WEBVTT

00:00.840 --> 00:02.370
-: Hello and welcome to the course

00:02.370 --> 00:04.380
on Artificial Intelligence A to Z.

00:04.380 --> 00:05.400
My name's Kirill Eremenko

00:05.400 --> 00:07.050
and on behalf of Hadelin de Ponteves,

00:07.050 --> 00:08.820
the co-instructor in this course,

00:08.820 --> 00:11.910
I'd like to welcome you into this amazing adventure

00:11.910 --> 00:13.320
that we're about to start.

00:13.320 --> 00:15.240
And what's up with this timer?

00:15.240 --> 00:17.070
Well, the reason for this timer

00:17.070 --> 00:19.020
is this is my third time recording this tutorial

00:19.020 --> 00:21.450
and I get so excited about these things

00:21.450 --> 00:22.380
and it just goes over,

00:22.380 --> 00:24.390
like blows over 10 minutes or 20 minutes.

00:24.390 --> 00:27.630
So I decided to limit myself to five minutes

00:27.630 --> 00:30.330
to get you excited about this course,

00:30.330 --> 00:33.000
and why it's the best time to be in AI.

00:33.000 --> 00:34.200
So let's dive into it.

00:34.200 --> 00:36.000
We've got four and a half minutes left.

00:36.000 --> 00:36.950
Why AI?

00:36.950 --> 00:39.570
Well, we've all heard about applications

00:39.570 --> 00:40.620
of artificial intelligence

00:40.620 --> 00:42.120
that are popping up around us.

00:42.120 --> 00:45.660
Self-driving cars, medicine, heavy machinery,

00:45.660 --> 00:47.880
customer service, the list just keeps going on

00:47.880 --> 00:48.713
and on and on.

00:48.713 --> 00:49.546
But why right now?

00:49.546 --> 00:52.380
Why not 10 years ago or not 10 years later?

00:52.380 --> 00:55.110
Well, the answer is hidden in this curve,

00:55.110 --> 00:56.250
which so called Moore's law.

00:56.250 --> 01:00.300
It was originally identified in 1965 by Gordon Moore,

01:00.300 --> 01:04.110
and what it says is that the power of the average computer,

01:04.110 --> 01:05.880
like in very simple terms,

01:05.880 --> 01:07.050
the power of the average computer

01:07.050 --> 01:08.400
that you can buy for $1,000

01:08.400 --> 01:11.460
will double every 18 to 24 months.

01:11.460 --> 01:13.320
And that's been, as you can see,

01:13.320 --> 01:16.500
the law has held constant through good times and bad,

01:16.500 --> 01:20.130
war and peace, and the recession, depressions.

01:20.130 --> 01:21.750
Nothing stops this curve.

01:21.750 --> 01:24.870
And we've already surpassed the brain of an insect

01:24.870 --> 01:26.081
for an average computer

01:26.081 --> 01:27.330
that you can buy for 1,000 bucks.

01:27.330 --> 01:30.840
We're sitting at about the processing power

01:30.840 --> 01:32.040
of the brain of a rat.

01:32.040 --> 01:34.800
By 2025, we'll be at the brain of a human.

01:34.800 --> 01:36.870
As you can see, it's exponentially growing

01:36.870 --> 01:39.450
and so this is a great time

01:39.450 --> 01:40.710
to be in artificial intelligence.

01:40.710 --> 01:42.150
It's just gonna be doubling,

01:42.150 --> 01:44.250
the power is gonna be doubling every time.

01:45.240 --> 01:46.531
Three minutes.

01:46.531 --> 01:48.750
So some of my favorite examples

01:48.750 --> 01:53.190
are related to games like chess and Go.

01:53.190 --> 01:55.021
So in 1997,

01:55.021 --> 01:56.700
DeepMind was the first computer

01:56.700 --> 01:58.830
to beat Garry Kasparov in chess.

01:58.830 --> 01:59.940
It was a huge thing.

01:59.940 --> 02:03.420
It was like one of the most broadcasted chess games

02:03.420 --> 02:05.040
on the planet ever.

02:05.040 --> 02:06.776
And at the same time...

02:06.776 --> 02:09.930
And considered an impossible feat.

02:09.930 --> 02:12.870
Now on any phone, you can install an app with chess,

02:12.870 --> 02:14.970
which no human player will ever beat

02:14.970 --> 02:17.730
because computers are that good now.

02:17.730 --> 02:19.350
You just cannot beat a computer in chess

02:19.350 --> 02:21.330
if you put it on the hardest setting.

02:21.330 --> 02:24.879
And then just recently in March 2016,

02:24.879 --> 02:29.250
a computer won against the 18 world times champion,

02:29.250 --> 02:32.850
18 world champion of the game of Go, Lee Sedol.

02:32.850 --> 02:35.220
Again, considered an impossible feat,

02:35.220 --> 02:36.680
even at the start of the match,

02:36.680 --> 02:38.520
it was considered to be impossible.

