1
00:00:00,570 --> 00:00:07,860
For the final lecture, I'm going to draw some conclusions about Searle and Turing lost,

2
00:00:07,860 --> 00:00:11,730
the last lecture may have seemed to come down rather more in cell's favour than

3
00:00:11,730 --> 00:00:16,500
Turing's because we were seeing some serious problems with the Turing test.

4
00:00:16,500 --> 00:00:22,590
Today, I'm going to be going somewhat in the other direction, as you'll see.

5
00:00:22,590 --> 00:00:25,920
So a summary of where we are so far on Turing and on.

6
00:00:25,920 --> 00:00:35,760
So first of all, Turing's predictions, the predictions that he made for the year 2000 I suggested look pretty plausible.

7
00:00:35,760 --> 00:00:46,650
But as a criterion of intelligence, the Turing test looks very dubious if we interpret it generously, which one of his predictions kind of hints at.

8
00:00:46,650 --> 00:00:48,030
Namely,

9
00:00:48,030 --> 00:01:01,710
can it reduce the success of the interrogator to less than or equal to 70 per cent after five minutes of questioning and an average interrogator?

10
00:01:01,710 --> 00:01:07,050
Then the trouble is that chat bots show it. It's an insufficient criterion.

11
00:01:07,050 --> 00:01:11,430
It's too easy to fool the average interrogator for five minutes.

12
00:01:11,430 --> 00:01:15,990
And even if current chat bots don't do that very well.

13
00:01:15,990 --> 00:01:19,590
If A.I. researchers were to put their mind seriously to it,

14
00:01:19,590 --> 00:01:27,180
I think they'd be able to adjust those in such a way as to get over that threshold relatively straightforwardly.

15
00:01:27,180 --> 00:01:31,860
But if, on the other hand, we interpret the criterion of the Turing test more strictly,

16
00:01:31,860 --> 00:01:36,690
then it becomes too demanding because computers give themselves away.

17
00:01:36,690 --> 00:01:49,650
By not following the cultural cues and things like that, it becomes too easy to tell when the response is from a non-human rather than a human.

18
00:01:49,650 --> 00:01:53,800
What about the Chinese room thought experiment of Searle?

19
00:01:53,800 --> 00:01:57,870
Well, first of all, it's wildly implausible in terms of practicality.

20
00:01:57,870 --> 00:02:04,500
And I've suggested that that casts some doubt on the intuitive responses we are inclined to make to it.

21
00:02:04,500 --> 00:02:15,030
We should be at least hesitant about taking a thought experiment that's completely devoid from reality and simply applying our naive intuitions to it.

22
00:02:15,030 --> 00:02:22,200
I'll be saying more on that later. We've noted two main objections the system reply and the robot reply.

23
00:02:22,200 --> 00:02:29,880
We noted that the system reply, as Copeland points out, begs the question, but still seems to retain a bit of force.

24
00:02:29,880 --> 00:02:37,920
Perhaps intuitive force, at least. We've yet to consider cells reply against the robot.

25
00:02:37,920 --> 00:02:48,360
I mean, we've seen it, but we've not discussed it. I ended by suggesting a better test or what I think is a better test than the Turing test,

26
00:02:48,360 --> 00:02:52,680
which is the idea of producing an intelligent tutoring system,

27
00:02:52,680 --> 00:02:58,020
a system where the intelligence of the computer is revealed,

28
00:02:58,020 --> 00:03:07,230
rather than resorting to the sort of hidden subtlety that is involved in the chat bots with chat bots.

29
00:03:07,230 --> 00:03:12,300
What the chat bot is trying to do is disguise its lack of intelligence with the tutoring system.

30
00:03:12,300 --> 00:03:18,690
The whole idea is to reveal the understanding of the complex web of concepts in,

31
00:03:18,690 --> 00:03:25,830
say, chemistry or whatever it might be in order to present them to the duty.

32
00:03:25,830 --> 00:03:30,480
Now suppose we have a system like that, a computerised tutoring system.

33
00:03:30,480 --> 00:03:41,250
It's capable of highly sophisticated information processing in a complex domain, and it's able to convey that to the human with whom it's interacting.

34
00:03:41,250 --> 00:03:52,720
It is responsive to information structures in ways that reflect how human experts would think, and it aims to convey that expertise to the Tutsi.

35
00:03:52,720 --> 00:03:56,670
Now, cells objection to this is going to be exactly the same.

36
00:03:56,670 --> 00:04:00,810
He's going to simply say, Well, the system's processing is merely semantic.

37
00:04:00,810 --> 00:04:11,250
It is a syntactic sorry. It has no genuine semantics, no real understanding of the domain it's thought cannot reach out to the real world.

38
00:04:11,250 --> 00:04:16,350
So even if this system actually is faithful to,

39
00:04:16,350 --> 00:04:23,670
if you like the logical interactions and mathematical interactions and so on between the different concepts in chemistry,

40
00:04:23,670 --> 00:04:32,400
the system itself has absolutely no awareness that the processing it's doing links in any way to chemicals out in the real world,

41
00:04:32,400 --> 00:04:41,940
so therefore it has no intentionality. OK, now I want to highlight a distinction that cell draws.

42
00:04:41,940 --> 00:04:48,210
This is in his book Rediscovery of the Mind Between Intrinsic ASAT, as if and derived intentionality.

43
00:04:48,210 --> 00:04:51,300
And this lies behind quite a lot of what he says.

44
00:04:51,300 --> 00:04:59,930
So what he's denying with the computer is intrinsic intentionality, the kind of intentionality that involves.

45
00:04:59,930 --> 00:05:02,510
A real, intentional mental state.

46
00:05:02,510 --> 00:05:16,820
So, for example, when I say I am now thirsty, I'm expressing a real mental state that has relation to things in the world, water wanting water.

47
00:05:16,820 --> 00:05:24,860
Contrast that with as if intentionality. Suppose I say my lawn is thirsty, meaning my lawn could do with being watered.

48
00:05:24,860 --> 00:05:30,200
Actually, the lawn has no genuine, intentional state. I'm just talking about it as though it did.

49
00:05:30,200 --> 00:05:37,790
That's fake intentionality. Call it as if intentionality. And then there's another kind derived intentionality.

50
00:05:37,790 --> 00:05:44,120
So suppose I say in French jagran, Swoff means I am very thirsty.

51
00:05:44,120 --> 00:05:49,370
I'm expressing the meaning of a foreign phrase.

52
00:05:49,370 --> 00:06:00,710
But that phrase? Well, what I'm doing is expressing the intentionality that a French person would express in using that phrase.

53
00:06:00,710 --> 00:06:10,130
So we can say that the phrase is intentional in the sense, but it's a derived intentionality from the intrinsic intentionality of language uses.

54
00:06:10,130 --> 00:06:13,250
Okay, so you'll find that on the pages I've mentioned in.

55
00:06:13,250 --> 00:06:21,320
So. Okay, so let's go back now and think of the tutoring system,

56
00:06:21,320 --> 00:06:29,870
and let's accept that in that system, the processed information has no intrinsic semantics.

57
00:06:29,870 --> 00:06:36,410
There's no semantic or grasp within the system itself.

58
00:06:36,410 --> 00:06:44,160
But I want to suggest that we might nevertheless consider the processing of the information to be intelligent.

59
00:06:44,160 --> 00:06:51,380
This is something so seems not to consider. He assumes that intelligence requires intrinsic semantics.

60
00:06:51,380 --> 00:06:56,000
But I think we can challenge that. We can say, OK, the system.

61
00:06:56,000 --> 00:07:06,680
Admittedly, it has no clue that the processing it's doing relates to links between chemicals and, you know, atoms and molecules and quantum stuff.

62
00:07:06,680 --> 00:07:08,640
And of course, it doesn't have any clue.

