1
00:00:11,450 --> 00:00:15,840
Our first example today
is a story of IBM Watson.

2
00:00:15,840 --> 00:00:19,370
IBM research, a distinguished
industrial laboratory,

3
00:00:19,370 --> 00:00:22,350
strives to push the
limits of science.

4
00:00:22,350 --> 00:00:26,360
In the mid-1990s it created
Deep Blue, a computer that

5
00:00:26,360 --> 00:00:30,060
beat Garry Kasparov, the world
champion in Chess at the time,

6
00:00:30,060 --> 00:00:33,180
showing to the world that
machines can beat humans

7
00:00:33,180 --> 00:00:34,960
in tasks that
people thought were

8
00:00:34,960 --> 00:00:38,120
restricted to
human intelligence.

9
00:00:38,120 --> 00:00:41,130
In the 1990s-- in
the late 1990s--

10
00:00:41,130 --> 00:00:45,720
it created Blue Gene, a computer
to map the human genome.

11
00:00:45,720 --> 00:00:49,830
In 2005 IBM decided to
create a computer that

12
00:00:49,830 --> 00:00:53,000
would compete at Jeopardy,
a popular game show.

13
00:00:56,430 --> 00:00:59,620
From the T.J. Watson
Research Center

14
00:00:59,620 --> 00:01:02,390
in Yorktown Heights,
New York, this

15
00:01:02,390 --> 00:01:07,110
is Jeopardy-- The IBM Challenge.

16
00:01:07,110 --> 00:01:09,030
Please welcome back
our contestants.

17
00:01:09,030 --> 00:01:10,460
He has never been defeated.

18
00:01:10,460 --> 00:01:13,640
And his winnings of
more than $3.2 million

19
00:01:13,640 --> 00:01:17,110
make him Jeopardy's all
time biggest money winner.

20
00:01:17,110 --> 00:01:22,300
From Los Angeles, California,
here's Brad Rutter.

21
00:01:22,300 --> 00:01:26,110
An IBM computer system able to
rapidly understand and analyze

22
00:01:26,110 --> 00:01:29,080
natural language,
including puns, riddles,

23
00:01:29,080 --> 00:01:32,350
and complex questions across
a broad range of knowledge,

24
00:01:32,350 --> 00:01:35,850
please welcome Watson.

25
00:01:35,850 --> 00:01:41,220
In 2004 he captivated America by
winning 74 consecutive matches

26
00:01:41,220 --> 00:01:44,570
and $2.5 million on Jeopardy.

27
00:01:44,570 --> 00:01:49,800
From Seattle, Washington,
here's Ken Jennings.

28
00:01:49,800 --> 00:01:56,610
And now here is the host
of Jeopardy, Alex Trebek.

29
00:01:56,610 --> 00:01:58,080
Thank you, Johnny.

30
00:01:58,080 --> 00:01:59,590
Before we get into
the game, there

31
00:01:59,590 --> 00:02:01,080
are just a couple
of things I need

32
00:02:01,080 --> 00:02:02,430
to tell you about this match.

33
00:02:02,430 --> 00:02:05,350
Now as I said earlier,
Watson will receive the clues

34
00:02:05,350 --> 00:02:09,759
electronically as a text file
at the same moment the clues are

35
00:02:09,759 --> 00:02:12,000
revealed to Ken and Brad.

36
00:02:12,000 --> 00:02:14,070
And at the same
time I read them.

37
00:02:14,070 --> 00:02:15,660
This competition
will be a two game

38
00:02:15,660 --> 00:02:17,720
total point exhibition match.

39
00:02:17,720 --> 00:02:20,670
However, these two games will be
played out over the next three

40
00:02:20,670 --> 00:02:23,090
days so we can tell
the full story.

41
00:02:23,090 --> 00:02:25,960
Throughout the games you'll
get a glimpse of the thinking

42
00:02:25,960 --> 00:02:30,200
process, if you will, that
is behind Watson's responses.

43
00:02:30,200 --> 00:02:33,380
Now this will be done through
an answer panel display

44
00:02:33,380 --> 00:02:36,140
at the bottom of the screen.

