1
00:00:00,000 --> 00:00:11,980


2
00:00:11,980 --> 00:00:13,460
Well, good morning.

3
00:00:13,460 --> 00:00:16,720
My name is Steve Ward, and I
have the intimidating honor

4
00:00:16,720 --> 00:00:19,580
of chairing a panel of
distinguished Turing Award

5
00:00:19,580 --> 00:00:20,080
winners.

6
00:00:20,080 --> 00:00:23,170
As you know, the Turing Award
is the most prestigious award

7
00:00:23,170 --> 00:00:25,300
available in computer science.

8
00:00:25,300 --> 00:00:30,610
And our panelists include
Professor Andrew Yao,

9
00:00:30,610 --> 00:00:34,510
who has been on many faculties,
including MIT's, and is

10
00:00:34,510 --> 00:00:39,880
currently from
Tsinghua in Beijing.

11
00:00:39,880 --> 00:00:43,930
MIT's Ron Rivest, and
MIT's Barbara Liskov,

12
00:00:43,930 --> 00:00:48,970
and Butler Lampson, who hails
from both MIT and Microsoft,

13
00:00:48,970 --> 00:00:53,770
and Professor Fernando
Corbató of MIT.

14
00:00:53,770 --> 00:00:58,900
I thought we would get started
by asking each panelist in turn

15
00:00:58,900 --> 00:01:04,810
to say a few retrospective words
about the work that earned them

16
00:01:04,810 --> 00:01:08,230
the Turing Award,
perhaps shedding

17
00:01:08,230 --> 00:01:09,760
a little light on
their motivation

18
00:01:09,760 --> 00:01:12,100
and what their
retrospective view is

19
00:01:12,100 --> 00:01:14,230
of the impact of their work.

20
00:01:14,230 --> 00:01:18,440
And I thought we would do this
in kind of chronological order.

21
00:01:18,440 --> 00:01:25,330
So I'll start with Corby, who
was cited for pioneering work,

22
00:01:25,330 --> 00:01:27,160
organizing the
concepts and leading

23
00:01:27,160 --> 00:01:30,370
the development of the
general-purpose, large-scale,

24
00:01:30,370 --> 00:01:32,320
time-sharing, and
resource-sharing computer

25
00:01:32,320 --> 00:01:35,290
systems CTSS and Multics.

26
00:01:35,290 --> 00:01:37,260
Corby?

27
00:01:37,260 --> 00:01:39,220
Thank you.

28
00:01:39,220 --> 00:01:43,120
Well, I'm reminded that--

29
00:01:43,120 --> 00:01:50,650
first of all-- that it was 50
years ago this spring that MIT

30
00:01:50,650 --> 00:01:54,670
had its 100th anniversary.

31
00:01:54,670 --> 00:02:00,730
And right here, in Kresge,
there was a series of symposia.

32
00:02:00,730 --> 00:02:07,540
One of which was chaired by
Martin Greenberger of the Sloan

33
00:02:07,540 --> 00:02:14,430
School, who had a
symposium called

34
00:02:14,430 --> 00:02:19,500
Management and the
Computers of the Future.

35
00:02:19,500 --> 00:02:23,910
That was published eventually
in a book by MIT Press,

36
00:02:23,910 --> 00:02:33,600
and it got relabeled Computers
and the World of the Future.

37
00:02:33,600 --> 00:02:37,950
It's still in print, actually,
thanks to demand printing,

38
00:02:37,950 --> 00:02:41,770
and for $40 you can get it.

39
00:02:41,770 --> 00:02:48,120
Most notably, in that
particular session on computing,

40
00:02:48,120 --> 00:02:51,300
there was a lecture
by John McCarthy,

41
00:02:51,300 --> 00:02:54,370
who subsequently
went to Stanford

42
00:02:54,370 --> 00:02:56,820
and got a Turing Award also.

43
00:02:56,820 --> 00:03:06,450
But his seminal work
began here at MIT.

44
00:03:06,450 --> 00:03:11,100
But he, in particular, in
his talk in that symposium,

45
00:03:11,100 --> 00:03:17,160
advocated time-sharing and
made a good case for it,

46
00:03:17,160 --> 00:03:20,100
and the notion of
a computer utility.

47
00:03:20,100 --> 00:03:30,820
Well, the other thing is that
computing 50 years ago was only

48
00:03:30,820 --> 00:03:32,690
a little over 10 years old.

49
00:03:32,690 --> 00:03:37,090
10 years old, modern computing
that is, programming.

50
00:03:37,090 --> 00:03:41,090
And the thing that was
staggering at the time

51
00:03:41,090 --> 00:03:46,670
was how frozen the attitude
of the industrialists were,

52
00:03:46,670 --> 00:03:49,790
and the computing manufacturers.

53
00:03:49,790 --> 00:03:51,830
They really didn't
understand why

54
00:03:51,830 --> 00:03:56,600
they needed to change
anything, and time-sharing

55
00:03:56,600 --> 00:04:00,730
was against the grain.

56
00:04:00,730 --> 00:04:02,990
So CTSS was started out--

57
00:04:02,990 --> 00:04:07,620
it was almost kind of a hack,
it was going to be a demo.

58
00:04:07,620 --> 00:04:11,970
It gradually, as it
succeeded, grew and grew

59
00:04:11,970 --> 00:04:14,550
and was quite successful.

60
00:04:14,550 --> 00:04:17,180


61
00:04:17,180 --> 00:04:18,829
One of the other
things I would notice

62
00:04:18,829 --> 00:04:23,630
is that time-sharing
was the phrase,

63
00:04:23,630 --> 00:04:26,870
but the real goal was
man-machine interaction.

64
00:04:26,870 --> 00:04:35,530
Or as Licklider said,
man-machine symbiosis.

65
00:04:35,530 --> 00:04:38,690
There were two key
people in that evolution,

66
00:04:38,690 --> 00:04:43,810
one was Licklider, who
was an inspiring advocate,

67
00:04:43,810 --> 00:04:51,490
and Bob Fano, my friend,
who was also very wise

68
00:04:51,490 --> 00:04:59,470
and had the leadership to start
Project Mac, from which CSAIL

69
00:04:59,470 --> 00:05:00,865
evolved directly.

70
00:05:00,865 --> 00:05:05,990


71
00:05:05,990 --> 00:05:09,050
Fano gave a presentation
in anticipation

72
00:05:09,050 --> 00:05:13,160
of this celebration which
is on the web, I believe.

73
00:05:13,160 --> 00:05:16,250


74
00:05:16,250 --> 00:05:20,270
And then finally,
a last remark would

75
00:05:20,270 --> 00:05:25,010
be that CTSS and Multics,
which succeeded it,

76
00:05:25,010 --> 00:05:29,090
were precursors to the
contemporary systems

77
00:05:29,090 --> 00:05:31,340
and networks.

78
00:05:31,340 --> 00:05:39,710
And all of the interfaces and
ideas that have shown up now

79
00:05:39,710 --> 00:05:43,340
have set the stage
for personal computing

80
00:05:43,340 --> 00:05:46,380
and the Altos, which Butler
I'm sure will talk about.

81
00:05:46,380 --> 00:05:54,560
And the PC world that
we know today got

82
00:05:54,560 --> 00:05:56,300
started in
time-sharing, as people

83
00:05:56,300 --> 00:06:00,420
began to learn to interact
in a real time basis.

84
00:06:00,420 --> 00:06:06,500
It also led to things like Unix,
and the C programming language,

85
00:06:06,500 --> 00:06:09,260
and ultimately today
is being rediscovered

86
00:06:09,260 --> 00:06:10,820
in cloud computing.

87
00:06:10,820 --> 00:06:13,520


88
00:06:13,520 --> 00:06:15,430
Well thank you Corby.

89
00:06:15,430 --> 00:06:19,670
And if Corby was the consummate
system builder of the 1960s,

90
00:06:19,670 --> 00:06:21,970
I think Butler Lampson
certainly deserves

91
00:06:21,970 --> 00:06:24,040
that honor for the '70s.

92
00:06:24,040 --> 00:06:29,260
He was cited in 1992, when
he earned his Turing Award,

93
00:06:29,260 --> 00:06:31,780
for contributions to the
development of distributed

94
00:06:31,780 --> 00:06:34,840
personal computing
environments and the technology

95
00:06:34,840 --> 00:06:38,530
for their implementation,
workstations, networks,

96
00:06:38,530 --> 00:06:42,940
operating systems, programming
systems, displays, security,

97
00:06:42,940 --> 00:06:44,560
and document publishing.

98
00:06:44,560 --> 00:06:47,680
It's hard to know what
got left off of that list.

99
00:06:47,680 --> 00:06:48,460
Butler?

100
00:06:48,460 --> 00:06:50,500
I can tell you exactly
what got left off.

101
00:06:50,500 --> 00:06:55,680


102
00:06:55,680 --> 00:06:58,080
Spreadsheets and the
web got left off.

103
00:06:58,080 --> 00:07:02,320


104
00:07:02,320 --> 00:07:06,430
When Xerox started the computing
part of the Xerox Palo Alto

105
00:07:06,430 --> 00:07:10,780
Research Center in 1969,
their short-term motivation

106
00:07:10,780 --> 00:07:12,820
was that they had just
made a foolish mistake

107
00:07:12,820 --> 00:07:14,890
and bought a computer company.

108
00:07:14,890 --> 00:07:16,810
Their longer term
motivation was that they

109
00:07:16,810 --> 00:07:20,740
were able to see that
the light lens copying

110
00:07:20,740 --> 00:07:23,350
business, over which they
had a monopoly at the time,

111
00:07:23,350 --> 00:07:25,750
was going to be very
nice for a decade or two,

112
00:07:25,750 --> 00:07:27,730
but it was not going
to last forever.

113
00:07:27,730 --> 00:07:30,580
And the senior
management of Xerox

114
00:07:30,580 --> 00:07:33,285
thought that they ought
to prepare for the future,

115
00:07:33,285 --> 00:07:34,660
so they started
this lab and they

116
00:07:34,660 --> 00:07:37,950
told us to invent the
office of the future.

117
00:07:37,950 --> 00:07:41,470
Of course, they had no idea what
that meant, and neither did we.

118
00:07:41,470 --> 00:07:45,210
But it turned out to be a
very fruitful instruction

119
00:07:45,210 --> 00:07:47,560
to give us, because
based on that,

120
00:07:47,560 --> 00:07:52,470
we were able to develop
pretty much all of the things

121
00:07:52,470 --> 00:07:55,770
that you're now
accustomed to in the world

122
00:07:55,770 --> 00:07:59,640
of personal computing.

123
00:07:59,640 --> 00:08:06,610
Networking, and local
systems, and drawing programs,

124
00:08:06,610 --> 00:08:09,760
and document editing
and the first internet

125
00:08:09,760 --> 00:08:12,280
was actually built
at Xerox PARC.

126
00:08:12,280 --> 00:08:15,040
For some mysterious
reason, Xerox

127
00:08:15,040 --> 00:08:17,420
decided that we shouldn't
publish anything about it,

128
00:08:17,420 --> 00:08:19,810
so the people who did the
work never got much credit.

129
00:08:19,810 --> 00:08:21,850
But they did go to the
seminar at Stanford

130
00:08:21,850 --> 00:08:25,960
where Vint Cerf was developing
the internet protocols

131
00:08:25,960 --> 00:08:28,900
and they weren't allowed to
talk about the internal network,

132
00:08:28,900 --> 00:08:31,270
but they were allowed
to ask questions.

133
00:08:31,270 --> 00:08:32,780
And they asked a
lot of questions.

134
00:08:32,780 --> 00:08:35,289
And finally Vint looked at
one of them and he said,

135
00:08:35,289 --> 00:08:38,679
you've done this
before, haven't you?

136
00:08:38,679 --> 00:08:39,730
So we did all that stuff.

137
00:08:39,730 --> 00:08:42,309
We didn't do spreadsheets, that
would have been technically

138
00:08:42,309 --> 00:08:45,970
possible on the
equipment that we had

139
00:08:45,970 --> 00:08:48,940
but we had no personal
use for spreadsheets.

140
00:08:48,940 --> 00:08:52,398
Since, although their use has
expanded a lot more recently,

141
00:08:52,398 --> 00:08:54,190
originally they were
meant for accountants,

142
00:08:54,190 --> 00:08:55,600
and we didn't do accounting.

143
00:08:55,600 --> 00:08:58,420
And we didn't get along
very well with the people

144
00:08:58,420 --> 00:09:01,553
at PARC who did do accounting.

145
00:09:01,553 --> 00:09:03,220
So that's why we
didn't do spreadsheets.

146
00:09:03,220 --> 00:09:04,210
And the reason we
didn't do the web

147
00:09:04,210 --> 00:09:06,940
was that we didn't have a big
enough sandbox to play in.