63
00:07:08,640 --> 00:07:18,350
But nevertheless, the processing that it's doing is intelligent because it is relevantly similar to the reasoning that a human expert does.

64
00:07:18,350 --> 00:07:23,030
OK, so that, it seems to me is is a possible reply here.

65
00:07:23,030 --> 00:07:29,060
We can also consider ways in which computer systems might perhaps achieve intrinsic semantics.

66
00:07:29,060 --> 00:07:41,720
So put that first response on one side for now, and let's consider whether a computer could achieve intrinsic semantics.

67
00:07:41,720 --> 00:07:46,550
OK. Well, the objection does seem quite strong sales objection in terms of physical things,

68
00:07:46,550 --> 00:07:53,150
you know that the system really it has this symbol inside it that represents H2O,

69
00:07:53,150 --> 00:08:04,100
but it doesn't have a clue that that has anything to do with what I may produce a system that you know, thinks about plants.

70
00:08:04,100 --> 00:08:09,140
It has a symbol for a tree, but it's got no idea that that actually relates to trees out there.

71
00:08:09,140 --> 00:08:15,350
Mere internal processing of formal symbols cannot genuinely constitute thought about real trees.

72
00:08:15,350 --> 00:08:20,510
That seems plausible. It's not so clear. We thought about abstract entities,

73
00:08:20,510 --> 00:08:28,820
say numbers or chess positions where intelligent processing needn't be responsive to the properties of concrete things like trees,

74
00:08:28,820 --> 00:08:33,680
but only to appropriate logical relations, and those can be represented formally.

75
00:08:33,680 --> 00:08:41,900
So I just want to give you an example here of a simple programme, and you can actually run this programme for yourself within the turtle system.

76
00:08:41,900 --> 00:08:50,030
It's available as a A. You can freely download this and it's one of the help, for example, files there.

77
00:08:50,030 --> 00:08:56,930
Here we've got a simple programme. I'm playing blue and the thing's going to beat me, right?

78
00:08:56,930 --> 00:08:59,930
Yeah, because I played a bad first move, right?

79
00:08:59,930 --> 00:09:07,840
And if I play in the corner, I can guarantee that it will play in the middle because that's the only move that doesn't lose.

80
00:09:07,840 --> 00:09:21,100
Gosh, it's taking a long time. Thinking hard, maybe are.

81
00:09:21,100 --> 00:09:26,290
OK, now I want to point out that programme is not driven by a look up table.

82
00:09:26,290 --> 00:09:30,310
OK? When it was working out to play in the middle,

83
00:09:30,310 --> 00:09:37,600
it really was calculating all the possible lines of there's a recursive procedure that does all the work.

84
00:09:37,600 --> 00:09:44,500
I mean, surprisingly short, really. But it is analysing all the way down the tree of analysis.

85
00:09:44,500 --> 00:09:57,580
So the the point here is that the processing there really does reflect the logic of the problem domain.

86
00:09:57,580 --> 00:10:05,890
When it decides which move to play, it is doing so in terms of the genuine relations between different positions.

87
00:10:05,890 --> 00:10:12,010
Of course, the positions are represented in a way quite different from how they're represented in our minds.

88
00:10:12,010 --> 00:10:17,650
But the links between the positions faithfully reflect the relationships that

89
00:10:17,650 --> 00:10:23,590
we are aware of in terms of a move changing from one position to another.

90
00:10:23,590 --> 00:10:27,310
So basically, it is doing what's called mini maxing.

91
00:10:27,310 --> 00:10:31,870
At the end of the handout. I've put six slides which explain how the programme works.

92
00:10:31,870 --> 00:10:38,050
If you've not come across this thing, this kind of thing that that will be illuminating, it's actually very straightforward.

93
00:10:38,050 --> 00:10:44,920
The programme obviously isn't aware, but its processing is appropriately responsive to the real logic of the game.

94
00:10:44,920 --> 00:10:53,050
So shouldn't that be enough for intelligence? It seems to me a reasonable claim or an arguable claim.

95
00:10:53,050 --> 00:10:56,680
It at least that it is OK.

96
00:10:56,680 --> 00:11:04,570
Having said that, I'm going to put all this on one side because I think getting into philosophy of mathematics,

97
00:11:04,570 --> 00:11:14,320
understanding our thinking about numbers and abstract relationships between games, positions and all the rest really is quite difficult and obscure.

98
00:11:14,320 --> 00:11:22,270
And in any case, soul is mainly concerned with the kind of thinking that we have in common with animals.

99
00:11:22,270 --> 00:11:28,570
And he thinks not with computers, namely thoughts about things in the real world, physical things that's much more central.

100
00:11:28,570 --> 00:11:39,250
So even if so, was to concede the point that that machines can be intelligent in an appropriate sense about abstract objects.

101
00:11:39,250 --> 00:11:47,050
They it's really thinking about more everyday physical objects and so on that are his main concern.

102
00:11:47,050 --> 00:11:58,330
Okay, so with apologies to Monty Python, you recognise that if you're familiar with it, I'm a lumberjack and I'm OK, I'm going to take on now.

103
00:11:58,330 --> 00:12:06,430
The robot reply to the Chinese room, and I'm going to combine it with the system or.

104
00:12:06,430 --> 00:12:16,040
So imagine a robotic crane armed with appropriate sensors and tours and tools, which is programmed to cut down trees intelligently and effectively.

105
00:12:16,040 --> 00:12:18,310
OK, I'm not begging the question by saying intelligently.

106
00:12:18,310 --> 00:12:24,310
All I mean is it's doing the kind of thing that an intelligent human would do in that position.

107
00:12:24,310 --> 00:12:29,710
It senses for itself which trees are suitable for chopping or pruning and which of best left.

108
00:12:29,710 --> 00:12:37,690
It takes account of relevant conservation needs. It's responsive to physical obstructions and other difficulties and real time events as it works.

109
00:12:37,690 --> 00:12:41,050
OK, so this robotic crane if, say,

110
00:12:41,050 --> 00:12:50,290
overnight some trees get blown down and are in the way it will respond appropriately will detect them using its sensors.

111
00:12:50,290 --> 00:12:54,430
It will chop them up. Carry the bits to where bits should be carried.

112
00:12:54,430 --> 00:13:03,710
Clear the lines and then get on with its job. So the internal states of the robot are responsive to physical things and impact causally on them.

113
00:13:03,710 --> 00:13:11,950
So we've got to weigh causation, right? It senses real things and is able to act on real things.

114
00:13:11,950 --> 00:13:16,990
OK, so the claim here is that robot.

115
00:13:16,990 --> 00:13:26,920
The robotic crane literally reaches out to the real world when the symbol tree occurs within its thinking.

116
00:13:26,920 --> 00:13:31,000
It actually does have an impact on the real world and is responsive to the real world.

117
00:13:31,000 --> 00:13:40,010
So why doesn't that have real intentionality? Well, remember his response to the robot reply?

118
00:13:40,010 --> 00:13:44,650
I would suggest we should consider a Chinese cabin in the grain fed,

119
00:13:44,650 --> 00:13:49,150
with messages from the sensors written in Chinese characters processed purely

120
00:13:49,150 --> 00:13:54,640
syntactically by the man inside and resulting in Chinese messages to the Cranes motors.

121
00:13:54,640 --> 00:13:58,780
All of these messages being incomprehensible to the guy in the cabin.

122
00:13:58,780 --> 00:14:03,010
So remember, his reply is a response to the robot reply.

123
00:14:03,010 --> 00:14:07,240
It's exactly like that.

124
00:14:07,240 --> 00:14:20,330
But this response is only conclusive is if his rebuttal to the system reply works okay because we can simply say, OK, so if you take the central.

