Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:01,929 --> 00:00:05,270
What about that other classic AI stuff?
We talked about theorem proving,
2
00:00:05,270 --> 00:00:08,759
can we prove mathematical theorems.
That stuff is happening too and it's also important.
3
00:00:08,759 --> 00:00:12,900
So, on the logic front for
example we can build amazing theorem provers
4
00:00:12,900 --> 00:00:16,330
compared to when people started
doing this. These provers are used for
5
00:00:16,330 --> 00:00:19,710
more than just proving theorems. For
example, NASA uses these for fault diagnosis.
6
00:00:19,710 --> 00:00:22,820
There are some question
answering systems which are no longer
7
00:00:22,820 --> 00:00:26,500
purely logical, but there is a
component of logical theorem proving. You try to
8
00:00:26,500 --> 00:00:28,089
prove that the statement
9
00:00:28,089 --> 00:00:31,809
that you think answers the question
actually entails the question in the appropriate way.
10
00:00:31,809 --> 00:00:33,050
11
00:00:33,050 --> 00:00:36,860
That's the way that logic and theorem proving
has made its way into very different
12
00:00:36,860 --> 00:00:40,780
domains like natural language processing.
The methods people use are things like
13
00:00:40,780 --> 00:00:44,380
deduction systems. The closest we'll get
to this in this class is constraint satisfaction,
14
00:00:44,380 --> 00:00:47,800
but you'll get a flavor for how this stuff works.
15
00:00:47,800 --> 00:00:51,170
Also it's worth pointing out that
satisfiability solvers have had
16
00:00:51,170 --> 00:00:55,580
huge advances in the past decade and are now
able to do really amazing things.
17
00:00:55,580 --> 00:01:00,920
What's shown here on the right is a proof of something called the Robbin's conjecture.
18
00:01:00,920 --> 00:01:01,990
You probably
19
00:01:01,990 --> 00:01:05,630
haven't heard the Robbin's conjecture
itself, but it was an open algebraic question,
20
00:01:05,630 --> 00:01:09,170
and there is a short, human
interpretable proof
21
00:01:09,170 --> 00:01:12,580
that a theorem prover chugged out that humans have been looking for for a while. This is a case of
22
00:01:12,580 --> 00:01:17,140
you see it, and you're like, yeah, that's right.
23
00:01:17,140 --> 00:01:20,690
That's a case of an open question
being proved by a computer in the way
24
00:01:20,690 --> 00:01:23,690
humans would prove it, in the sense that
the human can interpret the proof.
25
00:01:23,690 --> 00:01:26,810
There's also theorem proving by computers where
the computer does a brute force,
26
00:01:26,810 --> 00:01:29,800
where human doesn't want to check 4.7 billion cases, but the computer does,
27
00:01:29,800 --> 00:01:32,020
and so the computer does it.
2719
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.