The Ezra Klein Show: Best Of: Why Adults Lose the ‘Beginner’s Mind’
New York Times Opinion 5/9/23 - Episode Page - 1h 2m - PDF Transcript
Hey, it's Ezra. I am out today, but I wanted to share a past favorite episode with the
psychologist Allison Gopnik recorded in April 2021.
One thing I like about this particular episode is that there are parts of it, and you're
going to know which ones they are that are more relevant today and relevant for very
different reasons than when we recorded it. I like episodes at age well. Enjoy.
I'm Ezra Klein, and this is the Ezra Klein Show.
It probably won't surprise you that I'm one of those parents who reads a lot of books
about parenting, and they're mostly bad, particularly the books for dads. So many of those books
have this weird, dude, you're going to be a dad, bro, tone. It's a terrible literature.
But one of the great finds for me in the parenting book world has been Allison Gopnik's work.
Gopnik runs the Cognitive Development and Learning Lab at UC Berkeley. She's in both
the Psychology and Philosophy departments there. She's part of the AI working group there.
And one of the things about her work, the thing that sets it apart for me, is she uses
children and studies children to understand all of us. She takes childhood seriously as
a phase in human development. And why not, right? You're watching consciousness come
online in real time. You're watching language and culture and socials being absorbed and
learned and changed, importantly changed. Her books haven't just changed how I look
at my son. They've really changed how I look at myself, how I look at all of us. And one
of them in particular that I read recently is The Philosophical Baby, which blew my mind
a little bit, because what she does in that book is show through a lot of experiments
and research that there is a way in which children are a lot smarter than adults. I
think it's the right way to say that. A way in which their strangest, silliest-seeming
behaviors are actually remarkable. This is her core argument. Children are tuned to learn.
And when you tune a mind to learn, it actually has to work really differently than a mind
that already knows a lot. The efficiency that our minds develop as we get older, it has
amazing advantages. Unlike my son, and I don't want to brag here, unlike my son, I can make
it from his bedroom to the kitchen, but that any stops along the way, I can just get right
there. But also unlike my son, I take so much for granted. I have so much trouble actually
taking the world on its own terms and trying to derive how it works. I've learned so much
that I've lost the ability to unlearn what I know. And that means I've also sometimes
lost the ability to question things correctly. So this isn't just a conversation about kids
or for parents. It's a conversation about humans for humans. We spend so much time and
effort trying to teach kids to think like adults. A message of Gopnik's work, and one I take
seriously, is we need to spend more time and effort as adults trying to think more like kids.
As always, my email is aserclinenshow at nytimes.com. If you've got something to teach me, but
here is Alison Gopnik. You write that children aren't just defective adults, primitive grownups
who are gradually attaining our perfection and complexity. Instead, children and adults are
different forms of Homo sapiens. How so?
Well, from an evolutionary biology point of view, one of the things that's really striking is this
relationship between what biologists call life history, how our developmental sequence unfolds
and things like how intelligent we are. And there's a very, very general relationship between
how long a period of childhood an organism has and roughly how smart they are, how big their
brains are, how flexible they are. And an idea that I think a lot of us have now is that part of
that is because you've really got these two different creatures. So you've got one creature
that's really designed to explore, to learn, to change. That's the child form. And then you've got
this other creature that's really designed to exploit, as computer scientists say, to go out,
find resources, make plans, make things happen, including finding resources for that wild, crazy
explorer that you have in your nursery. And the idea is that those two different developmental
and evolutionary agendas come with really different kinds of cognition, really different kinds of
computation, really different kinds of brains. And I think with very different kinds of experiences
of the world. So the very way that you experience the world, your consciousness is really different
if your agenda is going to be get the next thing done, figure out how to do it, figure out what
the next thing to do after that is versus extract as much information as I possibly can from the
world. And I think adults have the capacity to some extent to go back and forth between those two
states. But I think that babies and young children are in that explorer state all the time. That's
really what they're designed to do. They're like a different kind of creature than the adult. You
know, you sort of might think about, well, are there other ways that evolution could have solved
this explorer exploit trade off this problem about how do you get a creature that can do things, but
can also learn things really widely. And Peter Godfrey Smith's wonderful book, I've just been
reading a Metazoa talks about the octopus. And the octopus is very puzzling, because the octaves
don't have a long childhood. And yet they seem to be really smart. And they have these big brains
with lots of neurons. But it also turns out that octaves actually have divided brains. So they have
one brain in the center in their head. And then they have another brain or maybe eight brains in
each one of the tentacles. And if you actually watch what the octaves do, the tentacles are out
there doing the explorer thing, they're getting information, figuring out what the water is like.
And then the central head brain is doing things like saying, okay, now it's time to squirt, now it's
time to get food. So my thought is that we could imagine an alternate evolutionary path by which
each of us was both a child and an adult. So, you know, imagine if your arms were like your two
year old, right, so that you were always trying to get them to stop exploring because you had to get
lunch. I suspect that may be what the consciousness of an octo is like. Now, we're obviously not like
that. But I think even human adults, that might be an interesting kind of model for some of what it's
like to be a human adult in particular. So I think we have children who really have this explorer
brain and in this explorer experience, they're kind of like our tentacles. They're going out and
figuring things out in the world. And then we have adults who are really the head brain, the one
that's actually going out and doing things. But I think even as adults, we can have this kind of split
brain phenomenon where a bit of our experience is like being a child again and vice versa.
I feel like that's an answer is going to launch 100 science fiction short stories.
Because people imagine this race you're describing here. One of the things I really liked about this is
that it pushes towards a real respect for the child's brain. If one defined intelligence as like the
ability to learn and to learn fast and to learn flexibly. A two year old is probably a lot more
intelligent right now than I am. I have more knowledge and I have more experience and I have more
ability to exploit existing learnings. But they have more capacity and flexibility and change
ability. Is that right? And to the extent it is, what gives it that flexibility? What are the
trade-offs to have that flexibility?