148
00:09:06,940 --> 00:09:10,960
My theory about the reason that
the web came along when it did

149
00:09:10,960 --> 00:09:14,080
is that that was the first time
the internet was big enough

150
00:09:14,080 --> 00:09:18,410
to make it worthwhile for people
to actually post things on it.

151
00:09:18,410 --> 00:09:20,330
Anyway, we did invent
all this stuff,

152
00:09:20,330 --> 00:09:25,840
and it's pretty much the direct
predecessors of all the stuff

153
00:09:25,840 --> 00:09:28,060
that people use today.

154
00:09:28,060 --> 00:09:31,450
Even the first
quote laptop unquote

155
00:09:31,450 --> 00:09:33,625
was built at PARC
in the late 70s.

156
00:09:33,625 --> 00:09:36,180


157
00:09:36,180 --> 00:09:38,990
OK, thank you.

158
00:09:38,990 --> 00:09:43,580
Andy, you were awarded the
Turing Award in the year 2000

159
00:09:43,580 --> 00:09:46,130
in recognition of
fundamental contributions

160
00:09:46,130 --> 00:09:49,070
to the theory of
computation, including

161
00:09:49,070 --> 00:09:51,530
the complexity-based theory
of pseudorandom number

162
00:09:51,530 --> 00:09:56,580
generation, cryptography,
and communication complexity.

163
00:09:56,580 --> 00:09:58,410
Can you give us your
retrospective view

164
00:09:58,410 --> 00:10:01,890
of that work, and its
impact and your motivation?

165
00:10:01,890 --> 00:10:09,000
I think my Turing Award was
mainly due to the proposal

166
00:10:09,000 --> 00:10:14,140
of two concepts around 1980.

167
00:10:14,140 --> 00:10:22,870
These concepts would lead to
the creation of two subfields

168
00:10:22,870 --> 00:10:25,780
in theoretical computer
science, which over the years

169
00:10:25,780 --> 00:10:30,340
have shown a lot of
depth and applications.

170
00:10:30,340 --> 00:10:34,840
And the first of these is
communication complexity,

171
00:10:34,840 --> 00:10:39,670
which captures the amount of
communications needed when

172
00:10:39,670 --> 00:10:43,510
several computers want to
perform a joint computation

173
00:10:43,510 --> 00:10:46,250
based on their private inputs.

174
00:10:46,250 --> 00:10:51,500
And the second concept
that I proposed

175
00:10:51,500 --> 00:10:57,470
was about the perfect
pseudorandom number generator.

176
00:10:57,470 --> 00:11:04,280
Pseudorandom number generation
had been a very important

177
00:11:04,280 --> 00:11:08,420
subject in simulations
and in other fields,

178
00:11:08,420 --> 00:11:12,860
and the standard
statistical theory about it

179
00:11:12,860 --> 00:11:18,350
is that whether a pseudorandom
number generator is good or not

180
00:11:18,350 --> 00:11:20,060
depends on the application.

181
00:11:20,060 --> 00:11:22,670
And typically you
have some sort of

182
00:11:22,670 --> 00:11:25,580
mathematical, statistical
test to see whether it's good

183
00:11:25,580 --> 00:11:26,920
or not.

184
00:11:26,920 --> 00:11:30,310
Now with the advent
of cryptography,

185
00:11:30,310 --> 00:11:32,260
we have to be more
stringent, and we

186
00:11:32,260 --> 00:11:38,140
have to think about is there
a way so that we can define

187
00:11:38,140 --> 00:11:41,860
a good definition of
pseudorandom number generator

188
00:11:41,860 --> 00:11:46,840
so that it can satisfy the
very rigorous requirements.

189
00:11:46,840 --> 00:11:52,780
And so I proposed a
definition, which I think

190
00:11:52,780 --> 00:11:54,730
is pretty outrageous.

191
00:11:54,730 --> 00:12:00,760
I was asking the question, is it
possible to have a pseudorandom

192
00:12:00,760 --> 00:12:05,950
number generator that can
pass all the statistical tests

193
00:12:05,950 --> 00:12:07,570
that people can think of?

194
00:12:07,570 --> 00:12:09,560
And it actually turned
out to be the case.

195
00:12:09,560 --> 00:12:13,720
And with the help of
computational complexity,

196
00:12:13,720 --> 00:12:17,230
and myself and many
of my colleagues

197
00:12:17,230 --> 00:12:21,160
together, we have
created to make it work.

198
00:12:21,160 --> 00:12:24,520
And now in retrospect, I
think it probably should not

199
00:12:24,520 --> 00:12:27,940
have been a surprise
that concepts like this

200
00:12:27,940 --> 00:12:31,750
would lead to ramifications.

201
00:12:31,750 --> 00:12:34,540
The reason is that
around 1980, there

202
00:12:34,540 --> 00:12:39,670
were two seminal intellectual
trends that started

203
00:12:39,670 --> 00:12:41,360
to happen in computer science.

204
00:12:41,360 --> 00:12:52,060
One is the change from single
machine to networked setting,

205
00:12:52,060 --> 00:12:59,070
I think as pioneered by
people like Butler and others.

206
00:12:59,070 --> 00:13:03,400
So the cost consideration
would no longer

207
00:13:03,400 --> 00:13:06,550
be just the computation
time for one single machine,

208
00:13:06,550 --> 00:13:09,070
but actually you have to
take the communication

209
00:13:09,070 --> 00:13:11,250
cost into account.

210
00:13:11,250 --> 00:13:17,010
And that the second
seminal intellectual trend

211
00:13:17,010 --> 00:13:20,970
that was started at time,
is that at that time

212
00:13:20,970 --> 00:13:26,790
the proposal of the
public-key cryptography

213
00:13:26,790 --> 00:13:30,330
led farsighted people to
think about in the future,

214
00:13:30,330 --> 00:13:32,790
maybe we will have
a big network and we

215
00:13:32,790 --> 00:13:35,820
can do electronic commerce,
and cryptography will

216
00:13:35,820 --> 00:13:39,030
be very useful and important.

217
00:13:39,030 --> 00:13:46,020
And now when you think
about cryptography,

218
00:13:46,020 --> 00:13:51,630
the traditional theory was the
Shannon Theory of Cryptography.

219
00:13:51,630 --> 00:13:56,010
But Shannon's theory
was very rigorous.

220
00:13:56,010 --> 00:13:58,590
However, it's
perhaps too rigorous.

221
00:13:58,590 --> 00:14:02,250
If you insist on that security
then you practically cannot do

222
00:14:02,250 --> 00:14:04,120
any electronic commerce.

223
00:14:04,120 --> 00:14:09,510
And so I think at that time,
near the end of the 1970s

224
00:14:09,510 --> 00:14:12,570
and the beginning of
the 1980s, many of us

225
00:14:12,570 --> 00:14:18,660
were struggling to try to break
away from the Shannon cage.

226
00:14:18,660 --> 00:14:23,430
And instead, we would embrace
the computational complexity

227
00:14:23,430 --> 00:14:27,840
to become our savior, so that
we can construct frameworks

228
00:14:27,840 --> 00:14:30,330
so that cryptography
will be rigorous

229
00:14:30,330 --> 00:14:32,520
and will be meaningful.

230
00:14:32,520 --> 00:14:37,880
And so I think that
given those two trends,

231
00:14:37,880 --> 00:14:40,190
I think that it should
have come as no surprise

232
00:14:40,190 --> 00:14:44,360
that those of us who were lucky
enough in those early days, who

233
00:14:44,360 --> 00:14:47,690
started to think
about such issues,

234
00:14:47,690 --> 00:14:53,890
would come up with concepts that
would become very important.

235
00:14:53,890 --> 00:14:59,910
And perhaps I should
explain why I'm here.

236
00:14:59,910 --> 00:15:02,800
I think all my
distinguished friends here

237
00:15:02,800 --> 00:15:05,380
are MIT faculty members.

238
00:15:05,380 --> 00:15:09,130
I think that firstly,
I was a faculty at MIT

239
00:15:09,130 --> 00:15:14,710
during 1975 to 1976.

240
00:15:14,710 --> 00:15:18,970
And secondly, I
think my work has

241
00:15:18,970 --> 00:15:25,280
been influenced by a lot of the
MIT people who are here today.

242
00:15:25,280 --> 00:15:29,680
I think, for example, in
the communication complexity

243
00:15:29,680 --> 00:15:33,500
that Harold Abelson and the
pseudorandom number generation

244
00:15:33,500 --> 00:15:41,140
by Ron Rivest, Shafi
Goldwasser, and Sylvia McCauley.

245
00:15:41,140 --> 00:15:47,530
And I think there's
an interesting story

246
00:15:47,530 --> 00:15:50,110
that I think--

247
00:15:50,110 --> 00:15:52,360
Ron if you'll forgive me--
that I will just repeat it.

248
00:15:52,360 --> 00:15:58,300
I think Ron always joked
with me that I missed

249
00:15:58,300 --> 00:16:03,010
the boat because I
left MIT in 1976,

250
00:16:03,010 --> 00:16:07,610
otherwise I would perhaps
be able to get a Turing

251
00:16:07,610 --> 00:16:09,980
Award together with them.

252
00:16:09,980 --> 00:16:13,320
But fortunately, I
went to Stanford,

253
00:16:13,320 --> 00:16:16,060
and then later on
had the opportunity

254
00:16:16,060 --> 00:16:18,100
to mix with the Berkeley crowd.

255
00:16:18,100 --> 00:16:23,980
And so I guess I got into
cryptography on the West Coast,

256
00:16:23,980 --> 00:16:29,290
and I guess everything is
probably preordained after all.

257
00:16:29,290 --> 00:16:32,170
Thank you.

258
00:16:32,170 --> 00:16:35,140
Thank you, Andy.

259
00:16:35,140 --> 00:16:39,340
Ron, you were awarded
the Turing Award in 2002

260
00:16:39,340 --> 00:16:43,750
for ingenious contribution for
making public-key cryptography

261
00:16:43,750 --> 00:16:45,820
useful in practice.

262
00:16:45,820 --> 00:16:47,830
Your retrospective view?

263
00:16:47,830 --> 00:16:48,560
Yes, sure.

264
00:16:48,560 --> 00:16:51,190
So Andy talked about
some of the themes

265
00:16:51,190 --> 00:16:52,520
that were around in those days.

266
00:16:52,520 --> 00:16:56,500
So our work was inspired by the
Diffie-Hellman paper talking

267
00:16:56,500 --> 00:16:59,380
about public-key cryptography,
and it asked the question

268
00:16:59,380 --> 00:17:02,500
whether public-key
cryptography was possible.

269
00:17:02,500 --> 00:17:06,160
And with Len Edelman
and Adi Shamir,

270
00:17:06,160 --> 00:17:08,050
who are also members of--

271
00:17:08,050 --> 00:17:10,450
not CSAIL, but LCS at the
time-- although members

272
00:17:10,450 --> 00:17:12,079
of the math department,
and that shows

273
00:17:12,079 --> 00:17:14,079
the strength of the
inter-departmental character

274
00:17:14,079 --> 00:17:18,490
of the enterprise
of computer science.

275
00:17:18,490 --> 00:17:21,381
We came up with a proposal
known as the RSA scheme.

276
00:17:21,381 --> 00:17:22,839
Now I guess if
you'd stayed around,

277
00:17:22,839 --> 00:17:25,674
it might have been RSAY or--

278
00:17:25,674 --> 00:17:26,364
Or YRSA.

279
00:17:26,364 --> 00:17:28,869


280
00:17:28,869 --> 00:17:33,040
YRSA or whatever, yeah.

281
00:17:33,040 --> 00:17:35,450
So, anyway, we came
up with a proposal,

282
00:17:35,450 --> 00:17:38,230
which has so far stood
the test of time,

283
00:17:38,230 --> 00:17:40,660
based on the difficulty of
factoring the product of two

284
00:17:40,660 --> 00:17:43,390
large prime numbers and doing
some modular exponentiation

285
00:17:43,390 --> 00:17:46,150
for encryption and
for decryption.

286
00:17:46,150 --> 00:17:48,910
We still don't know, in the
end, whether it's secure or not.

287
00:17:48,910 --> 00:17:52,360
We have many open problems in
the field of computer science

288
00:17:52,360 --> 00:17:54,520
and theory, and showing
that a problem is hard

289
00:17:54,520 --> 00:17:57,768
remains one of the most
challenging problems we have.

290
00:17:57,768 --> 00:17:59,560
And we don't know
whether specific problems

291
00:17:59,560 --> 00:18:03,430
like factoring are hard,
or particular cryptosystems

292
00:18:03,430 --> 00:18:04,240
are hard.

