Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,320 --> 00:00:04,430
What do we do in this course: we design rational agents. What's that?
2
00:00:04,430 --> 00:00:07,060
An agent is some entity that perceives
3
00:00:07,060 --> 00:00:08,700
and acts.
4
00:00:08,700 --> 00:00:10,619
What does that mean, well here's an agent,
5
00:00:10,619 --> 00:00:13,629
an agent that's trying to get at some kind of apple. We'll come back to
6
00:00:13,629 --> 00:00:15,719
this agent again later.
7
00:00:15,719 --> 00:00:18,710
We can think about this agent with the following diagram--we'll draw these
8
00:00:18,710 --> 00:00:20,189
agent diagrams from time to time.
9
00:00:20,189 --> 00:00:23,590
There's an environment and there's an agent.
10
00:00:23,590 --> 00:00:26,640
What's the difference? The environment is what you don't control,
11
00:00:26,640 --> 00:00:28,179
the agent is what you do.
12
00:00:28,179 --> 00:00:31,710
So, in particular, where you draw that boundary--you might think when
13
00:00:31,710 --> 00:00:35,980
you're driving the agent is like the car, it can turn left and it can turn right. But the agent
14
00:00:35,980 --> 00:00:39,170
is also your arms. You have to move the steering wheel.
15
00:00:39,170 --> 00:00:42,949
If the car is not working right maybe you want to think of the car as the environment, or
16
00:00:42,949 --> 00:00:46,050
maybe your arms not working right--you threw out your elbow driving so hard.
17
00:00:46,050 --> 00:00:49,659
Now the line between the agent and the environment is in your head. You can send
18
00:00:49,659 --> 00:00:52,670
the signal to your arm but who knows what your trick arm is going to do.
19
00:00:52,670 --> 00:00:55,699
And so where we draw this abstraction, just like everything else in computer science,
20
00:00:56,310 --> 00:00:57,840
depends on the problem we're solving
21
00:00:57,840 --> 00:01:00,240
and what abstraction is safe.
22
00:01:00,240 --> 00:01:01,839
Okay, so what can an agent do?
23
00:01:01,839 --> 00:01:03,380
It can perceive,
24
00:01:03,380 --> 00:01:05,810
we have percepts coming in.
25
00:01:05,810 --> 00:01:08,620
We have actions going to the environment. What does the environment do?
26
00:01:08,620 --> 00:01:11,320
It does whatever the environment does. The world does its thing.
27
00:01:11,320 --> 00:01:12,950
And then the agent
28
00:01:12,950 --> 00:01:15,890
takes what comes in and it formulates an action
29
00:01:15,890 --> 00:01:17,659
using the question mark.
30
00:01:17,659 --> 00:01:21,090
What the heck is the question mark? That's the all important agent function.
31
00:01:21,090 --> 00:01:23,780
That's the thing that maps inputs (sensation)
32
00:01:23,780 --> 00:01:27,360
to outputs (actuation).
33
00:01:27,360 --> 00:01:31,100
So a rational agent chooses actions that maximizes its utilities. Since it doesn't
34
00:01:31,100 --> 00:01:33,540
know what's gonna happen, like this agent's not really sure whether it's going to succeed,
35
00:01:33,540 --> 00:01:38,009
we have to talk about expectations. The characteristics of the
36
00:01:38,009 --> 00:01:41,570
percepts, the environment, the action space tell us how to solve the problem.
37
00:01:41,570 --> 00:01:45,880
If an agent can see everything, that's one thing. If the agent can't, that's another thing.
38
00:01:45,880 --> 00:01:48,470
So, for example, if you think about a video game. Sometimes you can see the whole board,
39
00:01:48,470 --> 00:01:53,240
sometimes there's the fog-of-war, and in general, you don't always have full information.
40
00:01:53,240 --> 00:01:56,390
The techniques you use are different.
41
00:01:56,390 --> 00:01:59,960
This class is largely about general AI techniques
42
00:01:59,960 --> 00:02:02,370
that apply to a variety of problems types,
43
00:02:02,370 --> 00:02:05,959
and about learning to recognize when and how a problem
44
00:02:05,959 --> 00:02:09,920
can be solved by one of those techniques, by thinking about what kinds of
45
00:02:09,920 --> 00:02:13,599
percepts do I have, what kinds of environment do I have. Is it adversarial,
46
00:02:13,599 --> 00:02:14,189
is it fully observed,
47
00:02:14,189 --> 00:02:17,700
and so on. And as we go through this course you'll get more and more of these tools
48
00:02:17,700 --> 00:02:19,140
to deal with more and more complicated environments.
4356
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.