Yeah. So I think a really deep idea that comes out of computer science originally, in fact, came out
of the original design of the computer, is this idea of the explore exploit trade-off is what they
call it. So if you're thinking about intelligence, there's a real genuine trade-off between your
ability to explore as many options as you can versus your ability to quickly efficiently commit to
a particular option and implement it. And it turns out that even if you just do the math, it's
really impossible to get a system that optimizes both of those things at the same time. That is
exploring and exploiting simultaneously because they're really deeply intentioned with one another.
And the way that computer scientists have figured out to try to solve this problem very
characteristically is give the system a chance to explore first, give it a chance to figure out all
the information. And then once it's got the information, it can go out and it can exploit later
on. So it's explore first and then exploit. And I think that evolution has used that strategy in
designing development, human development in particular, because we have this really long
childhood. But I think you can see the same thing in non-human animals and not just in mammals, but
in birds and maybe even in insects. So you see this really deep tension, which I think we're
facing all the time between how much are we considering different possibilities and how much
are we acting efficiently and swiftly? There's, again, an intrinsic tension between how much you
know and how open you are to new possibilities. So again, just sort of something you can formally
show is that if I know a lot, then I should really rely on that knowledge. And I should, to some
extent, discount something new that somebody tells me. Whereas if I don't know a lot, then almost
by definition, I have to be open to more knowledge. But I think it's more than just the fact that you
have what the Zen masters call beginner's mind, that you start out not knowing as much. I think we
can actually point to things like the physical makeup of a child's brain and an adult brain that
makes them differently adapted for exploring and exploiting.
Use some work on this. What does look different in the two brains?
So there's two big areas of development that seem to be different. So one of them is that the young
brain seems to start out making many, many new connections. So what you'll see when you look at
a chart of synaptic development, for instance, is you've got this early period when many, many, many
new connections are being made. And then you've got this later period where the connections that are
used a lot that are working well, they get maintained, they get strengthened, they get to be
more efficient. And then the ones that aren't are pruned as neuroscientists say they kind of
disappear. The consequence of that is that you have this young brain that has a lot of what
neuroscientists call plasticity, it can change really easily, essentially. But it's not very good at
like putting on its jacket and getting into preschool in the morning. It's not very good at
doing anything that is the sort of things that you need to act well. And it's especially not good
at things like inhibition. It's especially not good at doing things like having one part of the
brain restrict what another part of the brain is going to do. So that's one change. This change
from this lots of local connections, lots of plasticity to something that's got longer and
more efficient connections, but is less changeable. The other change that's particularly relevant to
humans is that we have the prefrontal cortex, that's the part of our brain, that's sort of the
executive office of the brain where long term planning inhibition focus, all those things seem
to be done by this part of the brain. And what happens with development is that that part of the
brain, that executive part gets more and more control over the rest of the brain as you get
older. So that the ability to have an impulse in the back of your brain and the front of your
brain can come in and shut that out. Or there's abstraction in the back of your brain, something
that is in your visual field that isn't relevant to what you do. And the frontal part can literally
shut down that other part of your brain. But that process takes a long time. So when you start
out, you've got much less of that kind of frontal control, more of, I guess, in some ways, almost
more like the octaves where parts of your brain are doing their own thing. And then as you get
older, you get more and more of that control.
And is that the dynamic that leads to this spotlight consciousness lantern consciousness
distinction? And can you talk about that? Because I don't think about it all the time.
So those are two really, really different kinds of consciousness. One kind of consciousness, this
is an old metaphor is to think about attention as being like a spotlight. It comes in, it
illuminates the thing that you want to find out about. And you don't see the things that are on
the other side. And I think that in other states of consciousness, especially the state of
consciousness, you're in when you're a child. But I think there are things that adults do that
put them in that state as well. You have something that's much more like a lantern. So you're
actually taking in information from everything that's going on around you. And the most important
thing is, is this going to teach me something? Is this new? Is this interesting? Is this curious?
Rather than focusing your attention and consciousness on just one thing at a time. So a lot of the
theories of consciousness start out from what I think of as professorial consciousness. So
surprise, surprise, when philosophers and psychologists are thinking about consciousness,
they think about the kind of consciousness that philosophers and psychologists have a lot of
the time. And that sort of consciousness is say, you're sitting in your chair, you have the paper
to write, you're desperately trying to focus on the specific things that you said that you would
do. And then you kind of get distracted in your mind wanders a bit, and you start ruminating
about other things. And that kind of goal directed focused consciousness, which goes very
much with the sense of a self. So there's a me that's trying to finish up the paper or answer
the emails or do all the things that I have to do. That's really been the focus of a lot of
theories of consciousness is if that kind of consciousness was what consciousness was all
about. And we even can show neurologically that for instance, what happens in that state is when
I attend to something when I pay attention to something, what happens is the thing that I'm
paying attention to becomes much brighter and more vivid. And I actually shut down all the other
things that I'm not paying attention to, you can even see that in the brain. So the part of your
brain that's relevant to what you're attending to becomes more active, more plastic, more
changeable. And the other nearby parts get shut down again, inhibited. So there's a really nice
picture about what happens in professorial consciousness. That's kind of how consciousness
works. And again, maybe not surprisingly, people have acted as if that kind of consciousness is
what consciousness is really all about. Like that's really what you want when you're conscious. And
what I would argue is there's all these other kinds of states of experience and not just me other
philosophers as well. There's all these other kinds of ways of being sentient ways of being aware
of the ways of being conscious that are not like that at all. So one interesting example that
there's actually some studies of is think about when you're completely absorbed in a really
interesting movie, you're kind of gone, yourself is gone, you're not deciding what to pay attention
to in the movie, the movie is just completely captivating. In the state of that focused goal
directive consciousness, those frontal areas are very involved and very engaged. And there seem to
actually be two pathways. One of them is the one that sort of hears the goal directed pathway, what
they sometimes call the task dependent activity. And then the other one is what sometimes called the
default mode. And that's the sort of ruminating or thinking about the other things that you have to
do being in your head, as we say, is the other mode. When you look at someone who's in the scanner,
who's really absorbed in a great movie, neither of those parts are really active. And instead, other
parts of the brain are more active. And that brain, the brain of the person who's absorbed in the
movie looks more like the child's brain.