45
00:02:36,140 --> 00:02:37,270
Let's play Jeopardy.

46
00:02:37,270 --> 00:02:37,820
Here we go.

47
00:02:40,920 --> 00:02:44,150
Our first round of play
contains these categories--

48
00:02:44,150 --> 00:02:50,470
Literary Character APB, All
Points Bulletin; Beatles

49
00:02:50,470 --> 00:02:56,120
People; Olympic Oddities;
Name the Decade;

50
00:02:56,120 --> 00:03:01,120
Final Frontiers; and
Alternate Meanings.

51
00:03:01,120 --> 00:03:03,090
A little while ago
we had a drawing

52
00:03:03,090 --> 00:03:05,530
to determine which player
would select first.

53
00:03:05,530 --> 00:03:06,970
Brad, you won that.

54
00:03:06,970 --> 00:03:09,270
So if you're ready
make your first choice.

55
00:03:09,270 --> 00:03:11,950
Let's take alternate
meanings for $200, Alex.

56
00:03:11,950 --> 00:03:16,040
Four letter word for a
vantage point or a belief.

57
00:03:16,040 --> 00:03:16,620
Brad?

58
00:03:16,620 --> 00:03:17,240
What is a view?

59
00:03:17,240 --> 00:03:17,760
Yes.

60
00:03:17,760 --> 00:03:19,260
Alternate meanings, $400.

61
00:03:19,260 --> 00:03:21,530
4-letter word for
the iron fitting

62
00:03:21,530 --> 00:03:25,940
on the hoof of a horse or a
card-dealing box in a Casino.

63
00:03:25,940 --> 00:03:26,510
Watson?

64
00:03:26,510 --> 00:03:27,500
What is shoe?

65
00:03:27,500 --> 00:03:28,130
You are right.

66
00:03:28,130 --> 00:03:28,990
You get to pick.

67
00:03:28,990 --> 00:03:32,260
Literary Character APB for $800.

68
00:03:32,260 --> 00:03:35,850
Answer the Daily Double.

69
00:03:35,850 --> 00:03:38,370
Now Watson, although
you have but $400,

70
00:03:38,370 --> 00:03:40,510
you know of course
that you can risk up

71
00:03:40,510 --> 00:03:42,550
to the maximum value
of a clue on the board.

72
00:03:42,550 --> 00:03:45,200
And that is $1,000.

73
00:03:45,200 --> 00:03:46,230
$1,000, please.

74
00:03:46,230 --> 00:03:46,730
All right.

75
00:03:46,730 --> 00:03:48,980
Here is the Daily
Double clue for you.

76
00:03:48,980 --> 00:03:52,240
Wanted for killing
Sir Danvers Carew.

77
00:03:52,240 --> 00:03:54,790
Appearance pale and dwarfish.

78
00:03:54,790 --> 00:03:58,160
Seems to have a
split personality.

79
00:03:58,160 --> 00:03:59,720
Who is Hyde?

80
00:03:59,720 --> 00:04:00,550
Hyde, yes.

81
00:04:00,550 --> 00:04:02,840
Dr. Jekyll and Mr. Hyde,
either one acceptable.

82
00:04:02,840 --> 00:04:05,840
You're now in the
lead with $1400.

83
00:04:05,840 --> 00:04:06,740
Go again.

84
00:04:06,740 --> 00:04:09,400
Beatles People for $200.

85
00:04:09,400 --> 00:04:13,590
And any time you feel the
pain, hey, this guy, refrain.

86
00:04:13,590 --> 00:04:17,079
Don't carry the world
upon your shoulders.

87
00:04:17,079 --> 00:04:17,730
Watson?

88
00:04:17,730 --> 00:04:18,959
Who is Jude?

89
00:04:18,959 --> 00:04:19,760
Yes.

90
00:04:19,760 --> 00:04:22,160
Olympic Oddities for $200.

91
00:04:22,160 --> 00:04:27,290
Milorad Cavic almost upset this
man's perfect 2008 Olympics,

92
00:04:27,290 --> 00:04:31,270
losing to him by one
hundredth of a second.

