world of chess. If you can remember back to
nineteen ninety seven, IBM had built an AI system called
Deep Blue, and it defeated the world champion, Gary Kasprov.
This was seen as a seismic moment. The game of
(11:41):
chess had been quote unquote solved, and everyone worried that
human mastery was obsolete and then about two decades later,
Alpha Go beat the world's top Go player, and the
choir of concerned voices grew louder. But then something unexpected happened.
(12:02):
In May of twenty seventeen, the world's number one Go player,
Could G, faced off against his toughest opponent. G was
the reigning champion in Go, and you know this is
the game where two players use smooth black rocks or
white rocks to surround more territory than their opponent. In
G's case, he was playing against AI. He was playing
(12:25):
against Alpha Go, which had been trained on many millions
of games, and it had deeply absorbed the statistics of
possible plays.
Speaker 2 (12:35):
So G lost the first game.
Speaker 1 (12:38):
Alpha Go had pulled moves that none of G's human
opponents had ever thought of, and then G lost the
second game. The fact is he didn't stand a chance.
The AI had won over a human in a game
that was way more complex than chess, and subsequent versions
of this AI will without doubt continue to win ever more.
(13:01):
But that's not the interesting part of the story. The
interesting part is what happened next. G got over his
embarrassment and he became mesmerized by what the heck had
just transpired. He studied the games that he lost. Now,
before he had played Alpha Go, G had won most
(13:23):
of the games against his human opponents, But after he
played Alpha Go, he was able to beat his human
opponents even more easily. In other words, after his species
shaming defeat in twenty seventeen, G went on to play
twelve straight matches against fellow humans, and he won.
Speaker 2 (13:42):
Them all in a row. Now, what had happened.
Speaker 1 (13:46):
G had been exposed to new kinds of moves and
strategies that AlphaGo was pulling off that lay outside the
traditional ideas. All these moves were legal and pop pole,
but they were different from what had been played over
the previous twenty five hundred years. For go eficionados, this
(14:08):
included novelties like playing a stone directly diagonal to your
opponent's loan stone, or commonly playing six space extensions while
humans tend to prefer five space. So G reported that
playing against the AI was like opening a door to
another world. Some people worry that AI might make games
(14:31):
of Chess and Go irrelevant, but amazingly, that is not
what's happened. When AI Trump's Chess champions and Go champions
It does so with moves that seem inhumanly creative, but
all the moves are allowed by the rules. Humans simply
never thought to go there before. And the key is
(14:53):
that once the moves are seen by humans, then they're
easily incorporated into our models. GE's experience with Alpha Go
illuminated new nooks and crannies in his landscape. It exposed
pathways that have never been lit up before. So AI
(15:14):
immediately became a tool for steep improvement. And nowadays all
chests and Go players above a certain level they all
train with AI. They study the surprising and sometimes counterintuitive
and alien strategies of the artificial mind, and just like Koje,
today's grand masters play a deeper and more creative game
(15:38):
than ever before. In other words, many commentators are worried
that AI is going to leave humans far behind, and
in some respects that's true.
Speaker 2 (15:48):
But as computation improves, so will we.
Speaker 1 (15:53):
AI will illuminate dark parts of our maps, allowing us
to see new roads we didn't even suspect. The key
point I want to make here is that instead of
dampening human excellence, AI sparked a renaissance. In these games,
human minds are elevated by learning from our artificial cousins.
(16:14):
And if it were just chests and go, that's one thing.
But I think we can detect this pattern happening all
over the place. For example, the same thing has happened
in poker. Professional poker players master the art of bluffing.
They project strength when they're weak or weakness when they're strong.
This has always been seen like a deep test of
(16:35):
human psychology. The players read faces, they watch for micro expressions,
they guess intentions, So in this way, poker is a
really human game. But then Carnegie Mellon cooked up an
AI called Librotus. Obviously, it doesn't have a face, it
doesn't sweat, it doesn't blink, so it wasn't really like
playing another human. So Lebronis started playing poker, and it
(16:59):
did things that baffled the human players. The AI began
making betting choices that looked bizarre. Sometimes it over bet
the pot by huge margins, a move that the human
players considered reckless. Other times it made really tiny bets
in situations where no human would bother and the human
(17:19):
players dismissed this all as network nonsense. But as the
games went on, they realized these inhuman strategies were working.