But now, you know, whether you're a philosopher or not, you know, or an academic or a journalist or
just somebody who spends a lot of time on their computer or a student, we now have a modernity
that is constantly training something more like spotlight consciousness, probably more so than
would have been true at other times in human history. And something that I took from your book is
that there is the ability to train, or at least experience different kinds of consciousness through
different kinds of either experiences like travel, or you talk about meditation. But one of the
thoughts that triggered for me is that somebody who's been pretty involved in meditation for the last
decade or so, there's a real dominance of the past and a style concentration meditation, single
point meditations, just watch the breath, just think about the breath right at the edge of the
nostril. And without taking anything away from from that tradition, it made me wonder if one reason
that has become so dominant in America, and particularly in Northern California, is because
it's a very good match for the kind of concentration that our and consciousness at our
economy is constantly trying to develop in us, this sort of, you know, this get things done, you
know, be very focused, don't ruminate too much, you know, like a neoliberal form of consciousness.
Do you think there's something to that?
I think that there's a paradox about, for example, going out and saying, I am going to meditate and
stop trying to get goals, because I have this goal, which is I want to be a much better meditator.
And I have done a bit of meditation and workshops. And it's always a little amusing when you see,
you know, the young men who are going to prove that they're better at meditating, they can sit for
longer than anybody else can. But I think it's important to say, when you're thinking about
things like meditation, or you're thinking about alternative states of consciousnesses in general,
that there's lots of different alternative states of consciousness. So it isn't just a choice between
lantern and spotlight. There's lots of different ways that we have of being in the world, lots of
different kinds of experiences that we have. And I suspect that they each come with a separate, a
different kind of focus, a different way of being. And in meditation, you can see the contrast
between some of these more pointed kinds of meditation versus what's sometimes called open
awareness meditation. So open awareness meditation is when you're not just focused on one thing,
when you try to be open to everything that's going on around you. And the phenomenology of that is
very much like this kind of lantern that everything at once is illuminated. And I think that kind of
open ended meditation and the kind of consciousness that it goes with is actually a lot like things
that, for example, the romantic poets like Wordsworth talked about. So there's this lovely concept
that I like of the numinous. And sometimes it's connected with spirituality, but I don't think it
has to be. It's this idea that you're going through the world. And often quite suddenly, if
you're an adult, everything in the world seems to be significant and important and important and
significant in a way that makes you insignificant by comparison. My colleague, Dr. Keltner, has
studied awe and awe is kind of an example of this. But the numinous is sort of turns up the dial on
awe. And part of the numinous is it doesn't just have to be about something that's bigger than you
like a mountain. It could just be your garden or the street that you're walking on. And suddenly
that becomes illuminated and everything around you becomes illuminated and you yourself sort of
disappear. And I think that's kind of the best analogy I can think of for the state that the
children are in. And it's worth saying, you know, it's not like the children are always in that
state. So the children, perhaps because they spend so much time in that state also can be fussy
and cranky and, you know, desperately wanting their next meal or desperately wanting comfort,
they're not always in that kind of broad state. But I think they spend much more of their time in
that state. That's more like their natural state than adults are.
Do you think for kids that play or imaginative play should be understood as a form of consciousness
estate?
Yeah, that's a really good question. So there's really a kind of coherent whole about what
childhood is all about. So if you think from this broad evolutionary perspective about these
creatures that are designed to explore, I think there's a whole lot of other things that go
with that. So one thing that goes with that is this broad based consciousness. But another thing
that goes with it is the activity of play. And if you think about play, the definition of play is
that it's the thing that you do when you're not working. Now it's not a form of experience and
consciousness so much. But it's a form of activity. It's a form of actually doing things that
nevertheless have this characteristic of not being immediately directed to a goal. If you look
across animals, for example, very characteristically, it's the young animals that are playing
across an incredibly wide range of different kinds of animals. Sometimes, you know, if they're
mice, they're play fighting. And if they're crows, they're playing with twigs and figuring out how
they can use the twigs. So what goes on in play is different. But it's really fascinating that it's
the young animals who are playing. And all of the theories that we have about play are plays
another form of this kind of exploration. So it's another way of having this explore state of
being in the world. Now it's not so much about your visually taking in all the information around
you the way that you do when you're exploring. Now it's more like you're actually doing things on
the world to try to explore the space of possibilities. You know, another thing that people
point out about play is play is fun. There's a certain kind of happiness and joy that goes
with being in that state when you're just playing. And again, it's not the state that kids are in
all the time, but the state that they're in a lot of the time and a state that they're in when
they're when they're actually engaged in play. One of the things that's really fascinating that's
coming out in AI now, and I've been spending a lot of time collaborating with people and
computer science at Berkeley who are trying to design better artificial intelligence systems. The
current systems that we have, I mean, the language is they're designed to optimize. They're really
exploit systems. What you do with these systems is say, here's what your goal is. Now you go out
and maximize that goal. And it turns out that if you have a system like that, it will be very good
at doing the things that it was optimized for, but not very good at being resilient, not very good
at changing when things are different. I've been really struck working with people in robotics,
for example, when people say, well, the robots have trouble generalizing, they don't mean they
have trouble generalizing from driving a Tesla to driving a Lexus. They mean they have trouble
going from putting the block down at this point to putting the block down a centimeter to the
left. I mean, they really have trouble generalizing even when they're very good. And it turns out
that if you get these systems to have a period of play, where they can just be generating things
in a, in a wilder way, or get them to train on a human playing, they end up being much more
resilient. They're much better at generalizing, which is of course, the great thing that children
are also really good at.
I was thinking about how a moment ago you said, play is what you do when you're not working. And
I was thinking, it's absolutely not what I do when I'm not working. I'm constantly like you
sitting here be like, don't work. And that's not playing. And in fact, I think I've lost a lot
of my capacity for play, you know, I've trained myself to be productive so often that it's
sometimes hard to put it down. And it takes like actual dedicated effort to not do things that feel
like work to me. What's lost in that? Because I think there's cultural pressure to not play. But I
think that that your research and some of the others suggest maybe maybe we've made a terrible
mistake on that by not honoring play more.
