0:00:00.835,0:00:02.990
Mark Twain summed up[br]what I take to be

0:00:02.990,0:00:06.110
one of the fundamental problems[br]of cognitive science

0:00:06.110,0:00:07.820
with a single witticism.

0:00:08.410,0:00:11.492
He said, "There's something[br]fascinating about science.

0:00:11.492,0:00:14.720
One gets such wholesale[br]returns of conjecture

0:00:14.720,0:00:17.924
out of such a trifling[br]investment in fact."

0:00:17.924,0:00:19.509
(Laughter)

0:00:20.199,0:00:22.803
Twain meant it as a joke,[br]of course, but he's right:

0:00:22.803,0:00:25.679
There's something[br]fascinating about science.

0:00:25.679,0:00:29.940
From a few bones, we infer[br]the existence of dinosuars.

0:00:30.910,0:00:34.781
From spectral lines,[br]the composition of nebulae.

0:00:35.471,0:00:38.409
From fruit flies,

0:00:38.409,0:00:41.352
the mechanisms of heredity,

0:00:41.352,0:00:45.601
and from reconstructed images[br]of blood flowing through the brain,

0:00:45.601,0:00:49.949
or in my case, from the behavior[br]of very young children,

0:00:50.519,0:00:53.138
we try to say something about[br]the fundamental mechanisms

0:00:53.138,0:00:55.716
of human cognition.

0:00:55.716,0:01:00.475
In particular, in my lab in the Department[br]of Brain and Cognitive Sciences at MIT,

0:01:00.475,0:01:04.129
I have spent the past decade[br]trying to understand the mystery

0:01:04.129,0:01:08.106
of how children learn so much[br]from so little so quickly.

0:01:08.666,0:01:11.644
Because, it turns out that[br]the fascinating thing about science

0:01:11.644,0:01:15.173
is also a fascinating[br]thing about children,

0:01:15.173,0:01:17.754
which, to put a gentler spin on Mark Twain

0:01:17.754,0:01:22.404
is precisely their ability[br]to draw rich, abstract inferences

0:01:22.404,0:01:27.065
rapidly and accurately[br]from sparse, noisy data.

0:01:28.355,0:01:30.753
I'm going to give you[br]just two examples today.

0:01:30.753,0:01:33.040
One is about a problem of generalization,

0:01:33.040,0:01:35.890
and the other is about a problem[br]of causal reasoning.

0:01:35.890,0:01:38.415
And although I'm going to talk[br]about work in my lab,

0:01:38.415,0:01:41.875
this work is inspired by[br]and indebted to a field.

0:01:41.875,0:01:46.158
I'm grateful to mentors, colleagues,[br]and collaborators around the world.

0:01:47.308,0:01:50.282
Let me start with the problem[br]of generalization.

0:01:50.652,0:01:54.785
Generalizing from small samples of data[br]is the bread and butter of science.

0:01:54.785,0:01:57.339
We poll a tiny fraction of the electorate

0:01:57.339,0:02:00.550
and we predict the outcome[br]of national elections.

0:02:00.550,0:02:04.165
We see how a handful of patients[br]respond to treatment in a clinical trial,

0:02:04.165,0:02:07.230
and we bring drugs to a national market.

0:02:07.230,0:02:11.595
But this only works if our sample[br]is randomly drawn from the population.

0:02:11.595,0:02:14.330
If our sample is cherry-picked in some way

0:02:14.330,0:02:16.402
-- say, we poll only urban voters,

0:02:16.402,0:02:20.790
or say, in our clinical trials[br]for treatments for heart disease,

0:02:20.790,0:02:22.671
we include only men --

0:02:22.671,0:02:25.829
the results may not generalize[br]to the broader population.

0:02:26.479,0:02:30.060
So scientists care whether evidence[br]is randomly sampled or not,

0:02:30.060,0:02:32.075
but what does that have to do with babies?

0:02:32.585,0:02:37.206
Well, babies have to generalize[br]from small samples of data all the time.

0:02:37.206,0:02:40.364
They see a few rubber ducks[br]and learn that they float,

0:02:40.364,0:02:43.939
or a few balls and learn that they bounce.