125
00:14:20,330 --> 00:14:27,060
As a unit of the computer that's there in the cabin of the robot and replace it with a person.

126
00:14:27,060 --> 00:14:32,390
OK, let's agree that that person has no semantic grasp of what's going on.

127
00:14:32,390 --> 00:14:38,000
They're just receiving symbols from the sensors and so forth and churning out symbols in response that work,

128
00:14:38,000 --> 00:14:49,760
the motors and the saw and all that kind of stuff. No idea at all that they're even controlling a robotic lumberjack crane.

129
00:14:49,760 --> 00:14:56,780
But as Copeland insists, that doesn't imply that the system as a whole lacks semantics that requires another argument.

130
00:14:56,780 --> 00:15:03,770
So I'm appealing here to Copeland's point, but still has not refuted the system reply.

131
00:15:03,770 --> 00:15:10,760
And because of that, the fact that the the central processor unit of the crane,

132
00:15:10,760 --> 00:15:18,950
all the man who's taking that place doesn't have any semantic link with the world doesn't imply that the crane doesn't.

133
00:15:18,950 --> 00:15:25,970
The crane has sensors and motors and saws and all the rest as well.

134
00:15:25,970 --> 00:15:32,780
And in that case, I'm suggesting that the man's unawareness of that is irrelevant.

135
00:15:32,780 --> 00:15:35,480
Or at least it's a challenge to sell to explain why.

136
00:15:35,480 --> 00:15:45,140
It's not because if, if intentionality, you know, reaching out to the world, having reference to the real world is watching question.

137
00:15:45,140 --> 00:15:54,510
Why doesn't the robotic crane achieve that? It's having a very real impact and is responding to things in the real world.

138
00:15:54,510 --> 00:16:02,030
So what is it that machines are supposed to lack according to So?

139
00:16:02,030 --> 00:16:03,650
Well,

140
00:16:03,650 --> 00:16:14,630
the style of his arguments seems to suggest that the crucial thing is that they lack understanding internal conscious awareness of what's going on.

141
00:16:14,630 --> 00:16:18,110
I note in passing that the word understanding is a very slippery one.

142
00:16:18,110 --> 00:16:27,800
It quite often comes up in philosophical discussion whenever you see some philosopher saying understanding is required or something like that,

143
00:16:27,800 --> 00:16:36,830
be very suspicious. Ask yourself what exactly that means, because it can mean a lot of different things anyway.

144
00:16:36,830 --> 00:16:45,500
Animals, by the way, do have what computers lack, according to so visual and auditory experiences,

145
00:16:45,500 --> 00:16:51,830
tactile sensations, hunger, thirst and sexual desire all caused by brain processes.

146
00:16:51,830 --> 00:16:56,600
And they are realised in the structure of the brain, and they are all intentional phenomena.

147
00:16:56,600 --> 00:17:02,780
It is just a plain fact about biological evolution that is produced certain type sorts of biological systems,

148
00:17:02,780 --> 00:17:12,730
namely human and certain animal brains that have subjective features, and there are other passages that same effect.

149
00:17:12,730 --> 00:17:24,260
OK, so it seems to be something like animal consciousness that so he's on to hear.

150
00:17:24,260 --> 00:17:33,860
At other points, he recognises that there is another issue here, though our intentional state need not be conscious.

151
00:17:33,860 --> 00:17:41,450
They can be unconscious, so not all of the processing that goes on in our brains is conscious.

152
00:17:41,450 --> 00:17:49,940
What about that? Well, still draws a distinction between consciousness and unconsciousness,

153
00:17:49,940 --> 00:17:55,370
but wants to insist isn't what characterises an unconscious, intentional state.

154
00:17:55,370 --> 00:18:04,760
That is, that it is at least potentially conscious. Only a being that could have conscious intentional states could have intentional states at all,

155
00:18:04,760 --> 00:18:09,920
and every unconscious, intentional state is at least potentially conscious.

156
00:18:09,920 --> 00:18:15,650
OK, now we might ask questions about that. You know, can it really count as exactly the same state?

157
00:18:15,650 --> 00:18:23,670
If it can be both conscious and unconscious? We won't go into that. We let so have that.

158
00:18:23,670 --> 00:18:32,430
But I won't suggest that he's at risk of conflating two quite different things when he uses terms like semantic and intentionality.

159
00:18:32,430 --> 00:18:40,020
And one notion is roughly that of internal symbols having objective significance representing external things in some

160
00:18:40,020 --> 00:18:47,130
intrinsic way rather than being just thought of by some other agent as having such representative significance.

161
00:18:47,130 --> 00:18:55,950
So take the robotic crane, for example, and contrast that with a piece of text.

162
00:18:55,950 --> 00:19:02,310
The piece of text in itself has no intrinsic link to anything in the world,

163
00:19:02,310 --> 00:19:10,530
any representational power that it has is purely in terms of our reading it and understanding it in a particular way.

164
00:19:10,530 --> 00:19:15,900
But the robotic crane does have a direct links with the outside world.

165
00:19:15,900 --> 00:19:21,420
It doesn't need an agent once it started going.

166
00:19:21,420 --> 00:19:26,550
It is going to have links with the outside world so that the symbols within it

167
00:19:26,550 --> 00:19:31,950
have objective significance without requiring any other agent in the picture.

168
00:19:31,950 --> 00:19:40,450
At least Wally Power keeps getting. OK, so that's one notion.

169
00:19:40,450 --> 00:19:45,460
The other notion is that of internal symbols having subjective significance to the agent

170
00:19:45,460 --> 00:19:51,280
in question and hence requiring either consciousness or at least potential consciousness.

171
00:19:51,280 --> 00:19:58,840
So this one is more demanding than that. And I'm suggesting that so kind of conflates them together.

172
00:19:58,840 --> 00:20:08,110
So if we do conflate these notes, it's going to pretty much follow straight away that only conscious beings can have intelligence.

173
00:20:08,110 --> 00:20:16,780
Now the point I'm making is that there's no obvious reason why we have to go with so on that.

174
00:20:16,780 --> 00:20:20,260
So here's how I think is the simplest way to oppose.

175
00:20:20,260 --> 00:20:26,320
So it's not to get into all sorts of nuances about philosophy of mind and all that kind of stuff.

176
00:20:26,320 --> 00:20:37,150
But just let's draw a distinction between information processing and its characteristics and the extent to which that processing is conscious.

177
00:20:37,150 --> 00:20:42,580
We could go further. The former category we could divide into information processing,

178
00:20:42,580 --> 00:20:51,160
which is embodied in such as such a way is to have direct queasy semantic connexions with the relevant subject matter.

179
00:20:51,160 --> 00:20:56,440
On the other hand, information processing, which requires interpretation by some external agent.

180
00:20:56,440 --> 00:21:05,980
So if we take the tutoring system, the system that tutors about chemistry that isn't embodied OK,

181
00:21:05,980 --> 00:21:12,710
in order to interpret its output as having anything to do with chemistry, it requires an agent doing it.

182
00:21:12,710 --> 00:21:21,460
Where is the robotic lumberjack there? The information processing is embedded in the system through the sensors and the motors and all the rest.

183
00:21:21,460 --> 00:21:25,870
So that has direct. I've called them crazy semantic connexions.

184
00:21:25,870 --> 00:21:29,650
I don't want to argue about the word semantic. They are connexions there.

185
00:21:29,650 --> 00:21:41,200
They link to the outside world. And that is a characteristic of that information processing, irrespective of whether it's conscious so distinguished.

186
00:21:41,200 --> 00:21:46,060
In this way, we could have information processing that's conscious and embodied.

187
00:21:46,060 --> 00:21:53,620
We have that higher animals have that unconscious and embodied the robotic lumberjack.