Yeah, I think there's a lot of evidence for that. And it's interesting that as I say, the, the hard
headed engineers who are trying to, you know, do things like design robots, are increasingly
realizing that play is something that's going to actually be able to get you systems that do
better in going through the world. Part of the problem. And this is a general explorer exploit
problem. Part of the problem with play is, if you think about it in terms of what its long term
benefits are going to be, then it isn't play anymore. And if you sort of set up any particular
goal, if you say, Oh, well, if you play more, you'll be more robust or more resilient. And you say,
Okay, so now I want to design you to do this particular thing. Well, then you're always going
to do better by just optimizing for that particular thing than by playing. So what play is really
about is about this ability to change, to be resilient in the face of lots of different
environments, in the face of lots of different possibilities. It's about dealing with something
new or unexpected. And it's interesting that if you look at what might look like a really
different literature, look at studies about the effects of preschool on later development in
children. So when they first started doing these studies, where you looked at the effects of an
enriching preschool, and these were play based preschools, the way preschools still are to some
extent, and certainly should be and have been in the past. So, you know, basically, you put a
child in a rich environment where there's lots of opportunities for play. And it turned out that if
you looked at things like just how well you did on a standardized test, after a couple of years, the
effects seem to sort of fade out. And that was an argument against early education. But it turns
out that if you look 30 years later, you can have these sleeper effects, where these children who
played are not necessarily getting better grades three years later, but they're not going to prison,
and their health is better, and their salaries are higher. And what that suggests is the things that
having a lot of experience with play was letting you do was to be able to deal with unexpected
challenges better, rather than that it was allowing you to attain any particular outcome. And it
really makes it tricky if you want to do evidence based policy, which we all want to do. And
that's optimistic.
Well, or what at least some people want to do any kind of metric that you said, almost by definition,
if it's the metric, you're going to do better if you teach to the test. So there's always this
temptation to do that, even though the advantages that play gives you seem to be these advantages of
robustness and resilience. So for instance, if you look at rats, and you look at the rats who get to
do play fighting versus rats who don't, it's not that the rats who play can do things that the rats
who can't play tap, like every specific fighting technique, the rats will have. But if you look at
their subtlety, at their ability to deal with context, at their ability to decide, when should I do
this versus that? How should I deal with the whole ensemble that I'm in? That's where play has its
great advantages.
Do you play?
Well, I was going to say, when you were saying that you don't play, you read science fiction, right?
And you watch the Marvel comic universe movies.
I do do that.
And I think for grownups, that's really the equivalent of the kind of, especially the kind of
pretend play and an imaginative play that you see in children. And those two things are very
parallel. There's even a nice study by Marjorie Taylor, who's studied a lot of this imaginative
play that when you talk to people who are adult writers, for example, they tell you that they
remember their imaginary friends from when they were kids. Everybody has imaginary friends. But
you know, it's sort of like they keep them in their Rolodex. They keep in touch with their
imaginary friends. And I think for adults, a lot of the function, which has always been kind of
mysterious, like why would reading about something that hasn't happened help you to understand
things that have happened? Or why would it be good in general? I think for adults, a lot of
that kind of activity is the equivalent of play. And I don't do that as much as I would like to,
or as much as I did 20 years ago, which makes me think a little about how the society has
changed. But I do think that counts as play for adults. And of course, you've got the best play
thing there could be, which is if you've got a two year old or a three year old or a four year
old, they kind of force you to be in that state, whether you start out wanting to be or not.
Yeah, there's definitely something to that. I'd have to spend a lot more time thinking about
pickle trucks now.
One of the arguments you make throughout the book is that children play a population level role.
Right, we're talking here about the way a child becomes an adult, how do they learn, you know,
how do they play in a way that keeps them from going to jail later. But you sort of say that
children are the R&D wing of our species, and that as generations turn over, we change in ways
and adapt to things in ways that the normal genetic pathway of evolution wouldn't wouldn't
necessarily predict it. And we do it partially through children. Can you talk about that? What
this sort of period of plasticity is doing at scale?
Yeah, so I think that's a good question. And we don't really completely know what the answer is.
But again, that sort of baseline is that humans have this really, really long period of
immaturity. So we have more different people who are involved and engaged in taking care of
children. And all that looks as if it's very evolutionarily costly. So there's a question
about why would it be? Now, of course, it could just be an epiphenomenon, but it seems to be a
really general pattern across so many different species at so many different times. So what kind
of function could that serve? Well, if you think about human beings, we're being faced with
unexpected environments all the time. One way you could think about it is our ecological niche is
the unknown unknowns. That's really what we're adapted to are the unknown unknowns. That's what
we're all about. And of course, once we develop culture, that just gets to be more true, because
each generation is going to change its environment in various ways that affect its culture. And that
means that now the next generation is going to have yet another new thing to try to deal with and
to understand. So I think more and more, especially in the cultural context, that having a new
generation that can look around at everything around it and say, let me try to make sense out of
this, or let me understand this, and let me think of all the new things that I could do, given this
new environment, which is the thing that children, and I think not just infants and babies, but up
through adolescence that children are doing, that could be a real advantage. And then once you've done
that kind of exploration of the space of possibilities, then as an adult, now in that
environment, you can decide which of those things you want to have happen.
Does this help explain why revolutionary political ideas are so much more appealing to sort of teens
and 20-somethings? And then why so much revolutionary political action comes from those age
groups, comes from students? It's partially this ability to exist within the imaginarium and have
a little bit more of a porous border between what exists and what could than you have when you're
50?
So we actually did some really interesting experiments where we were looking at how these kinds of
flexibility develop over the space of development. And one of the things that we discovered was that
if you look at your understanding of the physical world, the preschoolers are the most flexible and
then they get less flexible at school age and then less so at adolescence. But if you look at the
social world, there's really this burst of plasticity and flexibility in adolescence. And the
neuroscience suggests that too. So if you look at the social parts of the brain, you see this kind
of rebirth of plasticity and flexibility in adolescence. And I think that that's exactly what
you were saying, exactly what that's for, is that it gives the adolescents a chance to consider new
kinds of social possibilities and to take the information that they've got from the people
around them and say, okay, given that that's true, what something new that we could do, what's
something different from what we've done before. And if you look at the literature about cultural
evolution, I think it's true that culture is one of the really distinctive human capacities. There's
this constant tension between imitation and innovation. So to have a culture, one thing you
need to do is to have a generation that comes in and can take advantage of all the other things
that the previous generations have learned. But of course, what you also want is for that new
generation to be able to modify and tweak and change and alter the things that the previous
generation has done. And I think the period of childhood and adolescence in particular gives
you a chance to be that kind of cutting edge of change. And empirically, what you see is that
very often for things like music or clothing or culture or politics or social change, you see
that the adolescents are on the edge for better or for worse. And again, there's this kind of
trade off tension between all us cranky old people saying, what's wrong with what's wrong with
kids nowadays, because there's a reason why the previous generation is doing the things that
they're doing and the sense of here's this great range of possibilities that we haven't
considered before.