293
00:18:04,240 --> 00:18:05,940
But cryptography
has thrived, and one

294
00:18:05,940 --> 00:18:07,690
of the impacts that
our work had, I think,

295
00:18:07,690 --> 00:18:10,810
is to really help establish
the character of the questions

296
00:18:10,810 --> 00:18:11,710
that people ask.

297
00:18:11,710 --> 00:18:14,320
And the kind of questions
you can ask and answer

298
00:18:14,320 --> 00:18:16,570
is can you design
cryptosystems so

299
00:18:16,570 --> 00:18:19,150
that the difficulty of
breaking that cryptosystem

300
00:18:19,150 --> 00:18:22,630
is closely related or
identical to some hard problem

301
00:18:22,630 --> 00:18:24,010
that you believe to be hard.

302
00:18:24,010 --> 00:18:26,620
You make explicit assumptions
that factoring is hard,

303
00:18:26,620 --> 00:18:29,043
or discrete logarithm is
hard, or doing discrete log

304
00:18:29,043 --> 00:18:31,210
and elliptic curves is hard,
or something like that.

305
00:18:31,210 --> 00:18:32,627
And the field has
really blossomed

306
00:18:32,627 --> 00:18:35,140
in a very nice way, in
spite of our inability

307
00:18:35,140 --> 00:18:37,630
to show that these particular
problems are hard, because we

308
00:18:37,630 --> 00:18:39,843
can make very crisp assumptions
saying if you assume

309
00:18:39,843 --> 00:18:41,260
that these problems
are hard, then

310
00:18:41,260 --> 00:18:43,400
you can achieve these
cryptographic objectives.

311
00:18:43,400 --> 00:18:46,150
And so the field has really
blossomed in a wonderful way

312
00:18:46,150 --> 00:18:48,490
and achieved a lot of
paradoxical things.

313
00:18:48,490 --> 00:18:53,200
I think the public-key
work itself demonstrates

314
00:18:53,200 --> 00:18:56,920
the paradoxical character
that cryptography can have,

315
00:18:56,920 --> 00:18:59,620
that you can ask something which
seems like you shouldn't be

316
00:18:59,620 --> 00:19:02,380
able to do it, and
yet you can at the

317
00:19:02,380 --> 00:19:04,683
end with the clever use
of the tools at hand.

318
00:19:04,683 --> 00:19:06,850
And that's a theme that
appears over and over again,

319
00:19:06,850 --> 00:19:09,203
is that by making the
right assumptions, by being

320
00:19:09,203 --> 00:19:10,870
careful about your
reductions and asking

321
00:19:10,870 --> 00:19:13,928
what appear to be
paradoxical questions,

322
00:19:13,928 --> 00:19:15,970
paradoxical set of
requirements, you can actually

323
00:19:15,970 --> 00:19:17,560
achieve some of these things.

324
00:19:17,560 --> 00:19:18,230
I'll stop there.

325
00:19:18,230 --> 00:19:18,730
OK.

326
00:19:18,730 --> 00:19:19,840
Thank you, Ron.

327
00:19:19,840 --> 00:19:24,470
And Barbara, your
2008 Turing Award

328
00:19:24,470 --> 00:19:27,700
cited contributions to practical
and theoretical foundations

329
00:19:27,700 --> 00:19:30,640
for programming language and
system design, especially

330
00:19:30,640 --> 00:19:33,520
related to data abstraction,
fault tolerance,

331
00:19:33,520 --> 00:19:36,550
and distributed computing.

332
00:19:36,550 --> 00:19:37,050
OK.

333
00:19:37,050 --> 00:19:39,480
Well, Corby started by
mentioning John McCarthy,

334
00:19:39,480 --> 00:19:42,510
so I wanted to say that
John McCarthy was my thesis

335
00:19:42,510 --> 00:19:44,530
advisor at Stanford.

336
00:19:44,530 --> 00:19:46,240
And I was working
in AI at the time,

337
00:19:46,240 --> 00:19:49,410
but as soon as I got my PhD I
switched into computer systems.

338
00:19:49,410 --> 00:19:51,600
I went to work at the
Miter Corporation,

339
00:19:51,600 --> 00:19:53,820
and while I was
at Miter I started

340
00:19:53,820 --> 00:19:55,530
to work on programming
methodology,

341
00:19:55,530 --> 00:19:58,300
because there was a huge
concern at the time.

342
00:19:58,300 --> 00:20:01,410
And of course, there's still
a concern about what they then

343
00:20:01,410 --> 00:20:03,330
called the software crisis.

344
00:20:03,330 --> 00:20:04,800
By which they meant
we had no idea

345
00:20:04,800 --> 00:20:06,258
how to build software
that actually

346
00:20:06,258 --> 00:20:08,070
did what it was supposed to do.

347
00:20:08,070 --> 00:20:10,800
And I found this
problem very intriguing,

348
00:20:10,800 --> 00:20:12,870
and so I was thinking about it.

349
00:20:12,870 --> 00:20:16,890
And I moved to MIT
in 1972, and this

350
00:20:16,890 --> 00:20:18,700
was the problem that
was consuming me,

351
00:20:18,700 --> 00:20:22,080
what would be the right
way to organize software

352
00:20:22,080 --> 00:20:24,600
so that we could actually
control the complexity

353
00:20:24,600 --> 00:20:28,260
and reason about the
resulting programs.

354
00:20:28,260 --> 00:20:30,450
And I had in mind
a particular way

355
00:20:30,450 --> 00:20:33,060
of organizing programs,
which I referred to

356
00:20:33,060 --> 00:20:35,160
as the multi-operation module.

357
00:20:35,160 --> 00:20:38,260
So the idea was that
I had a big black box,

358
00:20:38,260 --> 00:20:40,980
and inside there was
complicated data structures

359
00:20:40,980 --> 00:20:45,000
and very complex code, but it
was all hidden inside the box.

360
00:20:45,000 --> 00:20:47,310
And the box provided
an interface

361
00:20:47,310 --> 00:20:50,108
that consisted of a number of
operations, and outside the box

362
00:20:50,108 --> 00:20:52,650
you could call these operations
but that was the only way you

363
00:20:52,650 --> 00:20:54,240
could interact with the box.

364
00:20:54,240 --> 00:20:56,250
And this seemed to me
like a really good way

365
00:20:56,250 --> 00:20:57,390
to organize programs.

366
00:20:57,390 --> 00:20:59,640
And there were other people
that had similar ideas,

367
00:20:59,640 --> 00:21:01,960
like Dave Parnas
at the same time,

368
00:21:01,960 --> 00:21:05,190
but I felt that it was very
hard to communicate this idea

369
00:21:05,190 --> 00:21:06,930
to normal programmers.

370
00:21:06,930 --> 00:21:11,320
And so, in the fall of 1972
shortly after I got to MIT,

371
00:21:11,320 --> 00:21:14,130
I had this wonderful aha moment.

372
00:21:14,130 --> 00:21:16,440
And very few people
are lucky enough

373
00:21:16,440 --> 00:21:17,580
to have a moment like this.

374
00:21:17,580 --> 00:21:19,710
And what happened to me
was I all of a sudden

375
00:21:19,710 --> 00:21:21,960
saw that I could
relate this idea

376
00:21:21,960 --> 00:21:24,630
of a very abstract,
multi-operation module

377
00:21:24,630 --> 00:21:25,830
to data types.

378
00:21:25,830 --> 00:21:28,260
And I felt that
making that connection

379
00:21:28,260 --> 00:21:31,050
was going to be really
beneficial, because programmers

380
00:21:31,050 --> 00:21:32,340
understood data types.

381
00:21:32,340 --> 00:21:34,500
They programmed with
data types all the time.

382
00:21:34,500 --> 00:21:38,400
And the connection was that
the module was a data object,

383
00:21:38,400 --> 00:21:40,860
the operations were the
operations of the object

384
00:21:40,860 --> 00:21:43,560
that you use to manipulate
it, and people already

385
00:21:43,560 --> 00:21:45,360
use things like arrays
and they understood

386
00:21:45,360 --> 00:21:47,175
that you could add
something to array,

387
00:21:47,175 --> 00:21:49,200
remove something from
the array, and they also

388
00:21:49,200 --> 00:21:51,390
understood that they
thought about these things

389
00:21:51,390 --> 00:21:55,170
abstractly and did not have to
actually understand how things

390
00:21:55,170 --> 00:21:57,020
were implemented underneath.

391
00:21:57,020 --> 00:21:59,968
And so I had this idea
of abstract data types.

392
00:21:59,968 --> 00:22:02,010
And then, of course, you
know you're lucky enough

393
00:22:02,010 --> 00:22:03,802
to have an idea like
that, but of course it

394
00:22:03,802 --> 00:22:04,950
doesn't stop there.

395
00:22:04,950 --> 00:22:08,070
I spent several years
trying to figure out

396
00:22:08,070 --> 00:22:11,640
what this idea really meant,
how did you define it formally,

397
00:22:11,640 --> 00:22:14,970
what was an abstract
data type conceptually,

398
00:22:14,970 --> 00:22:17,250
how did you specify
them, how did you reason

399
00:22:17,250 --> 00:22:20,160
about their correctness,
how did you design programs

400
00:22:20,160 --> 00:22:22,590
using the idea, and so on.

401
00:22:22,590 --> 00:22:25,750
Around about 1980 I switched
into distributed systems,

402
00:22:25,750 --> 00:22:29,310
so that was the other
part of the award.

403
00:22:29,310 --> 00:22:32,428
And so that was just
an example of what

404
00:22:32,428 --> 00:22:34,470
I was doing in my career,
I was opportunistically

405
00:22:34,470 --> 00:22:36,220
switching from one
field to another

406
00:22:36,220 --> 00:22:38,550
as I saw really
important problems.

407
00:22:38,550 --> 00:22:42,480
And when working in distributed
systems, where I work today,

408
00:22:42,480 --> 00:22:45,070
I got very interested in
replication and so forth.

409
00:22:45,070 --> 00:22:49,410
So I've had a lovely career,
and I was lucky enough

410
00:22:49,410 --> 00:22:51,000
to have some good ideas.

411
00:22:51,000 --> 00:22:54,220
And that's where I ended up.

412
00:22:54,220 --> 00:22:55,440
Well, thanks Barbara.

413
00:22:55,440 --> 00:22:57,690
I have here a
handful of questions

414
00:22:57,690 --> 00:23:00,210
that have been
submitted by attendees,

415
00:23:00,210 --> 00:23:02,610
and I'm unlikely to be
able to get to them all,

416
00:23:02,610 --> 00:23:06,180
but let me start
with several ones

417
00:23:06,180 --> 00:23:08,070
that I think should
be easy that deal

418
00:23:08,070 --> 00:23:11,170
with the institution of
the Turing Award itself.

419
00:23:11,170 --> 00:23:16,050
The first one is,
in 25 words or less,

420
00:23:16,050 --> 00:23:18,907
how did the Turing
Award change your life?

421
00:23:18,907 --> 00:23:20,490
Would anybody like
to respond to that?

422
00:23:20,490 --> 00:23:23,650


423
00:23:23,650 --> 00:23:24,150
OK.

424
00:23:24,150 --> 00:23:26,660
Well if you decline
responding to that,

425
00:23:26,660 --> 00:23:30,330
let me get to one that's even
shorter and considerably more

426
00:23:30,330 --> 00:23:32,010
blunt.

427
00:23:32,010 --> 00:23:34,230
Are Turing Awards
really a good idea?

428
00:23:34,230 --> 00:23:36,810


429
00:23:36,810 --> 00:23:39,270
I'm guessing this was
submitted by somebody who

430
00:23:39,270 --> 00:23:40,570
doesn't have a Turing Award.

431
00:23:40,570 --> 00:23:43,380


432
00:23:43,380 --> 00:23:47,280
If I had to guess, I would be
able to guess a one-bit answer

433
00:23:47,280 --> 00:23:50,400
that each of the
awardees would give, but.

434
00:23:50,400 --> 00:23:52,185
Is this meeting a good idea?

435
00:23:52,185 --> 00:23:53,400
Is anything a good idea?

436
00:23:53,400 --> 00:23:56,850
I think they're very
closely related questions.

437
00:23:56,850 --> 00:24:02,040
Perhaps, but I make the
assumption that none of you

438
00:24:02,040 --> 00:24:05,280
did the seminal work
that you did specifically

439
00:24:05,280 --> 00:24:07,240
in order to win a Turing Award.

440
00:24:07,240 --> 00:24:07,740
OK.

441
00:24:07,740 --> 00:24:10,770
So if it didn't have that
effect on your behavior,

442
00:24:10,770 --> 00:24:13,530
did it in fact have any
effect on your behavior?

443
00:24:13,530 --> 00:24:17,340
As a social institution,
is it a positive thing?

444
00:24:17,340 --> 00:24:19,350
Anybody want to comment on that?

445
00:24:19,350 --> 00:24:23,550
Well, it's not for individuals.