188
00:21:53,620 --> 00:21:58,300
And here I don't mean unconscious in the sense that Searle is talking about with our unconscious states.

189
00:21:58,300 --> 00:22:02,200
I mean, always unconscious, never conscious and embodied.

190
00:22:02,200 --> 00:22:06,820
So that's the robotic lumberjack and then unconscious and second hand.

191
00:22:06,820 --> 00:22:13,810
So that would be when we use artificial intelligence systems, we interpret the output in particular ways.

192
00:22:13,810 --> 00:22:19,750
So, so effectively, you're saying, well, we ought to restrict intelligence to the first of those.

193
00:22:19,750 --> 00:22:27,040
And I'm saying, why do you have a good reason for refusing to call these intelligent?

194
00:22:27,040 --> 00:22:33,070
Because after all, if you think about it, suppose you've got a number of robotic lumberjacks.

195
00:22:33,070 --> 00:22:40,600
Some of those might be better than others. Some might be extremely undiscriminating, you know, put them in the wrong place,

196
00:22:40,600 --> 00:22:44,560
and they're just going to hack down everything, whether it requires preservation or not.

197
00:22:44,560 --> 00:22:49,900
Others may be very discriminating. Well, don't we want to say that the latter is more intelligent than the former?

198
00:22:49,900 --> 00:22:57,280
Maybe we do. Likewise, you can have artificial intelligence systems that are better or worse at the information processing.

199
00:22:57,280 --> 00:23:04,390
Why can't we use the word intelligent genuinely to distinguish the ones that are good at it?

200
00:23:04,390 --> 00:23:14,590
Is there any reason other than prejudice for saying we're only going to count things as intelligent if they are conscious?

201
00:23:14,590 --> 00:23:22,960
A reminder now about something we saw a couple of lectures ago. Turing's response to Jefferson's argument from consciousness.

202
00:23:22,960 --> 00:23:27,880
Turing seems to accept that to support mean machine intelligence,

203
00:23:27,880 --> 00:23:34,780
he mistake machine consciousness to be a reasonable supposition, just as much with other people.

204
00:23:34,780 --> 00:23:45,220
So he he is actually here taking for granted what it seems to me Searle is trying to insist on.

205
00:23:45,220 --> 00:23:51,850
And I criticised him at the time, you may remember. I don't think Turing is right here.

206
00:23:51,850 --> 00:23:57,670
He says this argument, the argument from consciousness appears to be a denial of the validity of our test,

207
00:23:57,670 --> 00:24:03,760
according to the most extreme form of his view. The only way to know that either a machine or a man thinks is to be that particular man.

208
00:24:03,760 --> 00:24:08,290
It is, in fact, solipsistic point of view. So you may remember that.

209
00:24:08,290 --> 00:24:16,510
So Turing seems to be implying that in order to counter the argument from consciousness,

210
00:24:16,510 --> 00:24:23,620
he he has to appeal to the difficulty of establishing that other humans are conscious and then sort of saying,

211
00:24:23,620 --> 00:24:31,090
Well, we should apply the same presumption in favour of computers.

212
00:24:31,090 --> 00:24:34,810
He goes on to give the voter vote sample about the summit.

213
00:24:34,810 --> 00:24:39,400
You know, what would Jefferson say if the sonnet writing machine was able to answer like this in the Bible?

214
00:24:39,400 --> 00:24:48,010
Okay, if the answers were satisfactory and sustained, as in the above passage, I suggested back in slide 194,

215
00:24:48,010 --> 00:24:56,500
what Turing should have said is then there would be reason to call the machine intelligent, irrespective of whether or not it has genuine feelings.

216
00:24:56,500 --> 00:25:03,790
Intelligent thinking need not require consciousness, nor even potential consciousness.

217
00:25:03,790 --> 00:25:10,930
That's what I think he should have said. So let me now say more on that.

218
00:25:10,930 --> 00:25:16,600
So most discussions of these sorts of debates seem to proceed on the assumption that

219
00:25:16,600 --> 00:25:23,410
we should be looking for a verdict based on our current conceptual repertoire,

220
00:25:23,410 --> 00:25:29,920
our ordinary common sense conceptual repertoire that we grow up with.

221
00:25:29,920 --> 00:25:34,930
And I want to suggest that's wrong. That's not the way science works.

222
00:25:34,930 --> 00:25:39,740
If. Science at work like that, it would never have gone forward in the way it has.

223
00:25:39,740 --> 00:25:49,790
And I want to suggest the same applies to philosophy when we are considering a novel puzzle cases in philosophy.

224
00:25:49,790 --> 00:25:56,810
We ought to be prepared to expand our conceptual repertoire to deal with them because when we reach,

225
00:25:56,810 --> 00:26:06,140
when we meet new challenges, one of the things those challenges do is precisely to contest the traditional boundaries.

226
00:26:06,140 --> 00:26:12,380
I say more about this in a paper. I refer to their philosophical significance of the Turing machine in the Turing test.

227
00:26:12,380 --> 00:26:22,600
It was published in that that big collection of 2013 that I mentioned at the start of the lectures.

228
00:26:22,600 --> 00:26:27,910
So let's go back to Turing's prediction that we talked about this last time.

229
00:26:27,910 --> 00:26:33,220
The original question can machines think, I believe, to be too meaningless to deserve discussion?

230
00:26:33,220 --> 00:26:40,240
Nevertheless, I believe the end of the century the use of words and general educated opinion will have altered

231
00:26:40,240 --> 00:26:47,350
so much that one will be able to speak of machines thinking without expecting to be contradicted.

232
00:26:47,350 --> 00:26:58,450
So I'm suggesting that we take this as a hint, and this leads in a different direction from Turing's response to Jefferson saying,

233
00:26:58,450 --> 00:27:04,750
Oh, let's accept the fact that we're talking here about evolution of our concepts,

234
00:27:04,750 --> 00:27:16,700
so we don't need to go along with Jefferson in just identifying intelligence and consciousness or assuming that one requires the other.

235
00:27:16,700 --> 00:27:26,840
And an important point here, I want to make against a thought experiment or against their uses intuition pumps,

236
00:27:26,840 --> 00:27:29,600
which we've been looking at in the last couple of lectures,

237
00:27:29,600 --> 00:27:38,720
we saw Dennett discussing these very contemptuous into misleading or simplistic comparisons because

238
00:27:38,720 --> 00:27:48,690
precisely they appeal to our intuitive assumptions and that can drown out unexpected differences.

239
00:27:48,690 --> 00:27:54,630
So the problem we thought experiment is they are always appealing to something familiar

240
00:27:54,630 --> 00:28:00,210
and inviting us to see something novel as relevantly similar to the familiar thing.

241
00:28:00,210 --> 00:28:07,410
So let's make judgements accordingly. No, let's not. When we are faced with new realities,

242
00:28:07,410 --> 00:28:20,490
sometimes we have to think those through in novel ways and not assume the categories that we brought with us will apply to them satisfactorily.

243
00:28:20,490 --> 00:28:26,730
And I want to suggest that major scientific progress is very often involve these kinds of major novelties,

244
00:28:26,730 --> 00:28:34,290
bringing what are called paradigm shifts and fundamental new methods of explanation.

245
00:28:34,290 --> 00:28:46,260
And just as a matter of interest, before I get onto my rapid run through the history of science, it's very interesting to note that Luciano Floride,

246
00:28:46,260 --> 00:28:56,340
like me, sees Turing's discovery in 1936 as a sort of conceptual revolution, but he interprets it in a different way.

247
00:28:56,340 --> 00:29:05,790
I'm going to be talking about modes of explanation in science. He talks about humanity's fundamental nature and role in the universe.

248
00:29:05,790 --> 00:29:10,710
We are not immobile at the centre of the universe. That's what Copernicus showed.

