WEBVTT

00:00:01.765 --> 00:00:04.765
So, artificial intelligence

00:00:04.789 --> 00:00:08.318
is known for disrupting
all kinds of industries.

00:00:08.961 --> 00:00:11.004
What about ice cream?

00:00:11.879 --> 00:00:15.518
What kind of mind-blowing
new flavors could we generate

00:00:15.542 --> 00:00:18.518
with the power of an advanced
artificial intelligence?

00:00:19.011 --> 00:00:23.172
So I teamed up with a group of coders
from Kealing Middle School

00:00:23.196 --> 00:00:25.437
to find out the answer to this question.

00:00:25.461 --> 00:00:30.542
They collected over 1,600
existing ice cream flavors,

00:00:30.566 --> 00:00:36.088
and together, we fed them to an algorithm
to see what it would generate.

00:00:36.112 --> 00:00:39.865
And here are some of the flavors
that the AI came up with.

NOTE Paragraph

00:00:40.444 --> 00:00:41.915
[Pumpkin Trash Break]

NOTE Paragraph

00:00:41.939 --> 00:00:43.341
(Laughter)

NOTE Paragraph

00:00:43.365 --> 00:00:45.834
[Peanut Butter Slime]

NOTE Paragraph

00:00:46.822 --> 00:00:48.165
[Strawberry Cream Disease]

NOTE Paragraph

00:00:48.189 --> 00:00:50.315
(Laughter)

NOTE Paragraph

00:00:50.339 --> 00:00:54.936
These flavors are not delicious,
as we might have hoped they would be.

00:00:54.960 --> 00:00:56.824
So the question is: What happened?

00:00:56.848 --> 00:00:58.242
What went wrong?

00:00:58.266 --> 00:01:00.225
Is the AI trying to kill us?

00:01:01.027 --> 00:01:05.337
Or is it trying to do what we asked,
and there was a problem?

NOTE Paragraph

00:01:06.567 --> 00:01:09.031
In movies, when something
goes wrong with AI,

00:01:09.055 --> 00:01:11.767
it's usually because the AI has decided

00:01:11.791 --> 00:01:14.063
that it doesn't want to obey
the humans anymore,

00:01:14.087 --> 00:01:16.710
and it's got its own goals,
thank you very much.

00:01:17.266 --> 00:01:20.482
In real life, though,
the AI that we actually have

00:01:20.506 --> 00:01:22.369
is not nearly smart enough for that.

00:01:22.781 --> 00:01:25.763
It has the approximate computing power

00:01:25.787 --> 00:01:27.063
of an earthworm,

00:01:27.087 --> 00:01:30.490
or maybe at most a single honeybee,

00:01:30.514 --> 00:01:32.729
and actually, probably maybe less.

00:01:32.753 --> 00:01:35.347
Like, we're constantly learning
new things about brains

00:01:35.371 --> 00:01:39.731
that make it clear how much our AIs
don't measure up to real brains.

00:01:39.755 --> 00:01:45.418
So today's AI can do a task
like identify a pedestrian in a picture,

00:01:45.442 --> 00:01:48.425
but it doesn't have a concept
of what the pedestrian is

00:01:48.449 --> 00:01:53.273
beyond that it's a collection
of lines and textures and things.

00:01:53.792 --> 00:01:56.313
It doesn't know what a human actually is.

00:01:56.822 --> 00:02:00.104
So will today's AI
do what we ask it to do?

00:02:00.128 --> 00:02:01.722
It will if it can,

00:02:01.746 --> 00:02:04.472
but it might not do what we actually want.

NOTE Paragraph

00:02:04.496 --> 00:02:06.911
So let's say that you
were trying to get an AI

00:02:06.935 --> 00:02:09.554
to take this collection of robot parts

00:02:09.578 --> 00:02:13.775
and assemble them into some kind of robot
to get from Point A to Point B.

00:02:13.799 --> 00:02:16.280
Now, if you were going to try
and solve this problem

00:02:16.304 --> 00:02:18.655
by writing a traditional-style
computer program,

00:02:18.679 --> 00:02:22.096
you would give the program
step-by-step instructions

00:02:22.120 --> 00:02:23.449
on how to take these parts,

00:02:23.473 --> 00:02:25.880
how to assemble them
into a robot with legs

00:02:25.904 --> 00:02:28.846
and then how to use those legs
to walk to Point B.