The AI was winning, and it was doing so by
reinventing the language of poker. And now, just like with
chess and go, professional poker players study these strategies. You
(17:43):
have human players adopting patterns of play that just didn't
exist before the machines taught them. AI made humans better
poker players. And let me give you another example. There's
a strategy video game called StarCraft two where players build
armies and manage resources and try to outwit their opponent.
(18:04):
So the company DeepMind set some AI agents on it,
and those agents showed humans an entirely different way of
going about running the strategy. At first, people thought the
AI play seemed unfair and robotic, but then human players
began to adapt. They started rethinking their own strategies. They
(18:26):
started redesigning their build orders. They discovered things like sometimes
sacrificing entire groups of units early in the game could
produce long term advantages. These were not moves that humans
had come up with, or possibly ever would have come
up with, but once they were revealed, they became part
(18:47):
of the human playbook. Or take a different video game
called Dota two. You've got professional players. But then open
AI built a system that used strategies that looked to
human professionals like they were clumsy. The AI would push
aggressively when humans would retreat or take risks that seemed absurd,
(19:08):
but more often than not, the machines won, and just
like with Chests and Go and StarCraft, the human professionals
were forced to ask themselves had they been playing with
blinders on all along? After all, the machines were uncovering
landscapes that we just never knew were there. In other words,
(19:29):
whenever AI discovers new pathways inside the universe of a game,
it illuminates something for us about the fence lines of
our own imagination. Humans had decades to explore video games,
and centuries to explore poker, and millennia to explore the
games of Chess and Go. But within weeks AI uncovered
(19:53):
strategies that had never even struck us. And this is
the point of today's episode. AI can expand the possibility
space for us. What we've seen in the past few
years is the limits of our imagination. Even in domains
we thought we had mastered, our internal models might be
(20:14):
a lot more narrow than we had ever realized. But
by playing with machines, we're learning how they think, and
more importantly, how we might think differently. So beyond chests
or go or video games, the bigger game is how
we can expand our own internal models. A lot of
people cast this as man versus machine, but I think
(20:36):
a more productive lens is seeing it as man learning
from machine. The prize is a new way of seeing
the game, and maybe, by extension, a new way of
seeing the world.
Speaker 2 (20:51):
So let's come back to that.
Speaker 1 (20:53):
Amazing Swiss experiment where they made debate bots on Reddit
that performed way better than humans. I suggest it's possible
and maybe even likely, that a similar dynamic is going
to unfold in the world of persuasive argumentation that happened
in chess and go in video games. If AI agents
(21:14):
can model the best forms of debate, clear and structured
and empathetic and rational, then we humans can learn something
from our artificial cousins. We can try out new moves,
We can sharpen our skills on digital grinding stones. Imagine
(21:36):
a future where students practice crafting arguments by debating highly
skilled AI tutors. Imagine online discussions becoming more useful because
users have gotten used to high quality exchanges. Imagine politicians
and journalists and everyday citizens pushed to improve their things
(22:00):
and better articulate their positions. So, rather than dumbing down conversation,
the rise of high performing debate bots could nudge public
discourse toward a new level of reasoned discussion. If that
turns out to be the case, then we may come
to see AI not as an enemy but as a
(22:22):
sparring partner, just like in Chest and Go, and that
has very different implications. Now, some people will take just
(22:45):
the opposite position. If AI is so much better at
debating than we are, won't that cause us to lose
the skill entirely because there doesn't seem to be a
point in being a good debater if the computer can
do it better. But I think this is not an
issue because even when we extrapolate this era of AI,
(23:07):
people will still be talking with each other most of
the time. You'll still be with your family over dinner,
or with your friends over a coffee, or arguing with
your neighbor about where the fence goes, or debating with
a stranger at a town hall or whatever. It's not
like we are plugging into the matrix through a port
in the back of our neck, and we're only going
(23:27):
to be communicating with machines. We are intensely social creatures,
and the success of our species has been in part
because of our massive sociability. So I think we will
debate other humans all the time, and AI is just
going to train us to be a little bit better
(23:49):
at it. Now, I don't want to minimize the risks
we're facing. We need audit tools and authentication systems that
can verify whether content was written by humans or AIA.