0:02:43.939,0:02:46.890
And they develop expectations[br]about ducks and balls

0:02:46.890,0:02:49.606
that they're going to extend[br]to rubber ducks and balls

0:02:49.606,0:02:51.485
for the rest of their lives.

0:02:51.485,0:02:55.224
And the kinds of generalizations[br]babies have to make about ducks and balls

0:02:55.224,0:02:57.313
they have to make about almost everything:

0:02:57.313,0:03:01.230
shoes and ships and ceiling wax[br]and cabbages and [kings?].

0:03:02.200,0:03:05.161
So do babies care whether[br]the tiny bit of evidence they see

0:03:05.161,0:03:08.853
is plausibly representative[br]of a larger population?

0:03:09.763,0:03:11.663
Let's find out.

0:03:11.663,0:03:13.386
I'm going to show you two movies,

0:03:13.386,0:03:15.848
one from each of two conditions[br]of an experiment,

0:03:15.848,0:03:18.286
and because you're going to see[br]just two movies,

0:03:18.286,0:03:20.422
you're going to see just two babies,

0:03:20.422,0:03:24.369
and any two babies differ from each other[br]in innumerable ways.

0:03:24.369,0:03:27.420
But these babies, of course,[br]here stand in for groups of babies,

0:03:27.420,0:03:29.315
and the differences you're going to see

0:03:29.315,0:03:34.510
represent average group differences[br]in babies' behavior across conditions.

0:03:35.160,0:03:37.743
In each movie, you're going to see[br]a baby doing maybe

0:03:37.743,0:03:41.203
just exactly what you might[br]expect a baby to do,

0:03:41.203,0:03:45.220
and we can hardly make babies[br]more magical than they already are.

0:03:46.090,0:03:48.100
But to my mind the magical thing,

0:03:48.100,0:03:50.189
and what I want you to pay attention to,

0:03:50.189,0:03:53.300
is the contrast between[br]these two conditions,

0:03:53.300,0:03:56.829
because the only thing[br]that differs between these two movies

0:03:56.829,0:04:01.425
is the statistical evidence[br]the babies are going to observe.

0:04:01.425,0:04:04.608
We're going to show babies[br]a box of blue and yellow balls,

0:04:04.608,0:04:09.228
and my then-graduate student,[br]now colleague at Stanford, Yowan Gwan,

0:04:09.228,0:04:12.305
is going to pull three blue balls[br]in a row out of this box,

0:04:12.305,0:04:15.428
and when she pulls those balls out,[br]she's going to squeeze them,

0:04:15.428,0:04:17.541
and the balls are going to squeak.

0:04:17.541,0:04:20.304
And if you're a baby,[br]that's like a TEDTalk.

0:04:20.304,0:04:22.208
It doesn't get better than that.

0:04:22.208,0:04:24.769
(Laughter)

0:04:26.968,0:04:30.627
But the important point is it's really[br]easy to pull three blue balls in a row

0:04:30.627,0:04:32.932
out of a box of mostly blue balls.

0:04:32.932,0:04:34.992
You could do that with your eyes closed.

0:04:34.992,0:04:37.988
It's plausibly a random sample[br]from this population.

0:04:37.988,0:04:41.720
And if you can reach into a box at random[br]and pull out things that squeak,

0:04:41.720,0:04:44.559
then maybe everything in the box squeaks.

0:04:44.559,0:04:48.209
So maybe babies should expect[br]those yellow balls to squeak as well.

0:04:48.209,0:04:50.728
Now, those yellow balls[br]have funny sticks on the end,

0:04:50.728,0:04:53.585
so babies could do other things[br]with them if they wanted to.

0:04:53.585,0:04:55.416
They could pound them or whack them.

0:04:55.416,0:04:58.002
But let's see what the baby does.

0:05:00.548,0:05:03.891
(Video) Yowan Gwan: See this?[br](Ball squeaks)

0:05:04.531,0:05:07.576
Did you see that?[br](Ball squeaks)

0:05:08.036,0:05:11.102
Cool.

0:05:12.706,0:05:14.656
See this one?

0:05:14.656,0:05:16.537
(Ball squeaks)

0:05:16.537,0:05:19.190
Wow.

0:05:21.854,0:05:23.967
Laura Schulz: Told you. (Laughter)

0:05:23.967,0:05:27.998
(Video) YG: See this one?[br](Ball squeaks)

0:05:27.998,0:05:32.617
Hey Clara, this one's for you.[br]You can go ahead and play.