249
00:29:10,710 --> 00:29:16,240
Personally, I'm a little bit more inclined to give the credit to Galileo. Copernicus came up with the theory.

250
00:29:16,240 --> 00:29:23,010
Galileo proved it with this telescope. We are not unnaturally distinct and different from the rest of the animal world.

251
00:29:23,010 --> 00:29:27,870
Darwin and we are far from being entirely transparent to ourselves.

252
00:29:27,870 --> 00:29:33,570
Freud here I would give more credit to David Hume.

253
00:29:33,570 --> 00:29:40,740
I would one day. But long before Freud, Hume had drawn this conclusion.

254
00:29:40,740 --> 00:29:49,170
I think Hume's arguments are a lot more solid than Freud psychology, but anyway, similar kind of point.

255
00:29:49,170 --> 00:29:57,210
We are now slowly accepting the idea that we might be informational organisms amongst many agents, Turing.

256
00:29:57,210 --> 00:30:07,240
So as I say, I'm just going to very quickly give a similar sort of rapid history, except in terms of modes of explanation.

257
00:30:07,240 --> 00:30:17,430
OK, so if you go back before 6500, before Galileo, more or less everything in the world is explained in terms of purpose.

258
00:30:17,430 --> 00:30:27,690
God's purpose, of course, God has designed the world in the way it is, but other things to act according to purpose.

259
00:30:27,690 --> 00:30:35,160
So obviously, humans, obviously animals, but actually also physical objects.

260
00:30:35,160 --> 00:30:39,630
If I take a stone and I drop it, the stone falls downwards.

261
00:30:39,630 --> 00:30:44,730
The reason it falls downwards is that it's striving to reach its natural place in the universe.

262
00:30:44,730 --> 00:30:48,540
Its natural place in the universe is at the centre of the universe.

263
00:30:48,540 --> 00:31:00,150
There are four main elements beneath the moon earth, water, air and fire, and they occupy places going outwards from the centre of the universe.

264
00:31:00,150 --> 00:31:07,890
So a stone being made mainly of Earth naturally moves towards its natural place, which is the centre of the universe.

265
00:31:07,890 --> 00:31:12,690
So stones are actually striving to achieve a certain purpose.

266
00:31:12,690 --> 00:31:19,540
So more or less, you know, all of science is based on purposes.

267
00:31:19,540 --> 00:31:24,760
Galileo turns his telescope up, shows that the Earth is not the centre of the universe, right?

268
00:31:24,760 --> 00:31:31,810
Yes. All right. That is perhaps a bit damaging to human hubris, as Florida points out.

269
00:31:31,810 --> 00:31:37,360
But I think more fundamentally, it completely scuppered the Aristotelian explanation.

270
00:31:37,360 --> 00:31:41,890
The whole Aristotelian physics goes out of the window because if the Earth is not the centre of the universe,

271
00:31:41,890 --> 00:31:48,670
then you can't explain the motion of rocks and things by their striving to reach the centre of the universe.

272
00:31:48,670 --> 00:31:54,940
So we get a completely new physics based on the idea of inertia that things naturally

273
00:31:54,940 --> 00:32:02,320
move at a constant speed in straight lines unless acted upon by some force.

274
00:32:02,320 --> 00:32:06,160
And gravity remains a big problem in those terms.

275
00:32:06,160 --> 00:32:14,260
Descartes comes up with a nice explanation in terms of the essence of matter being extension and therefore the universe working with vortices.

276
00:32:14,260 --> 00:32:25,600
Newton more successfully postulate a completely new way in which science can work, not explaining things in terms of physical contact.

277
00:32:25,600 --> 00:32:30,010
One thing bashing another, which is what Descartes had tried to do,

278
00:32:30,010 --> 00:32:38,020
but instead postulating a force that acts according to certain mathematical properties proportional

279
00:32:38,020 --> 00:32:43,030
to the the masses concerned and inversely proportional to the square of the distance.

280
00:32:43,030 --> 00:32:56,780
That's a new kind of explanation in science. In the 18th century, Hume throws doubt on traditional conceptions of human behaviour.

281
00:32:56,780 --> 00:33:02,000
Actually, humans are not nearly as based on reason as they like to think.

282
00:33:02,000 --> 00:33:05,540
Moreover, more importantly, they could not be they could not be.

283
00:33:05,540 --> 00:33:10,220
If you try to follow the ideal of Aristotle, for example,

284
00:33:10,220 --> 00:33:16,640
and Plato and so forth of living a life purely according to reason, you won't be able to do anything.

285
00:33:16,640 --> 00:33:25,640
So Hume is pointing out that animal instinct and feeling plays a much bigger role,

286
00:33:25,640 --> 00:33:31,010
and that is obviously putting us closer to the animals rather than to gods.

287
00:33:31,010 --> 00:33:38,360
And indeed, Hume argues that explicitly. Charles Darwin, of course, came along simultaneously with with with Wallace.

288
00:33:38,360 --> 00:33:45,560
I mean, Darwin published his theory in the same year as Wallace and was prompted to do so by Wallace.

289
00:33:45,560 --> 00:33:58,820
His theory of evolution biological organisms take the form they do not because of divine design, but because of evolution and natural selection.

290
00:33:58,820 --> 00:34:05,960
So we inherit characteristics from our parents or organisms do.

291
00:34:05,960 --> 00:34:15,440
And those characteristics that are better suited to our survival and reproduction naturally get better represented in the next generation.

292
00:34:15,440 --> 00:34:20,240
So you get evolution because this doesn't imply that there isn't a god behind it all.

293
00:34:20,240 --> 00:34:30,470
But the crucial point is it shows that a phenomenon which previously would incontrovertibly have been ascribed to intelligent design,

294
00:34:30,470 --> 00:34:36,470
namely the exquisite adaptation of organisms to their environment now has an

295
00:34:36,470 --> 00:34:46,930
alternative explanation doesn't require conscious purpose to get exquisite adaptation.

296
00:34:46,930 --> 00:34:49,900
Moving on to our series of revolutions again, I mean,

297
00:34:49,900 --> 00:34:56,170
do notice that both human Darwin there are bringing in new methods of explanation in science before Darwin.

298
00:34:56,170 --> 00:35:01,270
Right? An evolutionary explanation wasn't there after Darwin.

299
00:35:01,270 --> 00:35:09,070
We've got a new method of explaining things. General relativity brings in another one to do with curving space time.

300
00:35:09,070 --> 00:35:12,160
Quantum mechanics brings in yet another.

301
00:35:12,160 --> 00:35:24,580
So these major happenings in the history of science, all of them are bringing in new methods of explanation going beyond the natural,

302
00:35:24,580 --> 00:35:33,470
intuitive explanation in terms of purpose, which seems to come in a most easily to us.

303
00:35:33,470 --> 00:35:41,630
And they're bringing in fundamental changes in thinking they do change how we see our place in the universe, as Florida pointed out.

304
00:35:41,630 --> 00:35:49,190
They also extend our understanding of the possibilities of scientific explanation mathematical forces rather than strivings,

305
00:35:49,190 --> 00:35:56,240
inherited variation and selection rather than rational design curvature in spacetime rather than Newtonian forces.

306
00:35:56,240 --> 00:36:03,650
And I want to point out that the philosopher sorry, the scientists who made these big advances were very,

307
00:36:03,650 --> 00:36:06,680
very strongly influenced by philosophical thinking.

308
00:36:06,680 --> 00:36:14,050
I mean, the backing Galileo and Descartes and Boyle's time, and so they styled themselves natural philosophers.

309
00:36:14,050 --> 00:36:18,170
But it's interesting that Darwin was very influenced by Hume.

310
00:36:18,170 --> 00:36:22,430
He was reading Hume on the reason of animals at the time he came up with the theory of evolution.