02:38.520 --> 02:40.320
That it's only gonna happen 10 years from then.

02:40.320 --> 02:41.310
But it happened.

02:41.310 --> 02:42.933
And by the way,

02:42.933 --> 02:45.970
if you'd like, find out a bit more about the game of Go.

02:45.970 --> 02:49.260
It's very, very exciting, very interesting.

02:49.260 --> 02:52.110
It's got 40 million players, 3,000-year-old game.

02:52.110 --> 02:55.110
Very popular in China, Japan and Korea.

02:55.110 --> 02:58.230
And it's much more complex than chess.

02:58.230 --> 03:01.200
So just there's many more impossible combinations.

03:01.200 --> 03:03.840
So again, something that was considered impossible

03:03.840 --> 03:05.670
is happening and it's happening now,

03:05.670 --> 03:07.680
and much earlier than its time.

03:07.680 --> 03:08.700
So how are we going for time?

03:08.700 --> 03:10.710
1 minute 52 seconds.

03:10.710 --> 03:13.440
So that's what you'll see from this course.

03:13.440 --> 03:16.170
You'll see that we're using games to train AI.

03:16.170 --> 03:17.040
Why are we doing that?

03:17.040 --> 03:19.260
Because games are a confined environment

03:19.260 --> 03:21.570
where if we can beat a game,

03:21.570 --> 03:23.670
using an artificial intelligence,

03:23.670 --> 03:25.620
then we can use the same principles

03:25.620 --> 03:26.784
to apply it to business,

03:26.784 --> 03:28.624
and that's the exact same thing

03:28.624 --> 03:31.062
that the team at Google DeepMind

03:31.062 --> 03:33.390
who created AlphaGo who won against Lee Sedol,

03:33.390 --> 03:34.740
that's exactly what they did.

03:34.740 --> 03:36.750
They applied artificial intelligence

03:36.750 --> 03:40.830
to a Google warehouse to control the air cooling

03:40.830 --> 03:43.164
and what they found this is the electricity bill,

03:43.164 --> 03:44.370
the electricity consumption.

03:44.370 --> 03:46.260
You can see it's going like this.

03:46.260 --> 03:47.640
As soon they switch on the AI,

03:47.640 --> 03:49.080
bam, it goes down.

03:49.080 --> 03:50.880
And then they switch it back off, it goes up.

03:50.880 --> 03:52.890
They were able to save 40%

03:52.890 --> 03:54.420
on their electricity bill,

03:54.420 --> 03:55.890
and I have one minute left.

03:55.890 --> 03:56.723
40%.

03:56.723 --> 03:59.760
So you can find the full article on the DeepMind blog.

03:59.760 --> 04:01.050
That is insane.

04:01.050 --> 04:03.575
Imagine that for a company of Google's scale,

04:03.575 --> 04:06.420
which has millions and millions of servers.

04:06.420 --> 04:09.240
So there you go, that's your connection

04:09.240 --> 04:11.790
between applying artificial intelligence in games

04:11.790 --> 04:13.320
and applying it to business.

04:13.320 --> 04:16.020
Very, very closely related

04:16.020 --> 04:20.280
and that's what you will be able to do after this course.

04:20.280 --> 04:22.466
And finally, one of my favorite blogs,

04:22.466 --> 04:23.490
waitbutwhy.com.

04:23.490 --> 04:26.490
Check it out, amazing blog by Tim Urban.

04:26.490 --> 04:29.190
This is how he describes the situation we're in.

04:29.190 --> 04:31.800
This is us and technological progress.

04:31.800 --> 04:32.633
We're sitting here.

04:32.633 --> 04:33.466
This is what it looks like.

04:33.466 --> 04:35.100
It looks very linear.

04:35.100 --> 04:37.200
But because it's an exponential boom,

04:37.200 --> 04:39.240
this is what's gonna happen in the next years.

04:39.240 --> 04:40.950
And there's very strong evidence

04:40.950 --> 04:42.150
from the things we're seeing

04:42.150 --> 04:44.070
to suggest we're sitting right over here

04:44.070 --> 04:45.270
on the kink of the curve.

04:45.270 --> 04:47.010
And that, my friends, in short

04:47.010 --> 04:50.010
is why it's so exciting to be alive right now,

04:50.010 --> 04:51.780
why it's so exciting to be

04:51.780 --> 04:54.810
and get into and be in the space of artificial intelligence

04:54.810 --> 04:56.730
because the power you'll have,

04:56.730 --> 04:59.550
the applications that you will be able

04:59.550 --> 05:02.730
to create or the ways you'll be able to apply it

05:02.730 --> 05:05.250
is gonna be intense and insane.

05:05.250 --> 05:06.870
And I can't wait to see you

05:06.870 --> 05:08.100
on the first tutorial of this course.

05:08.100 --> 05:10.413
Let's get this started.