0:05:39.854,0:05:44.219
(Laughter)

0:05:44.219,0:05:47.214
LS: I don't even have to talk, right?

0:05:47.214,0:05:50.113
All right, it's nice that babies[br]will generalize properties

0:05:50.113,0:05:51.641
of blue balls to yellow balls,

0:05:51.641,0:05:54.737
and it's impressive that babies[br]can learn from imitating us,

0:05:54.737,0:05:58.406
but we've known those things about babies[br]for a very long time.

0:05:58.406,0:06:00.217
The really interesting question

0:06:00.217,0:06:03.069
is what happens when we show babies[br]exactly the same thing,

0:06:03.069,0:06:06.680
and we can ensure it's exactly the same[br]because we have a secret compartment

0:06:06.680,0:06:08.790
and we actually pull the balls from there,

0:06:08.790,0:06:12.268
but this time, all we change[br]is the apparent population

0:06:12.268,0:06:15.170
from which that evidence was drawn.

0:06:15.780,0:06:18.723
This time, we're going to show babies[br]three blue balls

0:06:18.723,0:06:21.857
pulled out of a box[br]of mostly yellow balls,

0:06:22.357,0:06:23.599
and guess what?

0:06:23.599,0:06:26.269
You cannot randomly draw[br]three blue balls in a row

0:06:26.269,0:06:28.753
out of a box of mostly yellow balls.

0:06:28.753,0:06:32.500
That is not plausibly[br]randomly sampled evidence.

0:06:32.500,0:06:37.623
That evidence suggests that maybe Yowan[br]was deliberately sampling the blue balls.

0:06:37.623,0:06:40.206
Maybe there's something special[br]about the blue balls.

0:06:40.846,0:06:43.822
Maybe only the blue balls squeak.

0:06:43.822,0:06:45.717
Let's see what the baby does.

0:06:45.717,0:06:48.621
(Video) YG: See this?[br](Ball squeaks)

0:06:50.851,0:06:53.496
See this toy?[br](Ball squeaks)

0:06:53.496,0:06:58.976
Oh, that was cool. See?[br](Ball squeaks)

0:06:58.976,0:07:03.370
Now this one's for you to play.[br]You can go ahead and play.

0:07:11.264,0:07:13.911
(Baby cries)

0:07:14.901,0:07:17.649
LS: So you just saw[br]two 15-month old babies

0:07:17.649,0:07:19.591
do entirely different things

0:07:19.591,0:07:23.190
based only on the probability[br]of the sample they observed.

0:07:23.190,0:07:25.511
Let me show you the experimental results.

0:07:25.511,0:07:28.275
On the vertical axis, you'll see[br]the percentage of babies

0:07:28.275,0:07:30.805
who squeezed the ball in each condition,

0:07:30.805,0:07:34.520
and as you'll see, babies are much[br]more likely to generalize the evidence

0:07:34.520,0:07:37.655
when it's plausibly representative[br]of the population

0:07:37.655,0:07:41.393
than when the evidence[br]is clearly cherry-picked.

0:07:41.393,0:07:43.808
And this leads to a fun prediction:

0:07:43.808,0:07:48.676
suppose you pulled just one blue ball[br]out of the mostly yellow box.

0:07:48.896,0:07:52.765
Now, you just can't pull three blue balls[br]in a row at random out of a yellow box,

0:07:52.765,0:07:55.220
but you could randomly sample[br]just one blue ball.

0:07:55.220,0:07:57.190
That's not an improbable sample.

0:07:57.190,0:07:59.414
And if you could reach into[br]a box at random

0:07:59.414,0:08:03.401
and pull out something that squeaks,[br]maybe everything in the box squeaks.

0:08:03.875,0:08:08.320
So even though babies are going to see[br]much less evidence for squeaking,

0:08:08.320,0:08:10.562
and have many fewer actions to imitate

0:08:10.562,0:08:13.905
in this one ball condition than in[br]the condition you just saw,

0:08:13.905,0:08:17.797
we predicted that babies themselves[br]would squeeze more,

0:08:17.797,0:08:20.691
and that's exactly what we found.