311
00:36:22,430 --> 00:36:31,850
He'd read lots of stuff on him. Einstein, rather nicely as well, credited Hume with the insight that led to the theory of relativity.

312
00:36:31,850 --> 00:36:45,230
I think that's particularly nice. You know, two centuries later and thinking in philosophical ways is prompting new thinking in science.

313
00:36:45,230 --> 00:36:48,620
Now, going back to the physics again,

314
00:36:48,620 --> 00:36:57,800
but relativity and quantum mechanics particularly illustrate how wrong our intuitive understanding of things can be.

315
00:36:57,800 --> 00:37:06,920
I think evolution has now got sufficiently into our general consciousness that an educated people no longer find that particularly strange relativity,

316
00:37:06,920 --> 00:37:13,250
curvature of space, time and quantum mechanics are seriously weird.

317
00:37:13,250 --> 00:37:21,690
And I think anybody who isn't trained very heavily in physics is going to find those weird and.

318
00:37:21,690 --> 00:37:24,510
Yet these are important conceptual advances.

319
00:37:24,510 --> 00:37:32,970
OK, science is advanced by showing that there's something wrong or inadequate in our intuitive understanding.

320
00:37:32,970 --> 00:37:46,830
And I want to suggest that Alan Turing's discovery of the Turing machine is comparable with those kind of innovations.

321
00:37:46,830 --> 00:37:53,400
Obviously, he and girdle and so on have shown that some mathematical questions for which there's no possible method of solution.

322
00:37:53,400 --> 00:37:58,590
So, you know, the the advances that came in the 1930s in philosophy of maths,

323
00:37:58,590 --> 00:38:07,650
those shook very fundamental assumptions that had hitherto been more or less universal,

324
00:38:07,650 --> 00:38:19,020
but also obviously more pertinent here with regard to the nature of thinking, we need to be open to the possibility of inanimate thought.

325
00:38:19,020 --> 00:38:24,600
So what Turing showed is that information processing can be understood in terms of symbolic inputs

326
00:38:24,600 --> 00:38:30,310
and outputs governed by explicit and automatic processes with a limited range of operations.

327
00:38:30,310 --> 00:38:34,620
OK, we saw that in going through his 1936 paper.

328
00:38:34,620 --> 00:38:38,520
We've got the Turing machine, very limited repertoire of possibilities.

329
00:38:38,520 --> 00:38:46,860
And Turing argues that anything that can be calculated systematically can be calculated by such a machine.

330
00:38:46,860 --> 00:39:00,130
And the Church Turing thesis backs that up. So information and information processing in this sense does not presuppose an understanding mind.

331
00:39:00,130 --> 00:39:10,600
So we now have sophisticated information processing machines capable of doing anything that we can conceive how to do in a systematic way,

332
00:39:10,600 --> 00:39:15,850
but they can do it without any conscious purpose at all.

333
00:39:15,850 --> 00:39:20,110
And I want to suggest that that's somewhat similar to Darwin's innovation.

334
00:39:20,110 --> 00:39:24,580
Darwin showed that you can have sophisticated adaptation without intentional design.

335
00:39:24,580 --> 00:39:32,680
Turing showed that you can have information processing without that information existing in a conscious mind,

336
00:39:32,680 --> 00:39:39,880
and that is a novelty, OK, before computers came along.

337
00:39:39,880 --> 00:39:49,690
Words like intelligent were only associated with conscious things that had purposes agents applying their intelligence to the situation.

338
00:39:49,690 --> 00:39:53,170
They understand themselves to be in performing sophisticated processing of

339
00:39:53,170 --> 00:39:58,870
information in order to adapt their behaviour to achieve their own purposes.

340
00:39:58,870 --> 00:40:05,410
By contrast, inanimate things just don't have purposes. So there's no question of them having intelligence.

341
00:40:05,410 --> 00:40:11,290
But then what we have since Turing is intelligent processing of information,

342
00:40:11,290 --> 00:40:17,110
apparently by things that are inanimate, that don't have their own purposes.

343
00:40:17,110 --> 00:40:22,990
So what I want to suggest is when we're faced with that, what we should do is modify our concepts.

344
00:40:22,990 --> 00:40:35,680
We shouldn't try to apply our pre-existing naive concepts to this new reality, just like with Darwin and Einstein and quantum mechanics and so on.

345
00:40:35,680 --> 00:40:42,180
We have to invent new concepts. We have to divide things in new ways.

346
00:40:42,180 --> 00:40:50,070
Now, there's a general point to be made actually about lots of our concepts, not just the special scientific ones.

347
00:40:50,070 --> 00:40:59,190
Friedrich Weissman, who's a sort of follower of and Stein he set himself to write up in relatively clear language.

348
00:40:59,190 --> 00:41:07,690
What he took to be the outcome of Vik and Stein's work. Of course, as a result, Vik and Stein hated it.

349
00:41:07,690 --> 00:41:12,600
He came up with the very nice notion of open texture. So it's a kind of tinian notion.

350
00:41:12,600 --> 00:41:16,860
But it's the idea is that most of our concepts are open, textured.

351
00:41:16,860 --> 00:41:21,750
It's not clear in advance how we would apply them to all possible cases.

352
00:41:21,750 --> 00:41:27,870
Now, open texture is actually particularly prominent as a concept in law because in law,

353
00:41:27,870 --> 00:41:35,670
you can imagine you set out laws to prescribe what should happen in a particular situation, and it really matters, right?

354
00:41:35,670 --> 00:41:40,230
Penalties could depend on this or what people are allowed to do depends on this.

355
00:41:40,230 --> 00:41:52,770
So the categories really matter. It's not just like philosophers, you know, playing their word games, and sometimes this can impact with reality.

356
00:41:52,770 --> 00:42:00,840
So suppose we have a society where marriage is understood as being between a man and a woman only no sex change is nothing like that.

357
00:42:00,840 --> 00:42:08,340
No, no, that doesn't happen. But then maybe things change and we do get sex changes.

358
00:42:08,340 --> 00:42:10,470
Or maybe we get new discoveries about, you know,

359
00:42:10,470 --> 00:42:17,310
chromosomal abnormalities and whatever we discover more about humans or humans are presented with new possibilities.

360
00:42:17,310 --> 00:42:22,440
Well, we have to wonder how the old law applies. We can't just apply it mindlessly.

361
00:42:22,440 --> 00:42:28,840
We have to ask ourselves, how should this law be adapted in this new situation?

362
00:42:28,840 --> 00:42:31,650
I mean, another example inheritance rights.

363
00:42:31,650 --> 00:42:40,230
Suppose you have a society in which children are always, always stay with their parents and inheritance is understood accordingly.

364
00:42:40,230 --> 00:42:47,940
But then adoption becomes common as a practise. Maybe because there's a nasty plague or something like that, a game like death.

365
00:42:47,940 --> 00:42:56,820
And suddenly you get totally new social practises. You have to ask how the traditional laws are going to be applied.

366
00:42:56,820 --> 00:43:02,880
And those of you who know about Theseus to shift the imagine Theseus, the Greek hero,

367
00:43:02,880 --> 00:43:09,360
has a ship which he is regularly getting repaired and he takes it back to the ship, right?

368
00:43:09,360 --> 00:43:16,080
Who built it and the ship right? Whenever he, whenever he replaces a plank, keeps the old one.

369
00:43:16,080 --> 00:43:26,640
And then some years later, when all the planks have been replaced, the ship right builds a new the initial ship with the original planks.

370
00:43:26,640 --> 00:43:35,040
Well, is that actually the same ship or not? Or should we say that the same ship is the one that Theseus has been sailing around it?

371
00:43:35,040 --> 00:43:41,670
Well, again, this could matter in a legal context because maybe CCC ship has mooring rights in perpetuity somewhere.