What is this somewhat deeper understanding of the child's brain imply for caregivers? What does
taking more seriously what these states of consciousness are like, say about how you should
act as a parent and uncle and aunt, a grandparent?
Well, I think here's the wrong message to take, first of all, which I think is often the message
that gets taken from this kind of information, especially in our time and our place and among
people in our culture. The wrong message is, Oh, okay, they're doing all this learning. So we
better start teaching them really, really early. We better make sure that all this learning is
going to be shaped in the way that we want it to be shaped. And we better make sure that we're
doing the right things, and we're buying the right apps, and we're reading the right books, and
we're doing the right things to shape that kind of learning in the way that we, as adults, think
that it should be shaped. And that's not the right thing. That's actually working against the
very function of this early period of exploration and learning. But I do think something
that's important is that the very mundane investment that we make as caregivers, you know,
keeping the kids alive, figuring out what it is that they want or need at any moment, those
things that are often, you know, very time consuming and require a lot of work. It's that
context of being secure and having resources and not having to worry about the immediate
circumstances that you're in, that context that caregivers provide, that's absolutely crucial.
It's absolutely essential for that broad based learning and understanding to happen. So just by
doing just by being a caregiver, just by caring, what you're doing is providing the context in
which this kind of exploration can take place. And, you know, we're pretty well designed to think
it's good to care for children in the first place. But I think especially for, you know, sort of
self-reflective parents, the fact that part of what you're doing is allowing that to happen is
really important. And then the other thing is that I think being with children in that way is a
great way for adults to get a sense of what it would be like to have that broader focus. So going
for a walk with a two year old is like going for a walk with William Blake, you know, you go to the
corner to get milk. And part of what we can even show from the neuroscience is that as adults, when
you do something really often, you become habituated, you do the same thing over and over
again, it kind of disappears from your consciousness. You're not doing it with much
experience. And again, that's a lot of the times that's a good thing, because there's other things
that we have to do. But if you do the same walk with a two year old, you realize wait a minute,
this three blocks is just amazing. It's so rich. There's dogs, and there's gates, and there's
pizza flyers, and there is plants and trees, and there's airplanes. I'm sure you've seen this with
your two year old with this this phenomenon of some plane, plane, plane. And then you suddenly
realize airplanes. Oh, wait a minute. I didn't know that there was an airplane there. But now that
you pointed out, sure enough, there is one there. So I think the other thing is that being with
children can give adults a sense of this broader way of being in the world. So I think both of you
can appreciate the fact that caring for children is is this fundamental foundational important
thing that is allowing exploration and learning to take place, rather than thinking that that's just
kind of the Scott work. And what you really need to do is go out and do explicit teaching. That's a
way of appreciating it. And I think having this kind of empathic relationship to the children who
are exploring so much is another
what should having more respect for the child's mind change, not for how we care for children, but
how we care for ourselves, or what kinds of things we open ourselves into. If I if I want to make my
mind a little bit more childlike, aside from trying to appreciate the William Blake like nature of
children, are there things of the the child's life that I should be trying to bring into mind?
Well, we know something about the sort of functions that this childlike brains her. So one thing is
being able to deal with a lot of new information. And if you think about something like traveling to a
new place, that's a good example for adults where just being someplace that you haven't been before, or
another example is just trying to learn a skill that you haven't learned before. Even if you're not
very good at it, someone once said that if something's worth doing, it's worth doing badly, just
trying to do something that's different from the things that you've done before, just that can
itself put you into a state that's more like the childlike state. And again, there's tradeoffs, because
of course, we get to be good at doing things. And then we want to do the things that we're good at, but
setting up a new place, a new technique, a new relationship to the world, that's something that
seems to help to put you in this childlike state. And to go back to the parenting point, socially,
putting people in a state where they feel as if they've got a lot of resources, and they're not under
immediate pressure to produce a particular outcome, that seems to be something that helps people to be in
this, helps even adults to be in this more playful exploratory state.
What do you think about the twin studies that people use to suggest parenting doesn't really matter? Do you
buy that evidence, or do you think it's off?
I think it's off, but I think it's often in a way that's actually kind of interesting. So what I've
argued is that if you think that what having children does is introduce more variability into the
world, right? So it actually introduces more options, more outcomes. Each of the children comes out
differently, you get this different combination of genetics and environment and temperament. And each one
of them is going to come out to be really different from anything you would expect beforehand, which is
something that I think anybody who has had more than one child is very conscious of. But if you think
that part of the function of childhood is to introduce that kind of variability into the world, and
that being a good caregiver has the effect of allowing children to come out in all these different
ways, then the basic methodology of the twin studies is to assume that if parenting has an effect, it's
going to have an effect by the child being more like the parent, and by, say, the three children that
are the children of the same parent being more like each other than, say, the twins who are adopted by
different parents. That's the kind of basic rationale behind those studies. But if you think that what being
a parent does is not make children more like themselves and more like you, but actually make them more
different from each other and different from you, then when you do a twin study, you're not going to see
that. And in fact, one of the things that I think people have been quite puzzled about in twin studies is
this idea of the non-shared environment. So it turns out that you look at genetics, and that's responsible
for some of the variants, and you look at parental environment, and that's responsible for some of it.
But a lot of it is just all this other stuff, right? And no one quite knows where all that variability is
coming from. But if you think that actually having all that variability is not a bad thing, it's a good
thing. It's what you want. It's what childhood and parenting is all about. Then having that kind of
variation that you can't really explain either by genetics or by what the parents do. That's exactly
what being a parent, being a caregiver is all about, is for.