446
00:24:23,550 --> 00:24:25,230
You don't do research
because you're

447
00:24:25,230 --> 00:24:26,730
thinking about the
Turing Award, you

448
00:24:26,730 --> 00:24:28,650
do research because
you find things

449
00:24:28,650 --> 00:24:30,990
that you are fascinated
with that are interesting,

450
00:24:30,990 --> 00:24:32,460
useful to work on.

451
00:24:32,460 --> 00:24:34,530
It's probably good
for the field,

452
00:24:34,530 --> 00:24:38,730
because it gives us a history,
it recognizes contributions,

453
00:24:38,730 --> 00:24:40,110
it gives people
a way of thinking

454
00:24:40,110 --> 00:24:41,670
about the history of the field.

455
00:24:41,670 --> 00:24:43,680
Of course, it's not perfect.

456
00:24:43,680 --> 00:24:45,630
Because the people
who get the award,

457
00:24:45,630 --> 00:24:47,550
they have to be
nominated, the nomination

458
00:24:47,550 --> 00:24:51,030
has to be phrased properly, the
committee has to be receptive,

459
00:24:51,030 --> 00:24:53,580
and so clearly there are
going to be people who didn't

460
00:24:53,580 --> 00:24:55,410
get the award who deserved it.

461
00:24:55,410 --> 00:24:58,230
And there might be people
that got the award who

462
00:24:58,230 --> 00:25:00,570
didn't deserve it, who can say.

463
00:25:00,570 --> 00:25:03,150
But I think if you think
about it, sort of on average,

464
00:25:03,150 --> 00:25:06,520
it's a useful thing
for the field.

465
00:25:06,520 --> 00:25:11,370
Is it fair to have the model
that, like tenure, people who

466
00:25:11,370 --> 00:25:12,780
have it think it's a good idea?

467
00:25:12,780 --> 00:25:15,605


468
00:25:15,605 --> 00:25:16,980
Well, let me move
on to something

469
00:25:16,980 --> 00:25:18,090
that's closely related.

470
00:25:18,090 --> 00:25:21,203
To also answer that, if I may,
I think I do have an opinion.

471
00:25:21,203 --> 00:25:22,620
I think the Turing
Award is great,

472
00:25:22,620 --> 00:25:24,660
and it adds some
value to the field,

473
00:25:24,660 --> 00:25:26,460
but I think more
important to the field

474
00:25:26,460 --> 00:25:29,337
are awards going to younger
people before their promotion

475
00:25:29,337 --> 00:25:30,420
and tenure and so on, too.

476
00:25:30,420 --> 00:25:34,350
I think that's a critical thing,
to recognize accomplishment

477
00:25:34,350 --> 00:25:36,780
early in people's career,
and the Turing Award

478
00:25:36,780 --> 00:25:39,470
tends not to do that.

479
00:25:39,470 --> 00:25:40,550
OK.

480
00:25:40,550 --> 00:25:43,280
A related question
is, can you think

481
00:25:43,280 --> 00:25:46,010
of work that has
already been done

482
00:25:46,010 --> 00:25:49,320
that deserves a Turing Award
and hasn't yet received one?

483
00:25:49,320 --> 00:25:51,560
For example, would you like
to make any predictions

484
00:25:51,560 --> 00:25:53,900
as to the 2012
Turing Award winner?

485
00:25:53,900 --> 00:25:56,510


486
00:25:56,510 --> 00:25:59,270
You, after all, are
pundits, and you

487
00:25:59,270 --> 00:26:02,280
need to exercise your
crystal ball a little bit.

488
00:26:02,280 --> 00:26:03,520
I'll step into that one.

489
00:26:03,520 --> 00:26:06,620


490
00:26:06,620 --> 00:26:09,770
I'm not sure I know
exactly who to credit,

491
00:26:09,770 --> 00:26:12,830
but I think there
probably is not

492
00:26:12,830 --> 00:26:14,420
a person in this
room that doesn't

493
00:26:14,420 --> 00:26:20,230
recognize the profundity of
the contribution of Google.

494
00:26:20,230 --> 00:26:21,430
It's become a verb.

495
00:26:21,430 --> 00:26:24,400


496
00:26:24,400 --> 00:26:26,710
People have thrown away
their encyclopedias,

497
00:26:26,710 --> 00:26:34,360
and it's a wonderful
contribution to our well-being

498
00:26:34,360 --> 00:26:35,920
and so forth.

499
00:26:35,920 --> 00:26:41,410
It had a predecessor, which
isn't well known, AltaVista.

500
00:26:41,410 --> 00:26:45,400
Butler maybe can
talk about that.

501
00:26:45,400 --> 00:26:49,330
Google has succeeded for
reasons I don't fully

502
00:26:49,330 --> 00:26:57,000
understand myself, but
it's miraculously fast.

503
00:26:57,000 --> 00:26:59,122
It's fast because they
have a lot of money.

504
00:26:59,122 --> 00:27:01,330
And they're clever, too,
but it takes a lot of money.

505
00:27:01,330 --> 00:27:02,460
And the reason they
have a lot of money

506
00:27:02,460 --> 00:27:04,800
is that they figured out what
very few people did, which

507
00:27:04,800 --> 00:27:07,110
is that you could
make a lot of money

508
00:27:07,110 --> 00:27:09,480
out of targeted advertising.

509
00:27:09,480 --> 00:27:11,350
Well aren't there
similar institutions?

510
00:27:11,350 --> 00:27:12,943
I'm thinking of
things like Wikipedia,

511
00:27:12,943 --> 00:27:14,610
that doesn't have a
lot of money, that's

512
00:27:14,610 --> 00:27:18,000
also had a significant impact.

513
00:27:18,000 --> 00:27:19,935
Yes, I would put that
in the same ballpark.

514
00:27:19,935 --> 00:27:23,800


515
00:27:23,800 --> 00:27:25,750
So money isn't essential.

516
00:27:25,750 --> 00:27:26,313
OK.

517
00:27:26,313 --> 00:27:27,730
Whether money is
essential depends

518
00:27:27,730 --> 00:27:29,870
on what it is you're
trying to accomplish.

519
00:27:29,870 --> 00:27:30,370
Perhaps.

520
00:27:30,370 --> 00:27:33,850


521
00:27:33,850 --> 00:27:36,700
For going to the moon,
you need a lot of money.

522
00:27:36,700 --> 00:27:40,600
To index the internet,
you need a lot of money.

523
00:27:40,600 --> 00:27:41,110
OK.

524
00:27:41,110 --> 00:27:42,820
Let me move on to
another question, which

525
00:27:42,820 --> 00:27:45,400
may be viewed as
superficial or profound

526
00:27:45,400 --> 00:27:47,170
depending on your
view, but it should

527
00:27:47,170 --> 00:27:49,660
be easy to answer it actually.

528
00:27:49,660 --> 00:27:52,780
I'll read it in its entirety.

529
00:27:52,780 --> 00:27:55,750
It's addressed to
each of the panelists.

530
00:27:55,750 --> 00:28:01,610
And the question is this, do
you have a Facebook account--

531
00:28:01,610 --> 00:28:06,520
I haven't finished-- a
Twitter account, and why not?

532
00:28:06,520 --> 00:28:07,020
So.

533
00:28:07,020 --> 00:28:11,590


534
00:28:11,590 --> 00:28:12,668
So Barbara, fess up.

535
00:28:12,668 --> 00:28:13,960
Do you have a Facebook account?

536
00:28:13,960 --> 00:28:15,340
Absolutely not.

537
00:28:15,340 --> 00:28:16,300
A Twitter account?

538
00:28:16,300 --> 00:28:18,850
Neither Facebook nor Twitter.

539
00:28:18,850 --> 00:28:20,260
Do you have an email account?

540
00:28:20,260 --> 00:28:22,480
Of course I have
an email account.

541
00:28:22,480 --> 00:28:26,870
I'm not particularly interested
in that kind of exposure.

542
00:28:26,870 --> 00:28:29,560
I'm not really interested in
following people or having

543
00:28:29,560 --> 00:28:33,700
them follow me, but I'm a
person of my generation.

544
00:28:33,700 --> 00:28:37,930
And if I were 40 years
younger I would probably

545
00:28:37,930 --> 00:28:40,060
have a completely
different answer, because I

546
00:28:40,060 --> 00:28:41,290
do think it's generational.

547
00:28:41,290 --> 00:28:43,120
We grew up in a
certain way, we're

548
00:28:43,120 --> 00:28:46,180
used to a certain way of
living, and you sort of

549
00:28:46,180 --> 00:28:48,340
settle into a rut.

550
00:28:48,340 --> 00:28:49,453
Yeah.

551
00:28:49,453 --> 00:28:52,120
I do think, though, that there's
something quite dangerous going

552
00:28:52,120 --> 00:28:52,912
on in the internet.

553
00:28:52,912 --> 00:28:56,110
And Facebook is a good example
of that, because of the way

554
00:28:56,110 --> 00:28:59,230
people expose themselves,
not understanding how

555
00:28:59,230 --> 00:29:00,760
they might regret that later.

556
00:29:00,760 --> 00:29:04,330
And I'm keenly aware of the
dangers of the internet.

557
00:29:04,330 --> 00:29:06,940
And thank God for Ron,
and all the crypto,

558
00:29:06,940 --> 00:29:09,580
so I can at least
do online banking,

559
00:29:09,580 --> 00:29:11,590
but I'm very careful
even with email

560
00:29:11,590 --> 00:29:14,050
and what I will send because--

561
00:29:14,050 --> 00:29:17,420
and you think politicians
would learn this after a while.

562
00:29:17,420 --> 00:29:20,350
Because everything you do
like this is exposable.

563
00:29:20,350 --> 00:29:22,800
And I just don't
care to be exposed.

564
00:29:22,800 --> 00:29:25,040
So actually, I do have
a Facebook account.

565
00:29:25,040 --> 00:29:26,350
I do have a Twitter account.

566
00:29:26,350 --> 00:29:28,308
I make sure to log in to
Facebook at least once

567
00:29:28,308 --> 00:29:33,800
a quarter, Twitter at
least once a decade.

568
00:29:33,800 --> 00:29:36,040
OK, so my follow on
question for you, Ron,

569
00:29:36,040 --> 00:29:39,250
is what are you going to do
about the 600 friend requests

570
00:29:39,250 --> 00:29:41,550
that you get as a result
of that announcement?

571
00:29:41,550 --> 00:29:43,735
No promises.

572
00:29:43,735 --> 00:29:44,235
OK.

573
00:29:44,235 --> 00:29:47,590


574
00:29:47,590 --> 00:29:50,620
It's sometimes argued that all
branches of computer science

575
00:29:50,620 --> 00:29:53,470
have enjoyed success because
of the remarkable exponential

576
00:29:53,470 --> 00:29:58,010
growth, the Moore's law
growth, for over five decades.

577
00:29:58,010 --> 00:30:01,270
And so an interesting
question is,

578
00:30:01,270 --> 00:30:04,990
would your work be as
celebrated as it is now?

579
00:30:04,990 --> 00:30:06,880
Would you even
have done it if it

580
00:30:06,880 --> 00:30:09,970
hadn't been for this
kind of Moore's law

581
00:30:09,970 --> 00:30:13,480
growth, this exponential growth,
and the impact of computers

582
00:30:13,480 --> 00:30:14,420
themselves?

583
00:30:14,420 --> 00:30:17,560
And there is kind of a rising
tide lifts all boats argument,

584
00:30:17,560 --> 00:30:20,590
that maybe good
research has turned

585
00:30:20,590 --> 00:30:26,290
into great research because
of the economic impact of all

586
00:30:26,290 --> 00:30:28,350
of this growth.

587
00:30:28,350 --> 00:30:29,460
Certainly.

588
00:30:29,460 --> 00:30:32,360
It's certainly true for most.

589
00:30:32,360 --> 00:30:33,360
It's certainly true for?

590
00:30:33,360 --> 00:30:35,790
For most practical systems work.

591
00:30:35,790 --> 00:30:37,800
We couldn't have
done what we did it

592
00:30:37,800 --> 00:30:40,260
at PARC five years earlier.

593
00:30:40,260 --> 00:30:44,820
Just with the hardware, would
have been impossible to build.

594
00:30:44,820 --> 00:30:47,870
Were you excited about building
that hardware in anticipation

595
00:30:47,870 --> 00:30:49,620
that someday it would
become cheap enough?

596
00:30:49,620 --> 00:30:49,935
Absolutely.

597
00:30:49,935 --> 00:30:51,643
That people could put
it in their pockets

598
00:30:51,643 --> 00:30:52,700
and have it at home?

599
00:30:52,700 --> 00:30:53,640
Absolutely.

600
00:30:53,640 --> 00:30:58,320
I wrote a guest editorial
on this subject in 1972,

601
00:30:58,320 --> 00:31:00,400
it was crystal clear
that that would happen.