0:08:20.691,0:08:25.102
So 15-month old babies,[br]in this respect, like scientists,

0:08:25.102,0:08:28.190
care whether evidence[br]is randomly sampled or not,

0:08:28.190,0:08:31.697
and they use this to develop[br]expectations about the world:

0:08:31.697,0:08:33.879
what squeaks and what doesn't,

0:08:33.879,0:08:37.024
what to explore and what to ignore.

0:08:38.384,0:08:40.450
Let me show you another example now,

0:08:40.450,0:08:43.180
this time about a problem[br]of causal reasoning.

0:08:43.180,0:08:45.619
And it starts with a problem[br]of confounded evidence

0:08:45.619,0:08:47.291
that all of us have,

0:08:47.291,0:08:49.311
which is that we are part of the world.

0:08:49.311,0:08:52.747
And this may not seem like a problem[br]to you, but like most problems,

0:08:52.747,0:08:55.084
it's only a problem when things go wrong.

0:08:55.464,0:08:57.275
Take this baby, for instance.

0:08:57.275,0:08:58.980
Things are going wrong for him.

0:08:58.980,0:09:01.251
He would like to make[br]this toy go, and he can't.

0:09:01.251,0:09:03.780
I'll show you a few second clip.

0:09:09.340,0:09:11.260
And there's two possibilities, broadly:

0:09:11.260,0:09:13.894
maybe he's doing something wrong,

0:09:13.894,0:09:18.110
or maybe there's something[br]wrong with the toy.

0:09:18.110,0:09:20.221
So in this next experiment,

0:09:20.221,0:09:23.518
we're going to give babies[br]just a tiny bit of statistical data

0:09:23.518,0:09:26.100
supporting one hypothesis over the other,

0:09:26.100,0:09:29.555
and we're going to see if babies[br]can use that to make different decisions

0:09:29.555,0:09:31.389
about what to do.

0:09:31.389,0:09:33.411
Here's the setup.

0:09:34.071,0:09:37.101
Yowan is going to try to make[br]the toy go and succeed.

0:09:37.101,0:09:40.421
I am then going to try twice[br]and fail both times,

0:09:40.421,0:09:43.533
and then Yowan is going[br]to try again and succeed,

0:09:43.533,0:09:46.705
and this roughly sums up my relationship[br]to my graduate students

0:09:46.705,0:09:49.540
in technology across the board.

0:09:50.030,0:09:53.322
But the important point here is[br]it provides a little bit of evidence

0:09:53.322,0:09:57.010
that the problem isn't with the toy,[br]it's with the person.

0:09:57.610,0:09:59.340
Some people can make this toy go,

0:09:59.340,0:10:00.299
and some can't.

0:10:00.799,0:10:04.212
Now, when the baby gets the toy,[br]he's going to have a choice.

0:10:04.212,0:10:06.400
His mom is right there,

0:10:06.400,0:10:09.715
so he can go ahead and hand off the toy[br]and change the person,

0:10:09.715,0:10:12.873
but there's also going to be[br]another toy at the end of that cloth,

0:10:12.873,0:10:16.425
and he can pull the cloth towards him[br]and change the toy.

0:10:16.425,0:10:18.515
So let's see what the baby does.

0:10:18.515,0:10:21.278
(Video) YG: Two, three. Go!

0:10:21.278,0:10:25.829
(Video) LS: One, two, three, go!

0:10:25.829,0:10:33.211
Arthur, I'm going to try again.[br]One, two, three, go!

0:10:33.677,0:10:36.277
(Video) YG: Arthur,[br]let me try again, okay?

0:10:36.277,0:10:40.827
One, two, three, go!

0:10:41.583,0:10:43.466
Look at that. Remember these toys?

0:10:43.466,0:10:46.730
See these toys? Yeah, I'm going[br]to put this one over here,

0:10:46.730,0:10:48.792
and I'm going to give this one to you.

0:10:48.792,0:10:51.127
You can go ahead and play.

0:11:11.213,0:11:15.950
LS: Okay, Laura, but of course,[br]babies love their mommies.

0:11:15.950,0:11:18.132
Of course babies give toys[br]to their mommies

0:11:18.132,0:11:20.162
when they can't make them work.

0:11:20.162,0:11:23.755
So again, the really important question[br]is what happens when we change

0:11:23.755,0:11:26.909
the statistical data ever so slightly.