So you just heard earlier in the conversation, they began doing a lot of work around AI. And I find the
direction coming into this from really, really interesting that there's this idea, we just create AI.
And now there's increasingly conversation over the possibility that we will need to parent AI. Tell me a
little bit about those collaborations and the angle you're taking on this.
So I've been collaborating with a whole group of people. It's been incredibly fun at the Berkeley
Artificial Intelligence Research Group. And what we've been trying to do is to try and see what would
you have to do to design an AI system that was as smart as a two-year-old, basically, right, that could
do the kinds of things that two-year-olds can do. And it's kind of striking that the very best state-of-the-art
systems that we have that are great at playing go and playing chess and maybe even driving in some
circumstances are terrible at doing the kinds of things that every two-year-old can do. And the idea is
maybe we could look at some of the things that the two-year-olds do when they're learning and see if that
makes a difference to what the AIs are doing when they're learning. So one way that I think about it
sometimes is it's sort of like, you know, if you look at the current models for AI, it's like we're giving
these AIs hyper-helicopter tiger moms. There's a programmer who's hovering over the AI and saying, oh,
yeah, yeah, you got that one right. That was a cat. That was a dog. That was another cat. That was
another dog. Or you have the AI that's saying, oh, good, your go score just went up. So do what you're
doing there. But nope, now you lost that game. So figure out something else to do. And as you might
expect, what you end up with is AI systems that are very, very good at doing the things that they were
trained to do and not very good at all at doing something different. So, you know, they can play
chess, but if you turn to a child and said, okay, we're just going to change the rules now so that
instead of the night moving this way, it moves another way, they'd be able to figure out how to adopt
what they're doing. And it's much harder for AI systems to do that. Now, one of the big problems that
we have in AI is what's come to be called the alignment problem is how can you get the AI values to
be aligned with the values of humans. So the famous example of this is the paperclip apocalypse where
you try to train the robot to make paperclips and it just goes around and turns everything in the
world, including all the humans and all houses and everything else into paperclips. If you've got this
kind of strategy of here's the goal, try to accomplish the goal as best as you possibly can, then
it's really kind of worrying about what the goal is, what the values are that you're giving these AI
systems. And one idea people have had is, well, are there ways that we can make sure that those values
are human values? But of course, one of the things that's so fascinating about humans is we keep
changing our objective functions. What counted as being the good thing the value 10 years ago might be
really different from the thing that we think is important or valuable. Now, we keep discovering that
the things that we thought were the right things to do are not the right things to do and we change what
we do as a result. And it seems as if parents are playing a really deep role in that ability. So if you
think about what it's like to be a caregiver, it involves passing on your values. That's a really
deep part of it. But it also involves allowing the next generation to take those values, look at them
in the context of the environment they find themselves in now, reshape them, rethink them, do all the
things that we were mentioning that teenagers do, consider different kinds of alternatives. And it's
having a previous generation that's willing to do both those things, that's willing to both pass on
tradition and tolerate, in fact, even encourage change that's willing to say, here's my values, but
your job is to figure out your own values. That's what lets humans keep altering their values and
goals. And most of the time for good. So the question is, if we really wanted to have AI's that
were really autonomous, and maybe we don't want to have AI's that are really autonomous. But if we
wanted to have AI's that had those kinds of capacities, they'd need to have grandmoms, they'd
need to have someone who would tell them, here's what our human values are. And here's enough
possibilities so that you could decide what your values are, and then hope that those values actually
turn out to be the right ones.
Something that strikes me about this conversation is exactly what you were touching on this idea that
you're going to have one objective function. The AI will have one goal and that will never change. You
look at any kid, right? And I think it's called social reference learning. I mean, they're constantly
doing something, then they look back at their parents to see if their parent is smiling or
frowning, then they do something else and they look back. And this constant touching back, I don't
think I appreciated what a big part of development it was until I was a parent. And I just saw how
constant it is, just all day doing something, touching back, doing something, touching back like
100 times an hour. And it seems like that would be one way to work through that alignment problem to
just assume that the learning is going to be social. It's not just going to be a goal function, it's
going to be a conversation.
AI people love acronyms, it turns out. So the acronym we have for our project is MESS, which stands
for model building exploratory social learning systems. So one piece that we think is really
important is this exploration, this ability to go out and find out things about the world, do
experiments, be curious. One of the things that we're doing right now is using some of these kind of
video game environments to put AI agents and children literally in the same environment. So the AI
is trying to work through amazing unity and the kids are working through the amazing unity. And we can
compare what it is that the kids and the AIs do in that same environment. So one thing is to get them
to explore, but another thing is to get them to do this kind of social learning. So look at a person
who's next to you and figure out what it is that they're doing. And in robotics, for example, there's a
lot of attempts to use this kind of imitative learning to train robots. But here's the catch. And the
catch is that innovation, imitation trade off that I mentioned. And in empirical work that we've done,
we've shown that when you look at kids imitating, it's really fascinating because even three year
olds will imitate the details of what someone else is doing. But they'll integrate, okay, I saw you do
this. I saw this other person do something a little different. I have some information about how this
machine works, for example, myself. And the children will put all those together to design the next
thing that would be the right thing to do. So they're constantly social referencing, they imitate
literally from the moment that they're born, they're imitating us, they're paying attention to us, they're
seeing what we do. But then they're taking that information and integrating it with all the other
information they have, say, from their own exploration, and putting that together to try to design a new
way of being to try and do something that's different from all the things that anyone has done before.
So the meta message of this conversation of what I took from your book is that learning a lot about a
child's brain actually throws a totally different light on the adult brain. As you've been learning so
much about the effort to create AI, has it made you think about the human brain differently?