602
00:31:00,400 --> 00:31:04,010
But I think it's fair
to say that was only--

603
00:31:04,010 --> 00:31:06,510
I suppose from the point of
view of the welfare of the Xerox

604
00:31:06,510 --> 00:31:09,570
corporation, that
was quite essential.

605
00:31:09,570 --> 00:31:11,910
We would probably have
tried to do the work anyway,

606
00:31:11,910 --> 00:31:13,930
even if we couldn't have
foreseen that future.

607
00:31:13,930 --> 00:31:15,500
But we certainly did foresee it.

608
00:31:15,500 --> 00:31:16,000
Yeah.

609
00:31:16,000 --> 00:31:16,500
Yeah.

610
00:31:16,500 --> 00:31:18,210
I think in backing
up what Butler

611
00:31:18,210 --> 00:31:21,750
said, during the '70s and
'80s, everybody working

612
00:31:21,750 --> 00:31:24,270
in any central area
of computer science

613
00:31:24,270 --> 00:31:27,000
was keenly aware of
the fact that computers

614
00:31:27,000 --> 00:31:28,380
were getting faster and faster.

615
00:31:28,380 --> 00:31:31,290
And so you could envision
doing things today

616
00:31:31,290 --> 00:31:33,660
that weren't going to
be practical today,

617
00:31:33,660 --> 00:31:35,760
but you were laying the
framework for something

618
00:31:35,760 --> 00:31:39,240
that would be practical a
little bit down the line.

619
00:31:39,240 --> 00:31:42,070
I think even for
theorists, I think

620
00:31:42,070 --> 00:31:45,790
that it's very important
to do theory problems that

621
00:31:45,790 --> 00:31:48,130
are related to technology.

622
00:31:48,130 --> 00:31:51,100
Somehow the advances
of technology

623
00:31:51,100 --> 00:31:56,240
bring out a lot of really
interesting problems,

624
00:31:56,240 --> 00:31:58,820
so that I don't know
whether it's a coincidence

625
00:31:58,820 --> 00:32:02,450
or there is a correlation
that somehow things

626
00:32:02,450 --> 00:32:06,440
that have very heavy
technological content

627
00:32:06,440 --> 00:32:09,570
also have very rich
theoretical content.

628
00:32:09,570 --> 00:32:12,980
And I think that I was
pretty lucky in getting

629
00:32:12,980 --> 00:32:18,620
into this field and I
have been quite interested

630
00:32:18,620 --> 00:32:22,920
in combinatorial problems
like [INAUDIBLE] problem.

631
00:32:22,920 --> 00:32:27,020
And all those problems
were very fascinating to me

632
00:32:27,020 --> 00:32:29,510
at the time when I
entered computer science.

633
00:32:29,510 --> 00:32:34,220
And I couldn't really
decide whether to work

634
00:32:34,220 --> 00:32:36,920
in combinatorial problems
or doing something else,

635
00:32:36,920 --> 00:32:38,360
like computer science.

636
00:32:38,360 --> 00:32:41,030
And at some point,
it suddenly hit me

637
00:32:41,030 --> 00:32:44,510
that there were so many
interesting mathematical

638
00:32:44,510 --> 00:32:48,650
problems in the world
that how do you decide

639
00:32:48,650 --> 00:32:50,270
which problems to work on?

640
00:32:50,270 --> 00:32:55,610
And computer science, being a
very rapidly advancing field,

641
00:32:55,610 --> 00:32:57,440
kind of breaks the symmetry.

642
00:32:57,440 --> 00:33:01,880
Somehow it singles
out problems that

643
00:33:01,880 --> 00:33:06,510
is different from the typical
mathematical problems,

644
00:33:06,510 --> 00:33:10,940
and I thought that
was a good choice.

645
00:33:10,940 --> 00:33:13,790
And I still don't understand
whether there's a correlation,

646
00:33:13,790 --> 00:33:16,890
but somehow it seems so.

647
00:33:16,890 --> 00:33:20,460
Well, to put it
more indelicately,

648
00:33:20,460 --> 00:33:24,150
was there a bit of
technological opportunism

649
00:33:24,150 --> 00:33:27,130
in your career choice, then?

650
00:33:27,130 --> 00:33:27,910
Not really.

651
00:33:27,910 --> 00:33:35,080
I think that
actually I first got

652
00:33:35,080 --> 00:33:40,370
into the theoretical computer
science from Duncan notebooks.

653
00:33:40,370 --> 00:33:46,720
And so I think that when I
first read the manuscript

654
00:33:46,720 --> 00:33:50,740
and did the exercises
and open problems,

655
00:33:50,740 --> 00:33:55,930
I didn't really think of the
opportunism in technology.

656
00:33:55,930 --> 00:34:00,790
But somehow, later
on, when I thought

657
00:34:00,790 --> 00:34:03,730
about deciding whether to
pursue computer science or not,

658
00:34:03,730 --> 00:34:08,500
then I realized that
beside of an interest

659
00:34:08,500 --> 00:34:12,980
in problems, computer science
or so is a new discipline.

660
00:34:12,980 --> 00:34:16,989
I think that at that
time, I understood

661
00:34:16,989 --> 00:34:20,380
computer science was very, very
new compared with mathematics.

662
00:34:20,380 --> 00:34:23,020
But now, after 40 years
when I think back,

663
00:34:23,020 --> 00:34:25,480
I realized that how
close to the beginning

664
00:34:25,480 --> 00:34:27,909
I was entering into that field.

665
00:34:27,909 --> 00:34:29,949
Because when I entered
the field there

666
00:34:29,949 --> 00:34:37,230
were already many big names
and people, but I think that--

667
00:34:37,230 --> 00:34:44,300


668
00:34:44,300 --> 00:34:46,420
I think that it
is very important

669
00:34:46,420 --> 00:34:50,770
to get into an area early, so
I think that perhaps all of us

670
00:34:50,770 --> 00:34:53,139
have that advantage here.

671
00:34:53,139 --> 00:34:58,900
I think the evolution
of the technologies

672
00:34:58,900 --> 00:35:03,490
was kind of foreseen by
most people in a misty way,

673
00:35:03,490 --> 00:35:06,970
without recognizing everything.

674
00:35:06,970 --> 00:35:09,850
Clearly, some of the
many computer vendors

675
00:35:09,850 --> 00:35:15,830
didn't see the impact of the
processor on a chip very well

676
00:35:15,830 --> 00:35:18,050
and are out of business.

677
00:35:18,050 --> 00:35:24,500
But from the earliest
days of time-sharing,

678
00:35:24,500 --> 00:35:29,300
we had already discovered
social interaction

679
00:35:29,300 --> 00:35:31,890
in the form of
messages to each other,

680
00:35:31,890 --> 00:35:34,550
which later evolved into email.

681
00:35:34,550 --> 00:35:38,450
Initially, they were just
messages within the system.

682
00:35:38,450 --> 00:35:41,270
And today, the
social interaction

683
00:35:41,270 --> 00:35:44,870
extends to Facebook
and now even Twitter.

684
00:35:44,870 --> 00:35:48,800
Yes, I do have a Facebook
account, but I'm like Ron,

685
00:35:48,800 --> 00:35:50,270
I don't log in very well.

686
00:35:50,270 --> 00:35:53,892
I was coaxed into it
by my grandchildren.

687
00:35:53,892 --> 00:35:56,990


688
00:35:56,990 --> 00:36:04,490
But, I too share with
Barbara the hesitation

689
00:36:04,490 --> 00:36:09,770
to expose more than I have
today on my website, which

690
00:36:09,770 --> 00:36:12,590
is a brief bio, which
is comparable to what

691
00:36:12,590 --> 00:36:17,180
you can find just by any
normal Google search.

692
00:36:17,180 --> 00:36:23,190
So I think some of
those reservations

693
00:36:23,190 --> 00:36:27,600
are because maybe we've been
around long enough to recognize

694
00:36:27,600 --> 00:36:31,320
some of the disasters that
occur when people make mistakes.

695
00:36:31,320 --> 00:36:38,700
But I think most people have
that aha moment when they send

696
00:36:38,700 --> 00:36:41,010
an email, they
regret it, and they

697
00:36:41,010 --> 00:36:44,850
realize they can't get
it back and there's

698
00:36:44,850 --> 00:36:47,370
nothing they can do.

699
00:36:47,370 --> 00:36:49,860
And that was a
revelation to everybody

700
00:36:49,860 --> 00:36:52,110
when they first discovered.

701
00:36:52,110 --> 00:36:54,870
Well maybe the next
Turing Award winner

702
00:36:54,870 --> 00:36:58,758
should be the person who
invents the unsend key.

703
00:36:58,758 --> 00:37:01,300
Steve, I wanted to respond to
your question, too, if I could.

704
00:37:01,300 --> 00:37:03,092
Because theory is a
bit different, I think.

705
00:37:03,092 --> 00:37:05,730
And the ability to
do theory untethered

706
00:37:05,730 --> 00:37:09,060
to the technology of the day
is a characteristic of theory,

707
00:37:09,060 --> 00:37:10,380
sort of by definition.

708
00:37:10,380 --> 00:37:13,290
And we see that today with
quantum computing, right.

709
00:37:13,290 --> 00:37:16,290
It's not clear at all whether
we can build quantum computers.

710
00:37:16,290 --> 00:37:18,962
There may be
possible, but there's

711
00:37:18,962 --> 00:37:20,170
a wonderful theory involving.

712
00:37:20,170 --> 00:37:23,200
Scott Aaronson and other
people are working on this.

713
00:37:23,200 --> 00:37:27,352
And it's an example of
theory evolving, independent

714
00:37:27,352 --> 00:37:28,810
of the technological
underpinnings.

715
00:37:28,810 --> 00:37:31,030
And I think had, in
response to your question,

716
00:37:31,030 --> 00:37:35,410
had the computer technology
frozen in the '60s or the '70s,

717
00:37:35,410 --> 00:37:37,738
we would have seen much of
the same theory evolve too.

718
00:37:37,738 --> 00:37:40,280
Maybe there wouldn't have been
as much funding, or something,

719
00:37:40,280 --> 00:37:40,780
I don't know.

720
00:37:40,780 --> 00:37:42,580
But it would have been the
kinds of questions asked

721
00:37:42,580 --> 00:37:43,960
and the kinds of
evolutions might

722
00:37:43,960 --> 00:37:45,430
have been very much the same.

723
00:37:45,430 --> 00:37:48,160
Recall that NP completeness
came about that time too.

724
00:37:48,160 --> 00:37:50,368


725
00:37:50,368 --> 00:37:51,910
Of course NP
completeness was sort of

726
00:37:51,910 --> 00:37:54,170
on the pessimistic
side of theory,

727
00:37:54,170 --> 00:37:56,170
what you can't do
maybe, so you don't

728
00:37:56,170 --> 00:37:58,840
need the hardware to do that.

729
00:37:58,840 --> 00:38:02,433
But even public-key
was ahead of its time,

730
00:38:02,433 --> 00:38:03,850
in the sense that
it really didn't

731
00:38:03,850 --> 00:38:06,940
have a utility until the
web was invented till Tim.

732
00:38:06,940 --> 00:38:08,410
But if the market
hadn't developed

733
00:38:08,410 --> 00:38:11,080
for public-key
cryptography, you think

734
00:38:11,080 --> 00:38:12,340
you still would have done it?

735
00:38:12,340 --> 00:38:14,530
You wouldn't just be an
obscure number theorist instead

736
00:38:14,530 --> 00:38:15,572
of a Turing award winner?

737
00:38:15,572 --> 00:38:19,030
There was a conference series
started in the early '80s,

738
00:38:19,030 --> 00:38:21,490
but the market for
cryptography didn't really

739
00:38:21,490 --> 00:38:22,850
happen until the '90s.

740
00:38:22,850 --> 00:38:25,060
So it's a whole decade
of vigorous theoretical

741
00:38:25,060 --> 00:38:28,060
exploration before the
market happened there.

742
00:38:28,060 --> 00:38:31,880
So theoretical work and
cryptography is part of that.

743
00:38:31,880 --> 00:38:33,630
There's been a big
market for cryptography

744
00:38:33,630 --> 00:38:35,190
for hundreds of years.

745
00:38:35,190 --> 00:38:39,870
Well big in the sense
of the military market

746
00:38:39,870 --> 00:38:43,020
has always been
there, of course.

747
00:38:43,020 --> 00:38:43,860
OK.

748
00:38:43,860 --> 00:38:45,610
Before leaving the
Moore's law thing,

749
00:38:45,610 --> 00:38:49,950
can we look at potentially
the dark side of it?