We need technical solutions like water marking or cryptographic content authentication,
and providence tracking. This is all essential, but beyond these
(24:11):
defensive measures, we should also recognize the opportunity because like
the chests and go engines that reshaped how champion players think,
debate bots will reshape how we reason and argue and
understand one another. If we take this on correctly, AI
(24:32):
might just up our game. It's still early days, but
my impression so far is that AI is already playing
this role in many areas, including for example, the arts.
At least in some ways, it's amplifying human creativity. It's
like a jazz partner who plays a riff you weren't expecting,
(24:52):
or a painting mentor who introduces a color that you
never thought to use. AI forces us out of our
group in the arts. It shows us that the boundaries
of our imagination can be stretched by encountering minds, even
synthetic minds that think differently. These AI systems presumably don't
(25:13):
have esthetic tastes or emotional longing in the way that
we do, but they're awesome at doing remixes and trying
strange new things out, and in this way they teach
us just how much more flexible and expansive our own
creativity can be. The takeaway is that AI shakes us
(25:36):
loose from esthetic grooves that we might never leave on
our own, and AI is doing exactly that in science
as well. Just as one example, in material science and
drug design, AI is proposing molecular structures that humans just
wouldn't think to try already. It's nudging us to rethink
(25:57):
what counts as a reasonable chemical design, and we're seeing
the same thing in AI assisted math proofs. There are
systems like lean and GPT fueled theorem provers. There are
these systems that are suggesting lemmas or strategies that mathematicians
(26:18):
just hadn't thought of. They could have thought of it,
they just never did. These AI math proofs sometimes come
out strange or elegant or messy, but in all cases
they force mathematicians to think differently about structure and possibility.
So AI can serve as a creativity engine in arts
(26:42):
and in science, pushing us outside of our intuition driven
blind spots. And there's something else that AI might be
able to help us with, which is our personal lives
and how AI can uncover blind spots there as well.
If you're a regular listener, you know I've talked on
many previous episodes about the ways in which we fool
(27:03):
ourselves because you're.
Speaker 2 (27:04):
Not one thing.
Speaker 1 (27:06):
Instead, you're built of many different drives, or, as I
wrote my book Incognito, you can think about the brain
as a team of rivals. So think about the way
that many people approach fitness. Someone might sign up for
an expensive year long gym membership because they're determined to
get in shape, and on paper, that looks like a
(27:26):
great decision. It's an investment in their health. But at
the same time, that person keeps telling themselves that they
just don't have the time to exercise, so months pass,
the membership goes unused, the rationalization continues. We've all seen
this sort of thing. An AI personal assistant reviewing their
spending and scheduling could point out the contradiction. It could say, Hey,
(27:51):
you know what, You've paid twelve hundred bucks for a
gym that you rarely visit, but you also spend seven
hours a week watching streaming shows and an hour every
day doom scrolling on social media. You say you don't
have time, but your calendar suggests otherwise. We humans are
so good at fooling ourselves with our little stories. We
(28:14):
smooth over inconsistencies without even realizing it. But a good
AI isn't going to buy the story. It's going to
see the data and it's going to highlight the bias
for us. And when it does, it forces us to
confront something we might have preferred to ignore.
Speaker 2 (28:31):
But in this way it can make us better.
Speaker 1 (28:34):
Finally, I'll just mention something in my life where I'm
noticing that AI is improving me. I have been driving
for decades, and knock on wood, I've never had an accident,
presumably because I'm a perfectly good driver. But for the
past half year, I haven't driven myself around too much
because I have a Tesla with full self driving mode
and I let it drive me everywhere. So you just
(28:57):
tell the Tesla where you want to go, and it
does all the every bit of it.