Well, I have to say, actually being involved in the AI project, in many ways, makes the differences more
salient than the similarities. Because over and over again, something that is so simple, say, for young
children that we just take it for granted, like the fact that when you go into a new maze, you explore
it, that turns out to be really hard to figure out how to do with the AI system. Or to take the
example about the robot imitators, this is a really lovely project that we're working on with some
people from Google Brain. They thought, okay, well, a good way to get a robot to learn how to do
things is to imitate what a human is doing. So what they did was have humans who were, say,
manipulating a bunch of putting things on a desk in a virtual environment. And the robot is sitting
there and watching what the human does when they take up the pen and put it in the drawer in the
virtual environment. And it turned out that the problem was if you train the robot that way, then
they learn how to do exactly the same thing that the human did. But as I say, and this is always
sort of amazing to me, you know, you put the pen five centimeters to one side, and now they have no
idea what to do. But it turns out that if instead of that, what you do is you have the human just
play with the things on the desk, you tell the human, I just want you to do stuff with the things
that are here, just play with them, just do the things that you think are interesting or fun. And
then you use that to train the robots, the robots are much more resilient. So part of it is kind of
goes in circles. So it's also for the children imitating the more playful things that the adults
are doing it, or at least for robots, that's helping the robots to be more effective. I think
anyone who's worked with human brains and then goes to try to do AI, the gulf is really pretty
striking. And the difference between just the things that we take for granted that say children
are doing and the things that even the very best, most impressive AI systems can do is really
striking. Now, here's a specific thing that I'm puzzled about, that I think we've learned from
looking at the AI example. In AI, you sort of have a choice often between just doing the thing
that's the obvious thing that you've been trained to do, or just doing something that's kind of
random and noisy. Those are sort of the options. The amazing thing about kids is that they do
things that are unexpected. They're not just doing the obvious thing, but they're not just
behaving completely randomly. And I think it's a really interesting question about how do you
search through a space of possibilities, for example, where you're searching and looking
around widely enough so that you can get to something that's genuinely new, but you aren't
just doing something that's completely random and noisy. I've been thinking about the old program,
kids say the darndest things, if you just think about the things that kids say, collect them. A
lovely example that one of my computer science postdocs gave the other day, was that her three
year old was walking on the campus and saw the Campanile at Berkeley. So the Campanile is the
big clock tower at Berkeley. And he looked up at the clock tower and he said, there's a clock at
the top there. There's a clock way, way up high at the top of that tower. And then he said, I guess
they want to make sure that the children and the students don't break the clock. So they put it
really, really high up.
It's very funny.
And that's exactly the example of the sort of things that children do. It's not something he's
ever heard anybody else say. It kind of makes sense. It's not random. But of course, it's not
something that any grown up would say. In a sense, it's a really creative solution. And I think
that for AI, the challenge is how could we get a system that's capable of doing something that's
really new, which is what you want if you want robustness and resilience, and isn't just random
that is new, but appropriately new.
I always wonder if the AI two year old, three year old comparisons are just a category error
there in the sense that, you know, you might say, a small bat can do something that no children can
do, which is it can fly. GPT three, the open AI program can do something that no children can do
effortlessly, which is mimic, you know, the text of a certain kind of author. Is it just going to be
the case that there are certain collaborations of our physical forms and molecular structures and so
on that give our intelligence different categories? I always wonder if there's almost like a kind of
comfort being taken at how hard it is to do two year old style things. And meanwhile, I don't want
to put too much weight on it's beating everybody at go, but that what it does seem plausible, it
could do in 10 years will be will be quite remarkable. Now, again, that's different than the
conscious agent, right, that has to make its way through the world on its own. I'm curious how much
weight you put on the idea that that might just be the wrong comparison.
This is the old point about, you know, asking whether an AI can think is like asking whether a
submarine can swim, right? It feels like it's just a category. It's just a category error. And of
course, as I say, we have two year olds around a lot. So we don't really need more two year olds. We
should be designing these systems so they're complementary to our intelligence rather than
somehow being a reproduction of our intelligence. But on the other hand, there are very, again,
you know, just take something really simple, like it would be really good to have robots that
could pick things up and put them in boxes, right? Like that doesn't seem like such a high
falutin skill to be able to have and that could pick things up and put them in boxes. And now
when you gave it a screw that looked a little different from the previous screw and a box that
looked a little different from the previous box, that they could figure out, oh, yeah, no, that
one's a screw and it goes in the screw box, not the other box. And it turns out that even to do
just these really, really simple things that we would really like to have artificial systems do,
it's really hard. And those are things that two year olds do really well. And we can think about
what is it. On the other hand, you know, the two year olds don't get bored knowing how to put
things in boxes. So what is it that they've got? What mechanisms do they have that could help us
with some of these kinds of problems? And another example that we've been working on a lot with
the Bay Area group is just vision. So just look at a screen with a lot of pixels and make sense out
of it. And as you probably know, if you look at something like ImageNet, you can show, say, a
deep learning system, a whole lot of pictures of cats and dogs on the web. And eventually you'll
get it so that it can most of the time say this is a cat and this is a dog. But then you can give
it something that is just obviously not a cat or a dog. And they'll make a mistake. And they won't
be able to generalize even to, say, a dog on a video that's actually moving. So even if you take
something as simple as that, you would like to have your systems actually, you'd like to have a
computer in your car actually be able to identify this is a pedestrian or a car. It turns out that
even those simple things involve abilities that we see in very young children that are actually quite
hard to program into a computer. Some of the things that we're looking at, for instance, is with
children, when they're learning to identify objects in the world, one thing they do is they pick them
up and then they move around, look at them from different angles, look at them from the top, look
at them from the bottom, look at your hands this way, look at your hands that way, walk around to
the other side, pick things up and get into everything and make a terrible mess because you're
picking them up and throwing them around. But it turns out that maybe just the kind of thing that
you need to do, not to do anything fancy, just to have vision, just to be able to see the objects in
the way that adults see the objects.
I think it's a good place to come to a close. So let's get a variation on what's our final question.
What are three children's books you love and would recommend to the audience?
Yeah, so I was thinking a lot about this. And I actually had converged on two children's books. And
then yesterday, I went to see my grandchildren for the first time in a year, my beloved grandchildren.