372
00:43:41,670 --> 00:43:46,380
So if the ship right now turns up with the ship and says, actually, this is the sister ship,

373
00:43:46,380 --> 00:43:53,460
this is the one that has the mooring rights, well, does it or not? Probably not.

374
00:43:53,460 --> 00:44:00,630
I rather like this punch cartoon from 1869. I think it's a nice example of open texture.

375
00:44:00,630 --> 00:44:04,920
So it's a woman with a menagerie of creatures wanting to go on the train.

376
00:44:04,920 --> 00:44:14,340
Right? And the the port is, as station master says, mum of cats is dogs and rabbits is dogs and those parrots.

377
00:44:14,340 --> 00:44:20,220
But this year, tortoise is an insect, so there ain't no charge for it.

378
00:44:20,220 --> 00:44:25,020
So the porter knows how to charge for dogs and people, but but not for an insect.

379
00:44:25,020 --> 00:44:31,200
You don't charge for insects, but he doesn't know how to charge for cats and rabbits and parrots and tortoises,

380
00:44:31,200 --> 00:44:37,320
so they have to be categorised under the existing categories.

381
00:44:37,320 --> 00:44:43,290
OK, now open texture cut in two directions, first of all.

382
00:44:43,290 --> 00:44:45,960
And this, by the way, at least Weisman claims,

383
00:44:45,960 --> 00:44:54,090
I think quite plausibly is something a phenomenon that applies to a lot of our concepts, not just special ones.

384
00:44:54,090 --> 00:45:06,390
You know, arguably most of our concepts are derived from our experience in the world and not necessarily defined beyond that.

385
00:45:06,390 --> 00:45:11,160
We can't expect our concepts to be prepared in advance for all new eventualities.

386
00:45:11,160 --> 00:45:19,470
They may have to be revised or tuned to new contexts, and especially, of course, when big new innovations come.

387
00:45:19,470 --> 00:45:27,420
And another point is we needn't accept any requirement when reshooting our concepts to make them immune to future revision.

388
00:45:27,420 --> 00:45:32,760
We don't have to take all future possibilities, let alone all logical possibilities into account.

389
00:45:32,760 --> 00:45:40,170
So philosophers actually have a good reason when somebody says, Look, before we discuss X knowledge,

390
00:45:40,170 --> 00:45:45,480
personal identity, whatever it might be, we've got to define it precisely.

391
00:45:45,480 --> 00:45:53,270
Answer sorry, can't do it as the experience of philosophy shows.

392
00:45:53,270 --> 00:46:02,510
And a point I want to draw from that is when revising our concepts, we have every right to ignore crazy thought experiments.

393
00:46:02,510 --> 00:46:10,940
That's a controversial claim. I'm saying if it's true that in general, our concepts are tuned to the world as we experience it,

394
00:46:10,940 --> 00:46:15,830
then imagining some world in which things are wildly different and being asked,

395
00:46:15,830 --> 00:46:19,010
How is your concept going to apply there should you not be changing it?

396
00:46:19,010 --> 00:46:23,960
My answer is going to be no. Maybe I need to tune it to realities.

397
00:46:23,960 --> 00:46:36,380
Maybe I need to tune it to potential realities. But to impossibilities or complete implausibility is no, I'm not going to accept that obligation.

398
00:46:36,380 --> 00:46:42,230
OK, against that background. Let's now ask ourselves how in general, we judge intelligence,

399
00:46:42,230 --> 00:46:50,750
so I've argued that we should be prepared to consider intelligent things that are unconscious.

400
00:46:50,750 --> 00:46:59,150
I have acknowledged that our everyday concept of intelligence, the one that we and our parents and our parents,

401
00:46:59,150 --> 00:47:06,500
parents and our parents parents have grown up with has evolved to fit a world or being, you know,

402
00:47:06,500 --> 00:47:14,210
created to fit a world in which the only intelligent things are conscious, purposive creatures.

403
00:47:14,210 --> 00:47:26,240
I'm saying we now have a new reality. Apparently, intelligent processing going on within things that are not conscious or purposive.

404
00:47:26,240 --> 00:47:33,230
How are we going to judge those? OK, how do we think about intelligence in ordinary life?

405
00:47:33,230 --> 00:47:37,640
Well, first of all, its subject relative? Yeah.

406
00:47:37,640 --> 00:47:46,790
Many of us here would be considered very intelligent about some things we're probably considered very far from intelligent about some others.

407
00:47:46,790 --> 00:47:51,260
It's a matter of degree. We don't say some things intelligent or unintelligent.

408
00:47:51,260 --> 00:47:55,400
We say some things more intelligent or person is more intelligent than another.

409
00:47:55,400 --> 00:48:00,590
And hopefully, you know, when we study here, we get more intelligent and so forth.

410
00:48:00,590 --> 00:48:08,150
It's generally measured by performance, including flexibility, speed, appropriateness of response to new requirements or new information.

411
00:48:08,150 --> 00:48:16,250
So someone who's able to adapt to new things to learn new skills in relation to some domain,

412
00:48:16,250 --> 00:48:22,520
we think of them as more intelligent than someone who's simply been trained to deal with a fixed repertoire.

413
00:48:22,520 --> 00:48:35,040
It's not significantly correlated with feeling or consciousness. We don't judge someone as more intelligent because they Campbell agree.

414
00:48:35,040 --> 00:48:40,740
So suppose we now distinguish sharply between the sophistication of information

415
00:48:40,740 --> 00:48:45,120
processing and the phenomenology phenomenology is a technical philosopher's term,

416
00:48:45,120 --> 00:48:47,790
meaning what it's like from the inside, right?

417
00:48:47,790 --> 00:48:56,160
So we're going to distinguish between sophistication of information processing on one hand and the consciousness, the inner life on the other.

418
00:48:56,160 --> 00:49:00,900
They often come apart. I mean, we've already seen hints of some of this.

419
00:49:00,900 --> 00:49:06,270
Dogs can decide desire things as strongly as we do. Yes.

420
00:49:06,270 --> 00:49:14,100
I mean, anyone who's seen a dog with a ball coming up to you and holding its mouth looking appealingly, please, please throw the ball.

421
00:49:14,100 --> 00:49:18,350
I want to play. Right? Dogs can desire things very, very strongly.

422
00:49:18,350 --> 00:49:21,750
That doesn't make us judge them extra intelligent.

423
00:49:21,750 --> 00:49:28,530
I'm pretty intelligent compared to most animals, but it's not in proportion to the strength of their desires.

424
00:49:28,530 --> 00:49:32,100
Experts are often less conscious than novices.

425
00:49:32,100 --> 00:49:39,330
So, for example, expert drivers who are used to commuting along a particular road day after day after day,

426
00:49:39,330 --> 00:49:44,940
it's absolutely familiar that they can come to the end of their journey and walk into work.

427
00:49:44,940 --> 00:49:51,780
And somebody says, How was the drive today? I can't remember unless something special happened.

428
00:49:51,780 --> 00:49:55,110
You just drive along. I mean, you're obviously taking note of things.

429
00:49:55,110 --> 00:50:04,530
You're avoiding, you know, pedestrians, you're taking note of traffic lights, you're overtaking and being overtaken and, you know, driving safely.

430
00:50:04,530 --> 00:50:11,460
But you're almost I mean, you wouldn't say you're unconscious while you're doing it, but it barely impacts on your consciousness,

431
00:50:11,460 --> 00:50:20,400
whereas a novice driver someone is learning to drive is going to be aware of every single little thing.

432
00:50:20,400 --> 00:50:27,480
The same is true, for example, chess players. It's very frustrating actually talking to an expert chess player and sometimes and saying,

433
00:50:27,480 --> 00:50:32,340
Why did you play that move, just the right move to play in that kind of position?