00:02:29.441 --> 00:02:31.781
But when you're using AI
to solve the problem,

00:02:31.805 --> 00:02:32.979
it goes differently.

00:02:33.003 --> 00:02:35.385
You don't tell it
how to solve the problem,

00:02:35.409 --> 00:02:36.888
you just give it the goal,

00:02:36.912 --> 00:02:40.174
and it has to figure out for itself
via trial and error

00:02:40.198 --> 00:02:41.682
how to reach that goal.

00:02:42.254 --> 00:02:46.356
And it turns out that the way AI tends
to solve this particular problem

00:02:46.380 --> 00:02:47.864
is by doing this:

00:02:47.888 --> 00:02:51.255
it assembles itself into a tower
and then falls over

00:02:51.279 --> 00:02:53.106
and lands at Point B.

00:02:53.130 --> 00:02:55.959
And technically, this solves the problem.

00:02:55.983 --> 00:02:57.622
Technically, it got to Point B.

00:02:57.646 --> 00:03:01.911
The danger of AI is not that
it's going to rebel against us,

00:03:01.935 --> 00:03:06.209
it's that it's going to do
exactly what we ask it to do.

00:03:06.876 --> 00:03:09.374
So then the trick
of working with AI becomes:

00:03:09.398 --> 00:03:13.226
How do we set up the problem
so that it actually does what we want?

NOTE Paragraph

00:03:14.726 --> 00:03:18.032
So this little robot here
is being controlled by an AI.

00:03:18.056 --> 00:03:20.870
The AI came up with a design
for the robot legs

00:03:20.894 --> 00:03:24.972
and then figured out how to use them
to get past all these obstacles.

00:03:24.996 --> 00:03:27.737
But when David Ha set up this experiment,

00:03:27.761 --> 00:03:30.617
he had to set it up
with very, very strict limits

00:03:30.641 --> 00:03:33.933
on how big the AI
was allowed to make the legs,

00:03:33.957 --> 00:03:35.507
because otherwise ...

NOTE Paragraph

00:03:43.058 --> 00:03:46.989
(Laughter)

NOTE Paragraph

00:03:48.563 --> 00:03:52.308
And technically, it got
to the end of that obstacle course.

00:03:52.332 --> 00:03:57.274
So you see how hard it is to get AI
to do something as simple as just walk.

NOTE Paragraph

00:03:57.298 --> 00:04:01.118
So seeing the AI do this,
you may say, OK, no fair,

00:04:01.142 --> 00:04:03.722
you can't just be
a tall tower and fall over,

00:04:03.746 --> 00:04:07.181
you have to actually, like,
use legs to walk.

00:04:07.205 --> 00:04:09.964
And it turns out,
that doesn't always work, either.

00:04:09.988 --> 00:04:12.747
This AI's job was to move fast.

00:04:13.115 --> 00:04:16.708
They didn't tell it that it had
to run facing forward

00:04:16.732 --> 00:04:18.990
or that it couldn't use its arms.

00:04:19.487 --> 00:04:24.105
So this is what you get
when you train AI to move fast,

00:04:24.129 --> 00:04:27.663
you get things like somersaulting
and silly walks.

00:04:27.687 --> 00:04:29.087
It's really common.

00:04:29.667 --> 00:04:32.846
So is twitching along the floor in a heap.

NOTE Paragraph

00:04:32.870 --> 00:04:34.020
(Laughter)

NOTE Paragraph

00:04:35.241 --> 00:04:38.495
So in my opinion, you know what
should have been a whole lot weirder

00:04:38.519 --> 00:04:39.915
is the "Terminator" robots.

00:04:40.256 --> 00:04:44.011
Hacking "The Matrix" is another thing
that AI will do if you give it a chance.

00:04:44.035 --> 00:04:46.552
So if you train an AI in a simulation,

00:04:46.576 --> 00:04:50.689
it will learn how to do things like
hack into the simulation's math errors

00:04:50.713 --> 00:04:52.920
and harvest them for energy.

00:04:52.944 --> 00:04:58.419
Or it will figure out how to move faster
by glitching repeatedly into the floor.

00:04:58.443 --> 00:05:00.028
When you're working with AI,

00:05:00.052 --> 00:05:02.441
it's less like working with another human

00:05:02.465 --> 00:05:06.094
and a lot more like working
with some kind of weird force of nature.