And I was really pleased because my intuitions about the best books were completely confirmed by
this great reunion with the grandchildren. So my five year old grandson, who hasn't been in our
house for a year, first said, I love you, Grandma. And then said, you know, Grandma, you still have
that book, you know, that book that you have at your house with a little boy who has like the white
suit, and he goes to the island with the monsters on it. And then he comes back again. And I said, you
mean where the wild things are? That's the one with the wild things with the most you still have that
book. Could we read that book at your house? So I figured that's a pretty serious endorsement when a
five year old remembers something from a year ago. So that's the first one, especially for the
younger children, all the Morissette books, but especially where the wild things are is a fantastic
wonderful book. And then for older children, that same day, my nine year old, who is very into the
Marvel Universe and superheroes said, could we read a chapter from Mary Poppins, which is again
something that grandma reads. And we had a marvelous time reading Mary Poppins. And he said, you
know, the book is so much better than the movie. And he was absolutely right. And the reason is that
when you actually read the Mary Poppins books, especially the later ones, like Mary Poppins in
the park, and Mary Poppins opens the door, Mary Poppins is a much stranger, weirder, darker
figure than Julie Andrews is. So if you've seen the movies, you have no idea what Mary Poppins is
about. Essentially, what Mary Poppins is about is this very strange surreal set of adventures that
the children are having with this figure who, as I said to Augie, is much more like Iron Man or
Batman or Doctor Strange than like Julie Andrews, right? You know, he's this powerful, mysterious,
sometimes dark, but ultimately good creature in your experience. So I keep thinking, oh, yeah, now
what we really need to do is add Mary Poppins to the Marvel Universe. And that would be that would
be a much better version. And let me give you a third book, which is much more obscure. There's a
book called The Children of Green Know, K-N-O-W-E. I like this because it's a book about a
grandmother and her grandson, and he comes to visit her in this strange old house in the
Cambridge countryside. And gradually it gets to be clear that there are ghosts of the history of
this house. And what I like about all three of these books, in their different ways, is that I
think they capture this thing that's so distinctive about childhood, the fact that on the one hand,
you're in this safe place. So with the wild things, you know, he's in his room where mom is,
where supper is going to be. And all the time sitting in that room, he also adventures out in
this boat to these strange places where wild things are, including he himself as a wild thing. And
the same thing is true with Mary Poppins. So there are these children who are just leading this
very ordinary British middle-class life in the 30s, and they're going to the greengrass or in the
fishmonger. And yet there's all this strangeness, this weirdness, this surreal things, just about
those everyday experiences. And the same way with The Children of Green Know, you're going to visit
your grandmother in her house in the country. And then it turns out that that house is full of
spirits and ghosts and traditions and things that you've learned from the past. All three of those
books really capture what's special about childhood. It's that combination of a small, safe
world. And it's actually having that small, safe world that lets you explore a much wilder,
crazier, stranger set of worlds than any grown-up ever gets to.
Alison Gopnik, thank you very much.
Thank you, Esser.
Thank you to Alison Gopnik for being here. I'm going to keep it up with these little occasional
recommendations after the show. I'm a writing nerd. I mean, obviously I'm a writer, but I like
writing software. When I went to box media, partially I did that because of their great CMS or
publishing software, Chorus. And I'm always looking for really good clean composition apps. I
find Word and Pages and Google Docs to be just horrible to write in and having a good space to
write in. And it actually helps me think. But I found something recently that I like and I'm not
getting paid to promote them or anything. I just like it. It's called Calmly Writer. You can just
find it at CalmlyWriter.com. And it's the cleanest writing interface, simplest of these programs I
found. So if you're looking for a real lightweight, easy place to do some writing, Calmly Writer, but
I'd be interested to hear what you all like because I've become a little bit of a nerd about these
ads. That's it for the show. Thank you for listening as always. If you want to help the show out,
leave us a review wherever you are listening to it now, or send this episode to a friend, a family
member, somebody you want to talk about it with. It really does help the show grow.
The Is Your Clancho is a production of New York Times Opinion. It is produced by Roger Karma and
Jeff Geld fact checked by Michelle Harris, original music by Isaac Jones and mixing by Jeff Geld.
Machine-generated transcript that may contain inaccuracies.
Here’s a sobering thought: The older we get, the harder it is for us to learn, to question, to reimagine. This isn’t just habit hardening into dogma. It’s encoded into the way our brains change as we age. And it’s worsened by an intellectual and economic culture that prizes efficiency and dismisses play.
Alison Gopnik is a professor of psychology and philosophy at the University of California, Berkeley, where she runs the Cognitive Development and Learning Lab; she’s also the author of over 100 papers and half a dozen books, including “The Gardener and the Carpenter” and “The Philosophical Baby.” What I love about her work is she takes the minds of children seriously. The child’s mind is tuned to learn. They are, she writes, the R. & D. departments of the human race. But a mind tuned to learn works differently from a mind trying to exploit what it already knows.
So instead of asking what children can learn from us, perhaps we need to reverse the question: What can we learn from them?
In this conversation, recorded in April 2021, Gopnik and I discuss the way children think, the cognitive reasons social change so often starts with the young, and the power of play. We talk about why Gopnik thinks children should be considered an entirely different form of Homo sapiens, the crucial difference between “spotlight” consciousness and “lantern” consciousness, why “going for a walk with a 2-year-old is like going for a walk with William Blake,” what A.I. researchers are borrowing from human children, the effects of different types of meditation on the brain and more.
Book recommendations:
Where the Wild Things Are by Maurice Sendak
Mary Poppins in the Park by P.L. Travers
The Children of Green Knowe by L.M. Boston
Thoughts? Email us at ezrakleinshow@nytimes.com. (And if you're reaching out to recommend a guest, please write “Guest Suggestion" in the subject line.)
You can find transcripts (posted midday) and more episodes of “The Ezra Klein Show” at nytimes.com/ezra-klein-podcast, and you can find Ezra on Twitter @ezraklein. Book recommendations from all our guests are listed at https://www.nytimes.com/article/ezra-klein-show-book-recs.
“The Ezra Klein Show” is produced by Emefa Agawu, Annie Galvin, Jeff Geld and Roge Karma. Fact-checking by Michelle Harris, Mary Marge Locker and Kate Sinclair. Original music by Isaac Jones. Mixing by Jeff Geld. Audience strategy by Shannon Busta. Special thanks to Kristina Samulewski.