750
00:38:49,950 --> 00:38:54,810
Is it possible that the Moore's
law growth of technology

751
00:38:54,810 --> 00:38:58,650
has made too much function
available too fast,

752
00:38:58,650 --> 00:39:01,170
that we haven't actually
had time to develop

753
00:39:01,170 --> 00:39:03,570
mature engineering disciplines?

754
00:39:03,570 --> 00:39:07,440
There's a lot of complaints
about the quality of software

755
00:39:07,440 --> 00:39:10,520
these days, for example.

756
00:39:10,520 --> 00:39:12,270
It's the law of nature
that there's always

757
00:39:12,270 --> 00:39:14,853
going to be a lot of complaints
about the quality of software.

758
00:39:14,853 --> 00:39:15,580
Why is that?

759
00:39:15,580 --> 00:39:20,620
Because we always
raise our aspirations

760
00:39:20,620 --> 00:39:22,270
to the point where
we can't quite do

761
00:39:22,270 --> 00:39:24,720
it or we can just barely do it.

762
00:39:24,720 --> 00:39:25,410
Yeah.

763
00:39:25,410 --> 00:39:30,180
So that's just-- words fail me.

764
00:39:30,180 --> 00:39:34,290
But I actually think that the
much more serious implications

765
00:39:34,290 --> 00:39:37,260
are the sociological
implications of computing.

766
00:39:37,260 --> 00:39:39,150
I mean, we heard
yesterday in the talks

767
00:39:39,150 --> 00:39:43,290
about the financial markets
about the danger of completely

768
00:39:43,290 --> 00:39:45,480
computer-controlled trading.

769
00:39:45,480 --> 00:39:47,880
And I think if you look
across our technology,

770
00:39:47,880 --> 00:39:49,920
you see this in many sectors.

771
00:39:49,920 --> 00:39:51,570
That we have developed
the technology--

772
00:39:51,570 --> 00:39:53,612
I mean, I don't think the
problem of the software

773
00:39:53,612 --> 00:39:56,160
is the big problem, I think
it's the control of technology.

774
00:39:56,160 --> 00:39:58,850
But of course, this is not
limited to our technology.

775
00:39:58,850 --> 00:40:02,580
Yeah it's not as bad
as nuclear fission.

776
00:40:02,580 --> 00:40:05,470
Just look at it that way when
you start wringing your hands.

777
00:40:05,470 --> 00:40:08,280
But nevertheless, it's
a serious problem.

778
00:40:08,280 --> 00:40:10,590
And it takes time for
a society to catch up

779
00:40:10,590 --> 00:40:12,830
with the innovations.

780
00:40:12,830 --> 00:40:16,440
It's not as bad as the
spinning jenny, either.

781
00:40:16,440 --> 00:40:21,230
OK, let's exercise your
crystal ball a little bit more.

782
00:40:21,230 --> 00:40:26,100
Put yourself in the position
of a new PhD student

783
00:40:26,100 --> 00:40:27,840
just starting your career.

784
00:40:27,840 --> 00:40:29,788
What problem would
you choose to work on?

785
00:40:29,788 --> 00:40:31,830
What do you think are the
important problems that

786
00:40:31,830 --> 00:40:34,020
are facing computer science
now, and the ones where

787
00:40:34,020 --> 00:40:37,118
there are still dramatic
progress to be made?

788
00:40:37,118 --> 00:40:39,660
I'll give the same answer I gave
last time a similar question

789
00:40:39,660 --> 00:40:40,535
was asked on a panel.

790
00:40:40,535 --> 00:40:42,394
P equals NP.

791
00:40:42,394 --> 00:40:44,382
[LAUGHTER]

792
00:40:44,382 --> 00:40:46,870


793
00:40:46,870 --> 00:40:47,860
Well I would venture--

794
00:40:47,860 --> 00:40:50,140
Not always silence is ever met.

795
00:40:50,140 --> 00:40:52,790
For a short time.

796
00:40:52,790 --> 00:40:58,270
Would you actually, as a PhD
supervisor with a new PhD

797
00:40:58,270 --> 00:41:00,940
student, would you
advise him to tackle

798
00:41:00,940 --> 00:41:04,690
as a thesis topic settling
the P equals NP question?

799
00:41:04,690 --> 00:41:06,437
No, that would be unwise.

800
00:41:06,437 --> 00:41:08,020
Think that would be
bad career advice.

801
00:41:08,020 --> 00:41:10,920
It's not quite what was asked.

802
00:41:10,920 --> 00:41:17,220
Well I would advise a
really brilliant PhD student

803
00:41:17,220 --> 00:41:19,860
to work on the factoring
of large integers

804
00:41:19,860 --> 00:41:21,000
by classical means.

805
00:41:21,000 --> 00:41:24,687
Which may irk Ron
a bit, but I think

806
00:41:24,687 --> 00:41:25,937
that that's the right problem.

807
00:41:25,937 --> 00:41:27,240
It's the question we want
to know the answer to.

808
00:41:27,240 --> 00:41:27,800
Yeah.

809
00:41:27,800 --> 00:41:28,380
Yeah, so.

810
00:41:28,380 --> 00:41:28,890
Yeah.

811
00:41:28,890 --> 00:41:33,360
I think machine learning
is a really promising area.

812
00:41:33,360 --> 00:41:35,110
It's vibrant right
now, a lot happening.

813
00:41:35,110 --> 00:41:36,810
And I think new
students with a lot

814
00:41:36,810 --> 00:41:40,710
of mathematical or other skills
could make progress there,

815
00:41:40,710 --> 00:41:45,140
and I think that's an area which
is going to be recommended.

816
00:41:45,140 --> 00:41:48,650
Well Ed Lazowska already
gave my spiel on this subject

817
00:41:48,650 --> 00:41:51,260
yesterday.

818
00:41:51,260 --> 00:41:54,980
I think it's very clear
that although sort

819
00:41:54,980 --> 00:42:00,350
of classical simulation
and communication

820
00:42:00,350 --> 00:42:02,210
applications of
computing will continue

821
00:42:02,210 --> 00:42:06,500
to be interesting and important,
the really hot area is going

822
00:42:06,500 --> 00:42:08,720
to be what I like
to call embodiment

823
00:42:08,720 --> 00:42:11,780
and what the NSF calls
by the bureaucratic name

824
00:42:11,780 --> 00:42:14,360
cyber-physical systems.

825
00:42:14,360 --> 00:42:17,240
Which to me means
computer systems

826
00:42:17,240 --> 00:42:18,920
that have non-trivial
interactions

827
00:42:18,920 --> 00:42:20,093
with the physical world.

828
00:42:20,093 --> 00:42:22,010
This is something that's
just getting started,

829
00:42:22,010 --> 00:42:25,460
and that's going to be
even more interesting

830
00:42:25,460 --> 00:42:29,750
and important than the areas
that have already been mined.

831
00:42:29,750 --> 00:42:32,750
So I thought Ed's talk
yesterday was great,

832
00:42:32,750 --> 00:42:35,910
and you gave me hope
about systems again.

833
00:42:35,910 --> 00:42:38,330
I think that the
characterization of systems

834
00:42:38,330 --> 00:42:40,250
going up and down
the trough of despair

835
00:42:40,250 --> 00:42:42,410
and so forth, I think
that's really right on.

836
00:42:42,410 --> 00:42:46,940
And it's very hard to answer
the question for any of us

837
00:42:46,940 --> 00:42:49,010
I think, of what
we would do today,

838
00:42:49,010 --> 00:42:50,930
because we're not young anymore.

839
00:42:50,930 --> 00:42:53,810
And when you're young you see
the world a little differently

840
00:42:53,810 --> 00:42:55,490
than when you're not so young.

841
00:42:55,490 --> 00:42:58,363
My thought about
today is that I still

842
00:42:58,363 --> 00:43:00,530
think there are very
interesting issues and systems,

843
00:43:00,530 --> 00:43:03,500
but I also think that it's
not really embodiment so more,

844
00:43:03,500 --> 00:43:07,260
as much as the combination of
computer science and something

845
00:43:07,260 --> 00:43:07,760
else.

846
00:43:07,760 --> 00:43:12,050
Like, I do think computational
biology is really interesting,

847
00:43:12,050 --> 00:43:15,830
and so if I were starting off
today that's the kind of thing

848
00:43:15,830 --> 00:43:16,730
I might do.

849
00:43:16,730 --> 00:43:18,740
I think all of us
benefited from the fact

850
00:43:18,740 --> 00:43:21,133
that we happened to
see this new field

851
00:43:21,133 --> 00:43:23,550
and we were lucky enough to
get in at the beginning of it.

852
00:43:23,550 --> 00:43:26,540
And I think that's a way that
you have a really interesting

853
00:43:26,540 --> 00:43:29,780
career, because there are just
so many open problems that you

854
00:43:29,780 --> 00:43:32,630
can work on that people
have not solved before.

855
00:43:32,630 --> 00:43:35,180
But it's very hard for us
today to see what goes--

856
00:43:35,180 --> 00:43:37,040
Well, isn't the flip
side of that argument,

857
00:43:37,040 --> 00:43:40,370
to put it very cynically,
that you guys have taken

858
00:43:40,370 --> 00:43:42,170
all the low hanging fruit?

859
00:43:42,170 --> 00:43:45,470
You've built these great
systems in the '60s and '70s

860
00:43:45,470 --> 00:43:49,275
and all the new PhDs can
do is kind of fine tune

861
00:43:49,275 --> 00:43:50,900
the fundamentals that
you've laid down?

862
00:43:50,900 --> 00:43:52,100
Absolutely not.

863
00:43:52,100 --> 00:43:53,840
Somehow they find
interesting things.

864
00:43:53,840 --> 00:43:56,060
That's right.

865
00:43:56,060 --> 00:43:57,080
OK.

866
00:43:57,080 --> 00:44:02,540
So another question from an
attendee, directed to all.

867
00:44:02,540 --> 00:44:06,740
And it reads, isn't it time for
parallel programming techniques

868
00:44:06,740 --> 00:44:09,070
that mortals can understand?

869
00:44:09,070 --> 00:44:10,160
Yeah.

870
00:44:10,160 --> 00:44:11,110
It is, indeed.

871
00:44:11,110 --> 00:44:11,930
Yes it is.

872
00:44:11,930 --> 00:44:15,470


873
00:44:15,470 --> 00:44:18,110
That's a really
important question,

874
00:44:18,110 --> 00:44:20,990
and it's not clear to me
that the answer is yes.

875
00:44:20,990 --> 00:44:23,360
That in the end we will succeed.

876
00:44:23,360 --> 00:44:26,120
And if we can't
succeed, then that's

877
00:44:26,120 --> 00:44:28,670
going to limit the
applicability of

878
00:44:28,670 --> 00:44:30,750
the huge multi-core computers.

879
00:44:30,750 --> 00:44:33,260
And that's not
necessarily bad, it just

880
00:44:33,260 --> 00:44:37,970
means that the technology
will go in one way or another.

881
00:44:37,970 --> 00:44:41,630
God did not say that computers
had to have thousands of cores.

882
00:44:41,630 --> 00:44:44,760
But I think as a
technical challenge,

883
00:44:44,760 --> 00:44:46,525
it's a huge interesting
problem right now.

884
00:44:46,525 --> 00:44:48,650
Is it a programming language
problem, do you think?

885
00:44:48,650 --> 00:44:49,400
I think so, yeah.

886
00:44:49,400 --> 00:44:50,960
It's a programming
language problem

887
00:44:50,960 --> 00:44:53,600
and it's a methodology
problem, even more.

888
00:44:53,600 --> 00:44:56,190
Because programming
languages follow methodology,

889
00:44:56,190 --> 00:44:57,440
so it's a methodology problem.

890
00:44:57,440 --> 00:45:01,050
If we can figure out how
to organize those programs,

891
00:45:01,050 --> 00:45:02,840
that's the first
thing we have to do.

892
00:45:02,840 --> 00:45:04,880
Then we can invent
programming languages

893
00:45:04,880 --> 00:45:06,230
that support the methodology.

894
00:45:06,230 --> 00:45:08,450
But so far we
don't have too much

895
00:45:08,450 --> 00:45:11,200
of a clue about what
to do with this.

896
00:45:11,200 --> 00:45:13,490
This one striking
counter-example

897
00:45:13,490 --> 00:45:14,900
to that, which is MapReduce.

898
00:45:14,900 --> 00:45:18,170
Yes, but I don't think
MapReduce goes all that far

899
00:45:18,170 --> 00:45:20,595
in solving the general
class of problems.

900
00:45:20,595 --> 00:45:22,970
It goes nowhere in solving
the general class of problems,

901
00:45:22,970 --> 00:45:23,630
that's certainly true.

902
00:45:23,630 --> 00:45:25,963
But there is a large class
of problems where we actually

903
00:45:25,963 --> 00:45:28,600
do have a reasonable answer.