Speaker 2 (29:01):
And here's the important part.
Speaker 1 (29:03):
I've had to admit that it's a better driver than
I am. There's the obvious stuff, like the fact that
it never blinks or sneezes or gets distracted by something,
but instead it has cameras that take thirty six frames
per second and never ever rest, even for the length
of an eyeblink. But more than that is a deeper,
(29:23):
more subtle issue. It's taught me that there are certain
reactions I have that aren't optimized. For example, when some
car pulls down traffic in front of me, I tend
to slow down, but on full self driving mode, the
Tesla just keeps going about the same speed that it
was going. And while I would have thought that that
(29:44):
seems a little aggressive, it turns out to be just
fine because the other car gets up to speed, so
there was no real need to slow down. It turns
out it's not aggressive, it's just optimized. And I see
lots of examples of this subtle differences in the way
that it drives versus me, and I am learning from it.
(30:05):
And there's also a relationship point to be made here,
because my wife generally thinks that I am a backseat driver,
because I'll often react with my body when she's driving,
and she's a perfectly great driver, but I can't help
myself because I would slow down when someone pulls out
in front of me and she doesn't. But ever since
I've seen how AI drives, I'm no longer an insufferable passenger.
(30:30):
Now I know what optimized driving is, and I admit
that she was closer to it than I was. So
I don't know if I'm going to argue that AI
is going to help marriages, but maybe Okay, so let
me zoom out to the big picture. We have very
limited internal models, and the main thing we're going to
get from AI is an illumination of ideas outside the
(30:54):
borders of our models. And to this end, I was
thinking the other day about the twenty sixteen movie Rival.
If you haven't seen it, this is a wonderful science
fiction drama where a linguist is recruited by the US
military after mysterious alien spacecraft appear around the world, So
(31:14):
rather than focusing on flashy battles, the film centers on communication,
decoding the alien strange language to understand why they have come. Now,
there's good suspense, but the movie is actually quietly building
towards its philosophical core. It's a concept from linguistics known
as the Sapier Wharf hypothesis, which I've talked about on
(31:37):
a couple of previous episodes. This hypothesis suggests that the
language you speak doesn't only allow you to express your thoughts,
but more than that, it shapes the way you think
and even possibly how you perceive reality itself. So in
the movie Arrival, this hypothesis becomes the lens through which
(32:00):
the entire story unfolds. So I won't give away the
spoiler of the movie, but it pivots on this point.
Once the linguist learns the alien's language, she's able to
perceive and experience the world differently. Upon learning their language,
she now has powers that humans don't normally have. So
(32:22):
I think that's a good metaphor for thinking about our
moment with AI landing here like an alien species. Are
we going to be exposed to ideas and concepts that
expand our thinking. When we think about the arrival of AI,
it's tempting to frame it as a contest will the
(32:44):
machine replace the worker and the scientist, and the gamer
and the composer.
Speaker 2 (32:50):
But the more interesting story is not about competition.
Speaker 1 (32:55):
It's about collaboration and how AI is going to stretch
the boundaries of human imagination. Go to eagleman dot com
slash podcast for more information and to find further reading.
Check out my newsletter on substack and be a part
(33:15):
of the online chats there. And you can watch the
videos of Inner Cosmos on YouTube, where you can leave comments.
Until next time, I'm David Eagleman, and this is Inner Cosmos.
Host
David Eagleman
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.
What Are We Even Doing? with Kyle MacLachlan
Join award-winning actor and social media madman Kyle MacLachlan on “What Are We Even Doing,” where he sits down with Millennial and Gen Z actors, musicians, artists, and content creators to share stories about the entertainment industry past, present, and future. Kyle and his guests will talk shop, compare notes on life, and generally be weird together. In a good way. Their conversations will resonate with listeners of any age whose interests lie in television & film, music, art, or pop culture.