1
00:00:09,100 --> 00:00:14,980
Playing a strategy game like chess or checkers couples intellectual activity with competition.

2
00:00:15,990 --> 00:00:20,900
By playing games we can often exercise and improve our intellectual skills.

3
00:00:21,310 --> 00:00:25,230
The competition adds excitement and allows us to compare our skills to those of others.

4
00:00:26,720 --> 00:00:33,860
Same motivation accounts for interest in computer game playing as a testbed for artificial intelligence.

5
00:00:34,290 --> 00:00:41,130
Programs that think better should be able to win more games and so we can use competition as an evaluation technique for intelligent systems.

6
00:00:42,410 --> 00:00:48,760
Unfortunately, building programs to play specific games has limited value in artificial intelligence.

7
00:00:49,780 --> 00:00:52,380
To begin with, specialized game players are very narrow.

8
00:00:53,400 --> 00:00:55,870
They are often good at one game but not so good at others.

9
00:00:56,700 --> 00:01:03,580
Deep Blue for example might have beaten the wolrd chess champion, but it has no clue how to play checkers and cannot even balance a checkbook.

10
00:01:05,250 --> 00:01:11,880
A second problem with specialized game playing is that such systems do only part of the work:

11
00:01:12,640 --> 00:01:18,000
most of the interesting analysis and design is done in advance by their programmers

12
00:01:18,410 --> 00:01:20,960
system themselves might just as well be teleoperated.

13
00:01:22,300 --> 00:01:23,470
However all is not lost.

14
00:01:25,000 --> 00:01:33,680
Many believe that the idea of game playing can be used to good effects to inspire and evaluate good work in artificial intelligence,

15
00:01:34,040 --> 00:01:38,310
but it requires moving more of the mental work to the computer itself.

16
00:01:39,150 --> 00:01:42,220
This can be done by focusing on general game playing.

17
00:01:44,160 --> 00:01:55,080
General game players are systems able to play arbitrary games effectively based solely on formal descriptions supplied at runtime.

18
00:01:56,140 --> 00:02:01,240
In a typical general game playing session, the players know nothing about the game in advance.

19
00:02:01,840 --> 00:02:09,570
Once the game begins, they receive a game description and based solely on this description they must figure out how to play the game legally and effectively.

20
00:02:10,340 --> 00:02:22,910
Furthermore they have to deal with uncertainty in the form of actions of the other players and usually ressource bounds limitations as well in the form of a game clock that limits their computation time.

21
00:02:24,950 --> 00:02:31,790
Given this arrangement, general game players cannot rely on algorithms designed in advance for specific games.

22
00:02:32,640 --> 00:02:41,770
General game playing expertise depends on intelligence on the part of the game player and not just intelligence of the programmer of the game player.

23
00:02:42,850 --> 00:02:55,030
In order to perform well, general game players typically incorporate ideas from multiple subareas of artificial intelligence such as knowledge representation, reasoning, rational decision making and so forth.

24
00:02:57,140 --> 00:03:02,040
The upshot of this is that general game players are able to play different kinds of games.

25
00:03:02,870 --> 00:03:07,610
They can play simple games like tic-tac-toe and complex games like chess.

26
00:03:08,430 --> 00:03:15,010
Games with simultaneous play like Diplomacy and games with alternating play like Risk.

27
00:03:15,620 --> 00:03:21,170
Gales with complete information or games with incomplete information such as Battleships.

28
00:03:22,020 --> 00:03:23,320
Games with different numbers of players.

29
00:03:24,430 --> 00:03:27,450
Games with or without communication among the players.

30
00:03:28,270 --> 00:03:32,090
Zero-sum games, cooperative games and so forth.