93
00:04:31,270 --> 00:04:31,820
Watson?

94
00:04:31,820 --> 00:04:33,620
Who is Michael Phelps?

95
00:04:33,620 --> 00:04:34,180
Yes.

96
00:04:34,180 --> 00:04:34,740
Go.

97
00:04:34,740 --> 00:04:36,930
Name the Decade for $200.

98
00:04:36,930 --> 00:04:40,590
Disneyland opens and the
peace symbol is created.

99
00:04:40,590 --> 00:04:41,090
Ken.

100
00:04:41,090 --> 00:04:41,890
What are the '50s?

101
00:04:41,890 --> 00:04:42,390
Yes.

102
00:04:42,390 --> 00:04:44,470
Final Frontiers
for $1,000, Alex.

103
00:04:44,470 --> 00:04:46,440
Tickets aren't needed
for this event,

104
00:04:46,440 --> 00:04:50,659
a black hole's boundary from
which matter cannot escape.

105
00:04:50,659 --> 00:04:51,540
Watson?

106
00:04:51,540 --> 00:04:53,470
What is event horizon?

107
00:04:53,470 --> 00:04:54,600
Yes.

108
00:04:54,600 --> 00:04:57,230
Why is Jeopardy hard?

109
00:04:57,230 --> 00:05:00,560
Jeopardy asks the contestants
to answer cryptic questions

110
00:05:00,560 --> 00:05:02,790
in a huge variety of categories.

111
00:05:02,790 --> 00:05:06,040
It is generally seen as a
test of human intelligence,

112
00:05:06,040 --> 00:05:07,920
reasoning and cleverness.

113
00:05:07,920 --> 00:05:10,500
No links to the outside
world are permitted.

114
00:05:10,500 --> 00:05:13,590
And new questions and categories
are created for every show.

115
00:05:16,530 --> 00:05:19,720
Watson is a supercomputer
with 3,000 processors

116
00:05:19,720 --> 00:05:23,090
and a database of 200
million pages of information.

117
00:05:23,090 --> 00:05:26,920
It has a massive number of data
sources like encyclopedias,

118
00:05:26,920 --> 00:05:30,910
texts, manuals,
magazines, Wikipedia, etc.

119
00:05:30,910 --> 00:05:33,210
IBM researchers who
developed Watson

120
00:05:33,210 --> 00:05:36,360
used over 100 different
analytical techniques

121
00:05:36,360 --> 00:05:40,080
for analyzing natural language,
finding candidate answers

122
00:05:40,080 --> 00:05:41,860
and selecting the final answer.

123
00:05:41,860 --> 00:05:46,460
We'll discuss this more
later in the class.

124
00:05:46,460 --> 00:05:50,070
In February, 2011, a two
game exhibition match

125
00:05:50,070 --> 00:05:53,350
aired on television six years
after the initial conception

126
00:05:53,350 --> 00:05:55,240
of the idea to build Watson.

127
00:05:55,240 --> 00:05:58,980
Watson competed against the best
two human players of all time

128
00:05:58,980 --> 00:06:01,640
and challenged the
meaning of intelligence.

129
00:06:01,640 --> 00:06:05,470
Now Watson is being used for
many applications including

130
00:06:05,470 --> 00:06:10,470
selecting the best course
of treatment for cancer.

131
00:06:10,470 --> 00:06:11,720
What is the edge in Watson?

132
00:06:11,720 --> 00:06:14,710
Watson combined many
algorithms to increase

133
00:06:14,710 --> 00:06:16,480
accuracy and confidence.

134
00:06:16,480 --> 00:06:18,840
We'll cover many of
them in this class.

135
00:06:18,840 --> 00:06:21,140
IBM approached the
problem in a different way

136
00:06:21,140 --> 00:06:23,310
than how a humans does it.

137
00:06:23,310 --> 00:06:26,320
Watson deals with
massive amounts of data,

138
00:06:26,320 --> 00:06:29,230
often in unstructured
form, which is important

139
00:06:29,230 --> 00:06:32,840
as 90% of the data in the
world is unstructured.