00:05:06.562 --> 00:05:11.185
And it's really easy to accidentally
give AI the wrong problem to solve,

00:05:11.209 --> 00:05:15.747
and often we don't realize that
until something has actually gone wrong.

NOTE Paragraph

00:05:16.242 --> 00:05:18.322
So here's an experiment I did,

00:05:18.346 --> 00:05:21.528
where I wanted the AI
to copy paint colors,

00:05:21.552 --> 00:05:23.298
to invent new paint colors,

00:05:23.322 --> 00:05:26.309
given the list like the ones
here on the left.

00:05:26.798 --> 00:05:29.802
And here's what the AI
actually came up with.

NOTE Paragraph

00:05:29.826 --> 00:05:32.969
[Sindis Poop, Turdly, Suffer, Gray Pubic]

NOTE Paragraph

00:05:32.993 --> 00:05:37.223
(Laughter)

NOTE Paragraph

00:05:39.177 --> 00:05:41.063
So technically,

00:05:41.087 --> 00:05:42.951
it did what I asked it to.

00:05:42.975 --> 00:05:46.283
I thought I was asking it for,
like, nice paint color names,

00:05:46.307 --> 00:05:48.614
but what I was actually asking it to do

00:05:48.638 --> 00:05:51.724
was just imitate the kinds
of letter combinations

00:05:51.748 --> 00:05:53.653
that it had seen in the original.

00:05:53.677 --> 00:05:56.775
And I didn't tell it anything
about what words mean,

00:05:56.799 --> 00:05:59.359
or that there are maybe some words

00:05:59.383 --> 00:06:02.272
that it should avoid using
in these paint colors.

00:06:03.141 --> 00:06:06.635
So its entire world
is the data that I gave it.

00:06:06.659 --> 00:06:10.687
Like with the ice cream flavors,
it doesn't know about anything else.

NOTE Paragraph

00:06:12.491 --> 00:06:14.129
So it is through the data

00:06:14.153 --> 00:06:18.197
that we often accidentally tell AI
to do the wrong thing.

00:06:18.694 --> 00:06:21.726
This is a fish called a tench.

00:06:21.750 --> 00:06:23.565
And there was a group of researchers

00:06:23.589 --> 00:06:27.463
who trained an AI to identify
this tench in pictures.

00:06:27.487 --> 00:06:28.783
But then when they asked it

00:06:28.807 --> 00:06:32.233
what part of the picture it was actually
using to identify the fish,

00:06:32.257 --> 00:06:33.615
here's what it highlighted.

00:06:35.203 --> 00:06:37.392
Yes, those are human fingers.

00:06:37.416 --> 00:06:39.475
Why would it be looking for human fingers

00:06:39.499 --> 00:06:41.420
if it's trying to identify a fish?

00:06:42.126 --> 00:06:45.290
Well, it turns out that the tench
is a trophy fish,

00:06:45.314 --> 00:06:49.125
and so in a lot of pictures
that the AI had seen of this fish

00:06:49.149 --> 00:06:50.300
during training,

00:06:50.324 --> 00:06:51.814
the fish looked like this.

NOTE Paragraph

00:06:51.838 --> 00:06:53.473
(Laughter)

NOTE Paragraph

00:06:53.497 --> 00:06:56.827
And it didn't know that the fingers
aren't part of the fish.

NOTE Paragraph

00:06:58.808 --> 00:07:02.928
So you see why it is so hard
to design an AI

00:07:02.952 --> 00:07:06.271
that actually can understand
what it's looking at.

00:07:06.295 --> 00:07:09.157
And this is why designing
the image recognition

00:07:09.181 --> 00:07:11.248
in self-driving cars is so hard,

00:07:11.272 --> 00:07:13.477
and why so many self-driving car failures

00:07:13.501 --> 00:07:16.386
are because the AI got confused.

00:07:16.410 --> 00:07:20.418
I want to talk about an example from 2016.

00:07:20.442 --> 00:07:24.897
There was a fatal accident when somebody
was using Tesla's autopilot AI,

00:07:24.921 --> 00:07:28.335
but instead of using it on the highway
like it was designed for,

00:07:28.359 --> 00:07:30.564
they used it on city streets.

00:07:31.239 --> 00:07:32.414
And what happened was,

00:07:32.438 --> 00:07:35.834
a truck drove out in front of the car
and the car failed to brake.