0:11:26.909,0:11:30.996
This time, babies are going to see the toy[br]work and fail in exactly the same order,

0:11:30.996,0:11:33.411
but we're changing[br]the distribution of evidence.

0:11:33.411,0:11:37.822
This time, Yowan is going to succeed[br]once and fail once, and so am I.

0:11:37.822,0:11:43.459
And this suggests it doesn't matter[br]who tries this toy, the toy is broken.

0:11:43.459,0:11:45.345
It doesn't work all the time.

0:11:45.345,0:11:47.310
Again, the baby's going to have a choice.

0:11:47.310,0:11:50.706
Her mom is right next to her,[br]so she can change the person,

0:11:50.706,0:11:53.910
and there's going to be another toy[br]at the end of the cloth.

0:11:53.910,0:11:55.288
Let's watch what she does.

0:11:55.288,0:11:59.636
(Video) YG: Two, three, go!

0:11:59.636,0:12:04.620
Let me try one more time.[br]One, two, three, go!

0:12:05.460,0:12:07.157
Hmm.

0:12:08.870,0:12:10.642
(Video) LS: Let me try, Clara.

0:12:10.642,0:12:14.587
One, two, three, go!

0:12:15.935,0:12:17.770
Hmm, let me try again.

0:12:17.770,0:12:22.870
One, two, three, go!

0:12:23.009,0:12:25.242
(Video) YG: I'm going[br]to put this one over here,

0:12:25.242,0:12:27.243
and I'm going to give this one to you.

0:12:27.243,0:12:30.840
You can go ahead and play.

0:12:46.376,0:12:51.273
(Applause)

0:12:52.993,0:12:55.385
LS: Let me show you[br]the experimental results.

0:12:55.385,0:12:57.860
On the vertical axis,[br]you'll see the distribution

0:12:57.860,0:13:00.437
of children's choices in each condition,

0:13:00.437,0:13:04.988
and you'll see that the distribution[br]of the choices children make

0:13:04.988,0:13:07.775
depends on the evidence they observe.

0:13:07.775,0:13:09.632
So in the second year of life,

0:13:09.632,0:13:12.209
babies can use a tiny bit[br]of statistical data

0:13:12.209,0:13:15.576
to decide between two[br]fundamentally different strategies

0:13:15.576,0:13:17.457
for acting on the world:

0:13:17.457,0:13:20.200
asking for help and exploring.

0:13:21.700,0:13:25.134
I've just shown you[br]two laboratory experiments

0:13:25.134,0:13:28.825
out of literally hundreds in the field[br]that make similar points,

0:13:28.825,0:13:31.217
because the really critical point

0:13:31.217,0:13:36.325
is that children's ability[br]to make rich inferences from sparse data

0:13:36.325,0:13:41.666
underlies all the species-specific[br]cultural learning that we do.

0:13:41.666,0:13:46.263
Children learn about new tools[br]from just a few examples.

0:13:46.263,0:13:50.980
They learn new causal relationships[br]from just a few examples.

0:13:51.928,0:13:56.799
They even learn new words,[br]in this case in American sign language.

0:13:56.799,0:13:59.110
I want to close with just two points.

0:14:00.050,0:14:03.738
If you've been following my world,[br]the field of brain and cognitive sciences,

0:14:03.738,0:14:05.665
for the past few years,

0:14:05.665,0:14:08.080
three big ideas will have come[br]to your attention.

0:14:08.570,0:14:11.516
The first is that this is[br]the era of the brain.

0:14:11.516,0:14:15.185
And indeed, there have been[br]staggering discoveries in neuroscience:

0:14:15.185,0:14:18.621
localizing, functionally specialized[br]regions of cortex,

0:14:18.621,0:14:21.222
turning mouse brains transparent,

0:14:21.222,0:14:24.998
activating neurons with light.

0:14:24.998,0:14:26.994
A second big idea

0:14:26.994,0:14:31.098
is that this is the era of big data[br]and machine learning,

0:14:31.098,0:14:34.239
and machine learning promises[br]to revolutionize our understanding

0:14:34.239,0:14:38.906
of everything from social networks[br]to epidemiology.