904
00:45:28,600 --> 00:45:31,840
Well should the question
be actually limited

905
00:45:31,840 --> 00:45:33,370
to parallel programming?

906
00:45:33,370 --> 00:45:37,590
I mean, what about just
everyday programming?

907
00:45:37,590 --> 00:45:39,420
One could ask the
skeptical question

908
00:45:39,420 --> 00:45:41,910
that despite all that you've
taught us about programming

909
00:45:41,910 --> 00:45:46,162
languages, Barbara,
JavaScript has emerged

910
00:45:46,162 --> 00:45:48,370
as the most popular programming
language in the world

911
00:45:48,370 --> 00:45:52,640
and that's hardly an
exemplar of your lessons.

912
00:45:52,640 --> 00:45:55,030
Can that be
interpreted as evidence

913
00:45:55,030 --> 00:45:58,528
that programming languages
don't really matter that much?

914
00:45:58,528 --> 00:46:00,070
The correct way to
look at JavaScript

915
00:46:00,070 --> 00:46:03,670
is it's a machine language.

916
00:46:03,670 --> 00:46:06,460
Well it's a machine language
that a lot of lines of code

917
00:46:06,460 --> 00:46:07,750
are written in every year.

918
00:46:07,750 --> 00:46:09,593
Only for relatively
small programs.

919
00:46:09,593 --> 00:46:10,510
People want to write--

920
00:46:10,510 --> 00:46:11,080
It's changing.

921
00:46:11,080 --> 00:46:13,000
People want to write big
programs, I don't think so.

922
00:46:13,000 --> 00:46:13,990
People want to
write big programs,

923
00:46:13,990 --> 00:46:14,950
they write them
in something else

924
00:46:14,950 --> 00:46:16,570
and they compile
them into JavaScript

925
00:46:16,570 --> 00:46:18,120
so that the browser
will run them.

926
00:46:18,120 --> 00:46:18,880
Yeah.

927
00:46:18,880 --> 00:46:21,460
I mean, it's a sad
statement about our field

928
00:46:21,460 --> 00:46:25,182
that a lot of big programs
start off with small programs.

929
00:46:25,182 --> 00:46:26,890
And when you're building
a small program,

930
00:46:26,890 --> 00:46:29,050
you don't have to pay much
attention to methodology

931
00:46:29,050 --> 00:46:31,360
because you can just do
it through brute force.

932
00:46:31,360 --> 00:46:32,140
Willpower.

933
00:46:32,140 --> 00:46:33,710
Willpower.

934
00:46:33,710 --> 00:46:34,850
Brute force.

935
00:46:34,850 --> 00:46:37,000
It's a form of willpower, right.

936
00:46:37,000 --> 00:46:39,550
And then, so this is
why there's so much

937
00:46:39,550 --> 00:46:42,850
stuff written in languages that
are not really all that great.

938
00:46:42,850 --> 00:46:44,433
Because you start
off this way, you

939
00:46:44,433 --> 00:46:45,850
don't have to think
about it hard.

940
00:46:45,850 --> 00:46:49,232
Then when you get to the big
problems, you're in trouble.

941
00:46:49,232 --> 00:46:50,940
But it's, I don't know
how to solve this.

942
00:46:50,940 --> 00:46:52,935
Yeah but we're OK, for
the reason that I said.

943
00:46:52,935 --> 00:46:54,310
JavaScript is a
machine language,

944
00:46:54,310 --> 00:46:57,400
you can write your program
in any language you want.

945
00:46:57,400 --> 00:46:57,900
OK.

946
00:46:57,900 --> 00:47:03,020
I have a question addressed
to Butler, Barbara, and Corby.

947
00:47:03,020 --> 00:47:04,530
The question is this.

948
00:47:04,530 --> 00:47:07,470
As computation becomes
increasingly distributed,

949
00:47:07,470 --> 00:47:09,890
should we rethink the role
of the operating system?

950
00:47:09,890 --> 00:47:13,570


951
00:47:13,570 --> 00:47:16,110
So there's a very healthy
operating system community that

952
00:47:16,110 --> 00:47:18,540
thinks about this all the time.

953
00:47:18,540 --> 00:47:20,970
My basic belief is that
distributed computing

954
00:47:20,970 --> 00:47:22,890
is probably the
source of inspiration

955
00:47:22,890 --> 00:47:24,480
for parallel computing.

956
00:47:24,480 --> 00:47:27,150
Distribute computing
actually is not in bad shape,

957
00:47:27,150 --> 00:47:31,200
and it is the way that Google
does all its work today.

958
00:47:31,200 --> 00:47:33,360
It is the way that
in fact, companies

959
00:47:33,360 --> 00:47:35,130
that have lots and
lots of processors,

960
00:47:35,130 --> 00:47:37,650
those are parallel
machines, in a sense.

961
00:47:37,650 --> 00:47:40,230
It's just that they
use a particular memory

962
00:47:40,230 --> 00:47:42,900
model and a particular way of
thinking about how computation

963
00:47:42,900 --> 00:47:44,220
is organized.

964
00:47:44,220 --> 00:47:46,020
They run MapReduce.

965
00:47:46,020 --> 00:47:48,840
Some of them run MapReduce,
some of them don't.

966
00:47:48,840 --> 00:47:50,280
Most of them do, actually.

967
00:47:50,280 --> 00:47:51,882
If you get down to
the systems level,

968
00:47:51,882 --> 00:47:53,340
they're not running
MapReduce, they

969
00:47:53,340 --> 00:47:54,750
are implementing MapReduce.

970
00:47:54,750 --> 00:47:56,760
That is undeniably true.

971
00:47:56,760 --> 00:47:58,948
They're running some
boring variant of Unix

972
00:47:58,948 --> 00:47:59,990
down at the system level.

973
00:47:59,990 --> 00:48:02,280
Yeah, it's not clear to me
that the operating systems

974
00:48:02,280 --> 00:48:03,060
are different.

975
00:48:03,060 --> 00:48:05,690
That the problems that people
look at in the operating system

976
00:48:05,690 --> 00:48:07,190
may change a little
bit, the way you

977
00:48:07,190 --> 00:48:09,840
do load balancing, your
notions of virtualization,

978
00:48:09,840 --> 00:48:11,190
they may evolve a bit.

979
00:48:11,190 --> 00:48:13,230
But it's not clear
that they're very

980
00:48:13,230 --> 00:48:15,642
different than what they were.

981
00:48:15,642 --> 00:48:17,350
Not much has changed
in operating systems

982
00:48:17,350 --> 00:48:19,423
since the mid '70s.

983
00:48:19,423 --> 00:48:21,340
In my opinion it's not
for commercial reasons,

984
00:48:21,340 --> 00:48:23,590
it's because we haven't
had any great new ideas.

985
00:48:23,590 --> 00:48:26,400
Aren't there people who
argue that the web is now

986
00:48:26,400 --> 00:48:27,870
our operating system?

987
00:48:27,870 --> 00:48:30,940
I have no idea what that means.

988
00:48:30,940 --> 00:48:32,940
Yes, certainly, there are
people who argue that,

989
00:48:32,940 --> 00:48:35,430
but I have no idea
what it means.

990
00:48:35,430 --> 00:48:35,930
OK.

991
00:48:35,930 --> 00:48:38,550


992
00:48:38,550 --> 00:48:41,880
Here's a question,
perhaps for our theorists.

993
00:48:41,880 --> 00:48:44,640
Has there been any progress
made in the question of P equals

994
00:48:44,640 --> 00:48:47,610
NP over the last 20 years?

995
00:48:47,610 --> 00:48:51,010
And do people do people
still work on this problem?

996
00:48:51,010 --> 00:48:53,580


997
00:48:53,580 --> 00:48:58,140
I think that there are people
who still work on this problem,

998
00:48:58,140 --> 00:49:02,220
as evidenced by the
claim of solving

999
00:49:02,220 --> 00:49:06,070
P equal to NP or the
other way every few years.

1000
00:49:06,070 --> 00:49:12,000
I think that the most recent
one appeared with a very high

1001
00:49:12,000 --> 00:49:16,350
publicity about a year ago.

1002
00:49:16,350 --> 00:49:20,020
So I think there are people
who still work on it,

1003
00:49:20,020 --> 00:49:26,070
and I think the
last one got some--

1004
00:49:26,070 --> 00:49:28,950


1005
00:49:28,950 --> 00:49:32,130
I think that most of the
claims, people just brushed off.

1006
00:49:32,130 --> 00:49:36,200
But somehow the last one got
a little bit more esteem.

1007
00:49:36,200 --> 00:49:44,390


1008
00:49:44,390 --> 00:49:51,185
My feeling is that there has
not been any real progress of P

1009
00:49:51,185 --> 00:49:52,070
equal to NP.

1010
00:49:52,070 --> 00:49:58,850


1011
00:49:58,850 --> 00:50:01,670
I think that if you want
to win a Turing Award

1012
00:50:01,670 --> 00:50:03,470
by working on P equal to NP--

1013
00:50:03,470 --> 00:50:07,160
Butler, sorry-- I think
that it probably is not

1014
00:50:07,160 --> 00:50:09,860
a good idea for a PhD student.

1015
00:50:09,860 --> 00:50:12,060
I don't deny that.

1016
00:50:12,060 --> 00:50:14,750
We have made some progress,
but of a negative sort.

1017
00:50:14,750 --> 00:50:17,300
I mean, we do know why
certain classes of arguments

1018
00:50:17,300 --> 00:50:18,572
don't work.

1019
00:50:18,572 --> 00:50:20,660
Some of the
arguments relativize,

1020
00:50:20,660 --> 00:50:22,580
or there's other
issues with them,

1021
00:50:22,580 --> 00:50:27,910
so we're getting
more frustrated.

1022
00:50:27,910 --> 00:50:29,770
We're understanding
the narrow path

1023
00:50:29,770 --> 00:50:32,350
we have to follow to come
up with a proof, maybe.

1024
00:50:32,350 --> 00:50:33,160
OK.

1025
00:50:33,160 --> 00:50:33,880
OK.

1026
00:50:33,880 --> 00:50:36,760
Let me ask you to turn up
the gain on your crystal

1027
00:50:36,760 --> 00:50:38,830
ball a lot.

1028
00:50:38,830 --> 00:50:42,460
There will presumably be a
celebration like this in 2061

1029
00:50:42,460 --> 00:50:47,050
of MIT'S 200th anniversary.

1030
00:50:47,050 --> 00:50:48,310
Will it be a Turing session?

1031
00:50:48,310 --> 00:50:50,680
Will there be Turing Awards?

1032
00:50:50,680 --> 00:50:53,440
Will it be computer science,
will computer science

1033
00:50:53,440 --> 00:50:56,800
have just assimilated into
other disciplines like math

1034
00:50:56,800 --> 00:50:58,330
and physics?

1035
00:50:58,330 --> 00:51:00,580
Any ideas?

1036
00:51:00,580 --> 00:51:03,610
Well I could speak up and say
I don't expect to be there.

1037
00:51:03,610 --> 00:51:07,990


1038
00:51:07,990 --> 00:51:10,600
Well we will certainly reserve
a place for you, nonetheless.

1039
00:51:10,600 --> 00:51:12,340
Thank you.

1040
00:51:12,340 --> 00:51:17,260
Well, I have a feeling
that at MIT 200

1041
00:51:17,260 --> 00:51:19,570
there will still
be Turing Awards.

1042
00:51:19,570 --> 00:51:21,340
And actually, at
that time, I think

1043
00:51:21,340 --> 00:51:23,170
contrary to what
you think, I think

1044
00:51:23,170 --> 00:51:28,043
that computer science has
absorbed physics and biology

1045
00:51:28,043 --> 00:51:28,969
and everything else.

1046
00:51:28,969 --> 00:51:35,380


1047
00:51:35,380 --> 00:51:36,460
That's a great answer.

1048
00:51:36,460 --> 00:51:36,960
OK.

1049
00:51:36,960 --> 00:51:39,160
The person moderating the
panel, Steve, will be a robot.

1050
00:51:39,160 --> 00:51:40,035
It won't be a person.

1051
00:51:40,035 --> 00:51:40,838
OK.

1052
00:51:40,838 --> 00:51:42,380
Well the panelists
be robots as well?

1053
00:51:42,380 --> 00:51:44,140
We don't know.

1054
00:51:44,140 --> 00:51:45,500
Intriguing possibility.

1055
00:51:45,500 --> 00:51:46,000
OK.

1056
00:51:46,000 --> 00:51:52,420
At the ACM's 50th anniversary
celebration, which was in 1997,

1057
00:51:52,420 --> 00:51:56,530
my then boss Nathan
Myhrvold made a prediction

1058
00:51:56,530 --> 00:51:58,300
about what would happen.

1059
00:51:58,300 --> 00:52:00,680
What would be going on
50 years in the future?