00:07:36.507 --> 00:07:41.269
Now, the AI definitely was trained
to recognize trucks in pictures.

00:07:41.293 --> 00:07:43.438
But what it looks like happened is

00:07:43.462 --> 00:07:46.393
the AI was trained to recognize
trucks on highway driving,

00:07:46.417 --> 00:07:49.316
where you would expect
to see trucks from behind.

00:07:49.340 --> 00:07:52.760
Trucks on the side is not supposed
to happen on a highway,

00:07:52.784 --> 00:07:56.239
and so when the AI saw this truck,

00:07:56.263 --> 00:08:01.090
it looks like the AI recognized it
as most likely to be a road sign

00:08:01.114 --> 00:08:03.387
and therefore, safe to drive underneath.

NOTE Paragraph

00:08:04.114 --> 00:08:06.694
Here's an AI misstep
from a different field.

00:08:06.718 --> 00:08:10.178
Amazon recently had to give up
on a résumé-sorting algorithm

00:08:10.202 --> 00:08:11.422
that they were working on

00:08:11.446 --> 00:08:15.354
when they discovered that the algorithm
had learned to discriminate against women.

00:08:15.378 --> 00:08:18.094
What happened is they had trained it
on example résumés

00:08:18.118 --> 00:08:20.360
of people who they had hired in the past.

00:08:20.384 --> 00:08:24.407
And from these examples, the AI learned
to avoid the résumés of people

00:08:24.431 --> 00:08:26.457
who had gone to women's colleges

00:08:26.481 --> 00:08:29.287
or who had the word "women"
somewhere in their resume,

00:08:29.311 --> 00:08:33.887
as in, "women's soccer team"
or "Society of Women Engineers."

00:08:33.911 --> 00:08:37.885
The AI didn't know that it wasn't supposed
to copy this particular thing

00:08:37.909 --> 00:08:39.887
that it had seen the humans do.

00:08:39.911 --> 00:08:43.088
And technically, it did
what they asked it to do.

00:08:43.112 --> 00:08:45.909
They just accidentally asked it
to do the wrong thing.

NOTE Paragraph

00:08:46.653 --> 00:08:49.548
And this happens all the time with AI.

00:08:50.120 --> 00:08:53.711
AI can be really destructive
and not know it.

00:08:53.735 --> 00:08:58.813
So the AIs that recommend
new content in Facebook, in YouTube,

00:08:58.837 --> 00:09:02.376
they're optimized to increase
the number of clicks and views.

00:09:02.400 --> 00:09:05.836
And unfortunately, one way
that they have found of doing this

00:09:05.860 --> 00:09:10.363
is to recommend the content
of conspiracy theories or bigotry.

00:09:10.902 --> 00:09:16.204
The AIs themselves don't have any concept
of what this content actually is,

00:09:16.228 --> 00:09:19.623
and they don't have any concept
of what the consequences might be

00:09:19.647 --> 00:09:21.756
of recommending this content.

NOTE Paragraph

00:09:22.296 --> 00:09:24.307
So, when we're working with AI,

00:09:24.331 --> 00:09:28.513
it's up to us to avoid problems.

00:09:28.537 --> 00:09:30.860
And avoiding things going wrong,

00:09:30.884 --> 00:09:35.410
that may come down to
the age-old problem of communication,

00:09:35.434 --> 00:09:39.179
where we as humans have to learn
how to communicate with AI.

00:09:39.203 --> 00:09:43.242
We have to learn what AI
is capable of doing and what it's not,

00:09:43.266 --> 00:09:46.352
and to understand that,
with its tiny little worm brain,

00:09:46.376 --> 00:09:50.389
AI doesn't really understand
what we're trying to ask it to do.

00:09:51.148 --> 00:09:54.469
So in other words, we have
to be prepared to work with AI

00:09:54.493 --> 00:09:59.751
that's not the super-competent,
all-knowing AI of science fiction.

00:09:59.775 --> 00:10:02.637
We have to be prepared to work with an AI

00:10:02.661 --> 00:10:05.599
that's the one that we actually have
in the present day.

00:10:05.623 --> 00:10:09.828
And present-day AI is plenty weird enough.

NOTE Paragraph

00:10:09.852 --> 00:10:11.042
Thank you.

NOTE Paragraph

00:10:11.066 --> 00:10:16.291
(Applause)