0:14:38.906,0:14:41.599
And maybe, as it tackles problems[br]of seen understanding

0:14:41.599,0:14:43.592
and natural language processing

0:14:43.592,0:14:46.916
to tell us something[br]about human cognition.

0:14:47.756,0:14:49.693
And the final big idea you'll have heard

0:14:49.693,0:14:53.080
is that maybe it's a good idea we're going[br]to know so much about brains

0:14:53.080,0:14:54.997
and have so much access to big data,

0:14:54.997,0:14:57.504
because left to our own devices,

0:14:57.504,0:15:01.335
humans are fallible, we take shortcuts,

0:15:01.335,0:15:04.772
we err, we make mistakes,

0:15:04.772,0:15:08.246
we're biased, and in innumerable ways,

0:15:09.326,0:15:11.425
we get the world wrong.

0:15:12.843,0:15:15.792
I think these are all important stories,

0:15:15.792,0:15:19.577
and they have a lot to tell us[br]about what it means to be human,

0:15:19.577,0:15:23.106
but I want you to note that today[br]I told you a very different story.

0:15:23.966,0:15:27.773
It's a story about minds and not brains,

0:15:27.773,0:15:30.779
and in particular, it's a story[br]about the kinds of computations

0:15:30.779,0:15:33.369
that uniquely human minds can perform,

0:15:33.369,0:15:37.313
which involve rich, structured knowledge[br]and the ability to learn

0:15:37.313,0:15:42.581
from small amounts of data,[br]the evidence of just a few examples.

0:15:44.301,0:15:48.600
And fundamentally, it's a story[br]about how starting as very small children

0:15:48.600,0:15:52.780
and continuing out all the way[br]to the greatest accomplishments

0:15:52.780,0:15:55.223
of our culture,

0:15:56.623,0:15:58.620
we get the world right.

0:16:00.433,0:16:05.700
Folks, human minds do not only learn[br]from small amounts of data.

0:16:06.285,0:16:08.386
Human minds think[br]of altogether new ideas.

0:16:08.746,0:16:11.437
Human minds generate[br]research and discovery,

0:16:11.787,0:16:16.320
and human minds generate[br]art and literature and poetry and theater,

0:16:17.070,0:16:20.600
and human minds take care of other humans:

0:16:21.330,0:16:24.257
our old, our young, our sick.

0:16:24.517,0:16:26.884
We even heal them.

0:16:27.624,0:16:30.667
In the years to come, we're going[br]to see technological innovation

0:16:30.667,0:16:33.284
beyond anything I can even envision,

0:16:34.594,0:16:36.614
but we are very unlikely

0:16:36.614,0:16:42.323
to see anything even approximating[br]the computational power of a human child

0:16:42.323,0:16:46.621
in my lifetime or in yours.

0:16:46.621,0:16:51.668
If we invest in these most powerful[br]learners and their development,

0:16:51.668,0:16:54.585
in babies and children

0:16:54.585,0:16:56.411
and mothers and fathers

0:16:56.411,0:16:59.110
and caregivers and teachers

0:16:59.110,0:17:03.280
the ways we invest in our other[br]most powerful and elegant forms

0:17:03.280,0:17:05.848
of technology, engineering, and design,

0:17:06.488,0:17:09.437
we will not just be dreaming[br]of a better future,

0:17:09.437,0:17:11.564
we will be planning for one.

0:17:11.564,0:17:13.909
Thank you very much.

0:17:13.909,0:17:17.810
(Applause)

0:17:17.810,0:17:22.236
Chris Anderson: Laura, thank you.[br]I do actually have a question for you.

0:17:22.236,0:17:24.595
First of all, the research is insane.

0:17:24.595,0:17:28.320
I mean, who would design[br]an experiment like that? (Laughter)

0:17:29.150,0:17:30.940
I've seen that a couple of times,

0:17:30.940,0:17:34.162
and I still don't honestly believe[br]that that can truly be happening,

0:17:34.162,0:17:37.320
but other people have done[br]similar experiments. It checks out.

0:17:37.320,0:17:38.953
They really are that genius.

0:17:38.953,0:17:41.960
LS: You know, they look really impressive[br]in our experiments,

0:17:41.960,0:17:44.592
but think about what they[br]look like in real life, right?

0:17:44.592,0:17:45.762
It starts out as a baby.