1060
00:52:00,680 --> 00:52:02,920
They asked a lot of
people to do this.

1061
00:52:02,920 --> 00:52:05,350
I would never have dared to
make a prediction about what

1062
00:52:05,350 --> 00:52:07,330
computing is going to
be like in 50 years,

1063
00:52:07,330 --> 00:52:08,330
but Nathan is fearless.

1064
00:52:08,330 --> 00:52:10,858
He predicted uploading.

1065
00:52:10,858 --> 00:52:13,150
So uploading means you get
to upload your consciousness

1066
00:52:13,150 --> 00:52:14,480
into a computer.

1067
00:52:14,480 --> 00:52:15,730
Why would you want to do that?

1068
00:52:15,730 --> 00:52:18,610
Well, you can be immortal.

1069
00:52:18,610 --> 00:52:19,210
Great.

1070
00:52:19,210 --> 00:52:22,840
And it seems to me, if
you're a materialist,

1071
00:52:22,840 --> 00:52:24,160
you can't really rule this out.

1072
00:52:24,160 --> 00:52:26,800


1073
00:52:26,800 --> 00:52:27,680
OK.

1074
00:52:27,680 --> 00:52:28,180
Well.

1075
00:52:28,180 --> 00:52:31,210


1076
00:52:31,210 --> 00:52:35,860
On that profound
thought, I'd like

1077
00:52:35,860 --> 00:52:41,350
to close this session by
resurrecting an old tradition

1078
00:52:41,350 --> 00:52:44,180
that Mike Dertouzos
started many decades ago.

1079
00:52:44,180 --> 00:52:47,560
He hosted the distinguished
lecturer series

1080
00:52:47,560 --> 00:52:51,850
of LCS, which is now called the
Dertouzos lectures, and made

1081
00:52:51,850 --> 00:52:55,150
a practice of asking each of
his distinguished lecturers

1082
00:52:55,150 --> 00:53:00,040
to summarize their life
experience into a one sentence

1083
00:53:00,040 --> 00:53:01,820
lesson.

1084
00:53:01,820 --> 00:53:04,330
So I'm going to put
each of you on the spot

1085
00:53:04,330 --> 00:53:08,380
and ask you for your sort of
one sentence words of wisdom.

1086
00:53:08,380 --> 00:53:10,660
And of course, you're all
Turing Award panelists,

1087
00:53:10,660 --> 00:53:13,960
so you have the strong
advantage that you

1088
00:53:13,960 --> 00:53:16,480
can say something that's
merely inscrutable

1089
00:53:16,480 --> 00:53:18,475
and the audience will
interpret it as profound.

1090
00:53:18,475 --> 00:53:21,820


1091
00:53:21,820 --> 00:53:25,030
So Corby, can I
ask you to start?

1092
00:53:25,030 --> 00:53:27,810
I'll give it a stab.

1093
00:53:27,810 --> 00:53:32,960
I think if there's any
single thing that--

1094
00:53:32,960 --> 00:53:35,180
a piece of advice
one could give,

1095
00:53:35,180 --> 00:53:39,140
it is to follow your
enthusiasm, whatever

1096
00:53:39,140 --> 00:53:42,080
that may turn out to be.

1097
00:53:42,080 --> 00:53:46,130
And in particular,
take advantage

1098
00:53:46,130 --> 00:53:56,810
of the ambience of
a place like MIT,

1099
00:53:56,810 --> 00:54:01,340
where one has to recognize
in order to enjoy your life

1100
00:54:01,340 --> 00:54:08,150
you have to consider
yourself a lifelong student,

1101
00:54:08,150 --> 00:54:13,250
and you really should take
a drink from the fire hose.

1102
00:54:13,250 --> 00:54:14,270
Thank you, Corby.

1103
00:54:14,270 --> 00:54:15,860
Butler?

1104
00:54:15,860 --> 00:54:17,360
I think Corby said
it very well, you

1105
00:54:17,360 --> 00:54:20,170
should do something that
you really like doing.

1106
00:54:20,170 --> 00:54:23,470
And if you like
doing lots of things,

1107
00:54:23,470 --> 00:54:29,440
I suggest that you consider
either biology, or chemistry,

1108
00:54:29,440 --> 00:54:35,020
or computing as the hot
fields for the next two

1109
00:54:35,020 --> 00:54:36,670
or three decades, at least.

1110
00:54:36,670 --> 00:54:37,480
Good.

1111
00:54:37,480 --> 00:54:38,825
Barbara?

1112
00:54:38,825 --> 00:54:40,450
So I think you should
be opportunistic.

1113
00:54:40,450 --> 00:54:42,770
I think opportunism
is a good thing.

1114
00:54:42,770 --> 00:54:44,470
And what it means
is you look around

1115
00:54:44,470 --> 00:54:47,710
and you see what's
moving and what's open

1116
00:54:47,710 --> 00:54:52,060
and then you should focus on
what are the important problems

1117
00:54:52,060 --> 00:54:54,350
and avoid doing
incremental work.

1118
00:54:54,350 --> 00:54:57,070
And if it requires
changing your field,

1119
00:54:57,070 --> 00:55:00,088
then you shouldn't
be afraid to do that.

1120
00:55:00,088 --> 00:55:01,430
Ron?

1121
00:55:01,430 --> 00:55:05,905
I guess it's very much in
the same sort of spirit.

1122
00:55:05,905 --> 00:55:07,280
The thing I would
advise students

1123
00:55:07,280 --> 00:55:10,580
would be to ask questions
based on the technology,

1124
00:55:10,580 --> 00:55:14,000
the hot technology of the day,
but to ask them in such a way

1125
00:55:14,000 --> 00:55:16,130
and try to derive
answers that will

1126
00:55:16,130 --> 00:55:19,920
be interesting a couple
of decades from now.

1127
00:55:19,920 --> 00:55:20,580
Good.

1128
00:55:20,580 --> 00:55:21,870
Andy?

1129
00:55:21,870 --> 00:55:24,780
My advice would be,
for the students,

1130
00:55:24,780 --> 00:55:27,750
would be to always
try to do something

1131
00:55:27,750 --> 00:55:31,290
that is just a little
bit beyond your ability.

1132
00:55:31,290 --> 00:55:31,883
Great.

1133
00:55:31,883 --> 00:55:34,050
Well, I'd like to thank our
distinguished panelists.

1134
00:55:34,050 --> 00:55:48,530


1135
00:55:48,530 --> 00:55:52,005
We have 4 and 1/2 minutes.

1136
00:55:52,005 --> 00:55:53,630
What would you like
to talk more about?

1137
00:55:53,630 --> 00:55:55,978


1138
00:55:55,978 --> 00:55:58,520
This is the first time I've ever
seen one of these things cut

1139
00:55:58,520 --> 00:56:01,580
off before its deadline.

1140
00:56:01,580 --> 00:56:02,990
I was going by my watch.

1141
00:56:02,990 --> 00:56:06,080


1142
00:56:06,080 --> 00:56:07,770
Oh, so there's--

1143
00:56:07,770 --> 00:56:10,880
Other questions?

1144
00:56:10,880 --> 00:56:13,300
Shall we wing it and take a
question from the audience?

1145
00:56:13,300 --> 00:56:13,800
Certainly.

1146
00:56:13,800 --> 00:56:14,430
Sure, why not?

1147
00:56:14,430 --> 00:56:14,930
Sure.

1148
00:56:14,930 --> 00:56:15,490
Yeah.

1149
00:56:15,490 --> 00:56:16,240
It's pretty scary.

1150
00:56:16,240 --> 00:56:17,115
A real-time question.

1151
00:56:17,115 --> 00:56:19,370
Yeah.

1152
00:56:19,370 --> 00:56:23,780
On the P equals NP
problem, what about proving

1153
00:56:23,780 --> 00:56:29,015
that neither P equals NP nor
P not equals NP is provable?

1154
00:56:29,015 --> 00:56:32,192


1155
00:56:32,192 --> 00:56:33,275
I would say it's possible.

1156
00:56:33,275 --> 00:56:38,500


1157
00:56:38,500 --> 00:56:42,200
But you may not be able to know
whether that's true or not.

1158
00:56:42,200 --> 00:56:49,000


1159
00:56:49,000 --> 00:56:51,130
Do you think you can
entertain another question?

1160
00:56:51,130 --> 00:56:52,000
Sure.

1161
00:56:52,000 --> 00:56:53,440
Stand close to the mic.

1162
00:56:53,440 --> 00:56:55,360
I just-- sorry-- I just
wanted to make sure

1163
00:56:55,360 --> 00:56:57,365
that I was allowed to.

1164
00:56:57,365 --> 00:56:58,970
You were talking
about the future,

1165
00:56:58,970 --> 00:57:00,910
I was just wondering
what you might think.

1166
00:57:00,910 --> 00:57:03,880
One of you said
machine learning was

1167
00:57:03,880 --> 00:57:09,010
a very hot topic in the future,
and I might have been busy

1168
00:57:09,010 --> 00:57:13,700
and haven't heard the
Turing IBM Watson computer.

1169
00:57:13,700 --> 00:57:20,060
And of course, at the feet of
humanity as a search engine

1170
00:57:20,060 --> 00:57:22,590
because it can do more
than what Google can do.

1171
00:57:22,590 --> 00:57:25,700
It can take care of human
language, idiosyncrasies

1172
00:57:25,700 --> 00:57:27,870
and puns, and many other things.

1173
00:57:27,870 --> 00:57:29,840
So I think that with
very little changes

1174
00:57:29,840 --> 00:57:31,700
it could be a
stronger search engine

1175
00:57:31,700 --> 00:57:34,090
than Google and content search.

1176
00:57:34,090 --> 00:57:36,890
Also that, it only
runs at 80 petaFLOPS,

1177
00:57:36,890 --> 00:57:39,920
which is a rather slow
speed, because that speed has

1178
00:57:39,920 --> 00:57:45,710
been around for a year or
two, maybe a year or so.

1179
00:57:45,710 --> 00:57:49,190
Even Anton chips with DE
Shaw, even those specialized,

1180
00:57:49,190 --> 00:57:52,970
will probably be running faster
floating point operations

1181
00:57:52,970 --> 00:57:53,690
than that.

1182
00:57:53,690 --> 00:57:56,258
Even though application
is not a natural language.

1183
00:57:56,258 --> 00:57:58,550
The other thing I want to
ask you about it, if you want

1184
00:57:58,550 --> 00:58:00,842
to comment on it, besides
the Watson Computer's machine

1185
00:58:00,842 --> 00:58:03,275
learning success and
search engine success,

1186
00:58:03,275 --> 00:58:07,800
is what do you think of
Google and YouTube broadcast

1187
00:58:07,800 --> 00:58:10,610
as network like MTV?

1188
00:58:10,610 --> 00:58:12,200
The public doesn't
like newspapers,

1189
00:58:12,200 --> 00:58:14,232
the public likes to
generate their own news

1190
00:58:14,232 --> 00:58:16,190
because they think the
newspapers and broadcast

1191
00:58:16,190 --> 00:58:19,620
news are a little bit corrupt,
have their own agenda.

1192
00:58:19,620 --> 00:58:24,140
So if YouTube and broadcast
and watch their own news,

1193
00:58:24,140 --> 00:58:25,970
it would be kind of
like MTV for the young.

1194
00:58:25,970 --> 00:58:27,845
Because YouTube used to
be the young, but now

1195
00:58:27,845 --> 00:58:29,250
it's replaced by Facebook.

1196
00:58:29,250 --> 00:58:32,870
So if it weren't cable TV, it
could be the MTV [? change ?]

1197
00:58:32,870 --> 00:58:36,995
for Facebook, which is really
YouTube with all the pictures.

1198
00:58:36,995 --> 00:58:39,660
Is that a question?

1199
00:58:39,660 --> 00:58:41,540
There's a lot there
in that question.

1200
00:58:41,540 --> 00:58:43,880
I think my piece of
advice would be to focus.

1201
00:58:43,880 --> 00:58:44,380
Focus.

1202
00:58:44,380 --> 00:58:46,352
[LAUGHTER]

1203
00:58:46,352 --> 00:58:47,810
Focus and get things
narrowed down.

1204
00:58:47,810 --> 00:58:51,040


1205
00:58:51,040 --> 00:58:53,560
Well there is the
whole YouTube kind

1206
00:58:53,560 --> 00:58:57,970
of revolution, where the
citizenship, the browsers,

1207
00:58:57,970 --> 00:59:02,470
and so on are generating
neural content.

1208
00:59:02,470 --> 00:59:04,690
We retract your invitation.

1209
00:59:04,690 --> 00:59:08,990
Sorry, but we are in fact
now, officially out of time.

1210
00:59:08,990 --> 00:59:11,790
So, I thank the panelists.

1211
00:59:11,790 --> 00:59:13,000