0:17:45.762,0:17:47.769
Eighteen months later,[br]it's talking to you,

0:17:47.769,0:17:50.810
and babies' first words aren't just[br]things like balls and ducks,

0:17:50.810,0:17:53.691
they're things like "all gone,"[br]which refer to disappearance,

0:17:53.691,0:17:55.974
or "uh oh," which refer[br]to unintentional actions.

0:17:55.974,0:17:57.536
It has to be that powerful.

0:17:57.536,0:18:00.311
It has to be much more powerful[br]than anything I showed you.

0:18:00.311,0:18:02.285
They're figuring out the entire world.

0:18:02.285,0:18:05.429
A four-year old can talk to you[br]about almost anything.

0:18:05.429,0:18:07.030
(Applause)

0:18:07.030,0:18:10.444
CA: And if I understand you right,[br]the other key point you're making is,

0:18:10.444,0:18:13.198
we've been through these years[br]where there's all this talk

0:18:13.198,0:18:15.130
of how quirky and buggy our minds are,

0:18:15.130,0:18:17.997
that behavioral economics[br]and the whole theories behind that

0:18:17.997,0:18:19.600
that we're not rational agents.

0:18:19.600,0:18:23.816
You're really saying that the biggest[br]story is how extraordinary,

0:18:23.816,0:18:28.760
and there really is genius there[br]that is underappreciated.

0:18:28.760,0:18:30.830
LS: One of my favorite[br]quotes in psychology

0:18:30.830,0:18:33.120
comes from the social[br]psychologist Solomon Asch,

0:18:33.120,0:18:35.927
and he said the fundamental task[br]of psychology is to remove

0:18:35.927,0:18:38.553
the veil of self-evidence from things.

0:18:38.553,0:18:43.104
There are a million orders of magnitude[br]more decisions you make every day

0:18:43.104,0:18:44.451
that get the world right.

0:18:44.451,0:18:46.583
You know about objects[br]and their properties.

0:18:46.583,0:18:49.612
You know them when they're occluded.[br]You know them in the dark.

0:18:49.612,0:18:50.920
You can walk through rooms.

0:18:50.920,0:18:54.452
You can figure out what other people[br]are thinking. You can talk to them.

0:18:54.452,0:18:56.682
You can navigate space.[br]You know about numbers.

0:18:56.682,0:18:59.704
You know causal relationships.[br]You know about moral reasoning.

0:18:59.704,0:19:02.060
You do this effortlessly,[br]so we don't see it,

0:19:02.060,0:19:04.972
but that is how we get the world right,[br]and it's a remarkable

0:19:04.972,0:19:07.290
and very difficult-to-understand[br]accomplishment.

0:19:07.290,0:19:09.918
CA: I suspect there are people[br]in the audience who have

0:19:09.918,0:19:12.156
this view of accelerating[br]technological power

0:19:12.156,0:19:15.114
who might dispute your statement[br]that never in our lifetimes

0:19:15.114,0:19:17.732
will a computer do what[br]a three-year old child can do,

0:19:17.732,0:19:20.980
but what's clear is that in any scenario,

0:19:20.980,0:19:24.750
our machines have so much to learn[br]from our toddlers.

0:19:26.230,0:19:29.446
LS: I think so. You'll have some[br]machine learning folks up here.

0:19:29.446,0:19:33.649
I mean, you should never bet[br]against babies or chimpanzees

0:19:33.649,0:19:37.294
or technology as a matter of practice,

0:19:37.294,0:19:41.822
but it's not just[br]a difference in quantity,

0:19:41.822,0:19:43.586
it's a difference in kind. Right?

0:19:43.586,0:19:45.746
We have incredibly powerful computers,

0:19:45.746,0:19:48.137
and they do do amazingly[br]sophisticated things,

0:19:48.137,0:19:51.341
often with very big amounts of data.

0:19:51.341,0:19:53.948
Human minds do, I think,[br]something quite different,

0:19:53.948,0:19:57.843
and I think it's the structured,[br]hierarchical nature of human knowledge

0:19:57.843,0:19:59.875
that remains a real challenge.

0:19:59.875,0:20:02.936
CA: Laura Schulz, wonderful[br]food for thought. Thank you so much.

0:20:02.936,0:20:05.858
LS: Thank you.[br](Applause)