434
00:50:32,340 --> 00:50:39,990
Thanks very much. Where is the novice will be able to tell you exactly, Oh, I thought this and this and this and this and this.

435
00:50:39,990 --> 00:50:45,960
Actually, the more expert we are, often the less conscious we become another.

436
00:50:45,960 --> 00:50:51,900
I think corroborative piece of evidence. Our intuitive judgements about what is easy are often badly wrong.

437
00:50:51,900 --> 00:50:56,430
So most people think arithmetic is really, really difficult.

438
00:50:56,430 --> 00:51:02,460
We know, you know, from computers, arithmetic is comparatively trivial in terms of information processing,

439
00:51:02,460 --> 00:51:06,240
whereas running and catching a ball, that's seriously difficult.

440
00:51:06,240 --> 00:51:12,330
You know, you've got a balance and you've got to predict where the ball or anticipate where the ball's going to be.

441
00:51:12,330 --> 00:51:18,180
You've got to reach out your hand appropriately and so forth. That's seriously difficult.

442
00:51:18,180 --> 00:51:25,050
And yet we tend to think of it as easy. So our intuitive judgements about information processing tend to be pretty bad.

443
00:51:25,050 --> 00:51:33,670
But dogs can do that pretty well. So in that respect, dogs are very intelligent.

444
00:51:33,670 --> 00:51:38,590
OK, suppose we distinguish sophistication of information processing from the phenomenology,

445
00:51:38,590 --> 00:51:43,980
I'm going to suggest it's clear that intelligence is far more a measure of the former than the latter.

446
00:51:43,980 --> 00:51:52,000
But all all of the evidence I've given in the last two slides suggest that when we judge something's intelligence,

447
00:51:52,000 --> 00:52:00,100
it's mainly a focus on the information processing, the power, the sophistication, the flexibility and so that the information processing.

448
00:52:00,100 --> 00:52:07,690
It's not the phenomenology. And in our new world of unconscious but highly sophisticated information processes,

449
00:52:07,690 --> 00:52:14,980
it makes sense to allow our concept of intelligence to evolve a court accordingly.

450
00:52:14,980 --> 00:52:23,350
So I think Turing's main claim in the paper is vindicated, but not by means of the test that he proposed.

451
00:52:23,350 --> 00:52:28,090
That is, I think, learning from the 1936 paper.

452
00:52:28,090 --> 00:52:37,450
We should accept that you can have intelligent computers.

453
00:52:37,450 --> 00:52:46,510
Coming back now to Turing's approach to this and what I call his solipsistic mistake, I mean, he didn't draw the distinction that I'm drawing.

454
00:52:46,510 --> 00:52:51,970
And I think thus he that weakened his argument and he talked about how.

455
00:52:51,970 --> 00:53:04,120
So if we followed the logical conclusion of Jefferson's objection, we would actually be reduced to solipsism denying consciousness to other humans.

456
00:53:04,120 --> 00:53:11,290
I've got a very different way. I've said I fully accept that computers aren't conscious. That doesn't stop them being intelligent.

457
00:53:11,290 --> 00:53:21,460
I don't want to go along with Turing in saying that if I deny consciousness to computers, I have to deny them to my fellow human beings.

458
00:53:21,460 --> 00:53:31,690
Those are very different. We know, at least in outline, why the robot does what it does.

459
00:53:31,690 --> 00:53:44,200
Suppose take a simple example, chess computer, I know in general terms what why the chess computer makes the moves that it does.

460
00:53:44,200 --> 00:53:48,070
I've programmed games, playing programmes. I've read about them.

461
00:53:48,070 --> 00:53:56,980
I know the kinds of algorithms that I know that the computer is doing what he's done because it's been programmed to do it.

462
00:53:56,980 --> 00:54:06,010
I don't have to postulate a desire, a conscious desire on the computer's part because I know full well, a conscious desire plays no part.

463
00:54:06,010 --> 00:54:09,970
Whatever in the causation of its doing its move.

464
00:54:09,970 --> 00:54:20,800
So I might say, Oh, it moved 96, because that was only the the only way of avoiding Queen H7 meant right.

465
00:54:20,800 --> 00:54:25,270
But I'm not attributing a conscious purpose with that.

466
00:54:25,270 --> 00:54:36,490
So I may be applying an intentional description to it, but I'm not under the illusion that there's some consciousness that we all the humans.

467
00:54:36,490 --> 00:54:38,770
It's quite different.

468
00:54:38,770 --> 00:54:48,070
I know that other humans function biologically in broadly the same way as I do, and I know that in my life, consciousness plays an important role.

469
00:54:48,070 --> 00:55:00,290
I'm aware of it. I'm aware of my conscious purposes and how they impact on what I do, a very good reason to think that other people are much the same.

470
00:55:00,290 --> 00:55:04,880
Moreover, the phenomenal reality of consciousness, the fact that we are aware of it,

471
00:55:04,880 --> 00:55:11,060
that we do ponder things consciously and that they affect how we behave strongly suggests it's something causally active.

472
00:55:11,060 --> 00:55:16,070
It's not just an abstraction of information processing patterns.

473
00:55:16,070 --> 00:55:27,620
So some people speculate that if you had a computer that reasoned in the same way as a human, then that would ipso facto make it conscious.

474
00:55:27,620 --> 00:55:33,320
I don't see any reason for saying that. It seems to me consciousness has a reality.

475
00:55:33,320 --> 00:55:40,520
It's terribly difficult to pin it down or to understand it. But it's not just a matter of information processing.

476
00:55:40,520 --> 00:55:48,650
The evolutionary role of it is quite mysterious. Those philosophers like postulating the idea of zombies.

477
00:55:48,650 --> 00:55:56,480
A zombie is a being that is physically the same as a human being.

478
00:55:56,480 --> 00:56:04,700
But with nobody at home, why wouldn't that be possible? Being with all the same neurones as me, but no awareness.

479
00:56:04,700 --> 00:56:08,930
Wouldn't they function physically in exactly the same way? Wouldn't they be just as successful?

480
00:56:08,930 --> 00:56:16,910
In which case, what difference does consciousness make? Well, I want to suggest that the close correlation between consciousness,

481
00:56:16,910 --> 00:56:26,030
the pleasures and pains that we feel and factors affecting our well-being is overwhelming evidence that consciousness is causally active.

482
00:56:26,030 --> 00:56:37,130
The fact that we get pleasure from things like eating honey say which in the environment where we evolved, honey is valuable.

483
00:56:37,130 --> 00:56:44,350
It's got calories in and getting calories, unlike in modern society 100000 years ago, is pretty difficult.

484
00:56:44,350 --> 00:56:51,110
You get honey, you want it. It's nice or injuring yourself that that's that does you harm.

485
00:56:51,110 --> 00:56:57,140
That's painful. And there are loads and loads of respect in which, you know, I mean, eating food is generally pleasurable.

486
00:56:57,140 --> 00:57:07,070
Sex is generally pleasurable. The things that conduce to our evolutionary success are generally pleasurable and the things that harm us are not.

487
00:57:07,070 --> 00:57:21,170
But that's not a coincidence. But conscious states can only track things that are good for us and bad for us if they are causally efficacious.

488
00:57:21,170 --> 00:57:22,760
Consciousness could only evolve.

489
00:57:22,760 --> 00:57:30,380
I want to suggest in a way that is so well tuned to our needs, giving us desires for things that are evolutionarily beneficial,

490
00:57:30,380 --> 00:57:37,220
dislike of things that are harmful if it's cause negative, even if we're unable to work out.

491
00:57:37,220 --> 00:57:44,150
How so? I fully accept that philosophers have great difficulty making sense of consciousness.

492
00:57:44,150 --> 00:57:46,640
You know, there's lots of debate about at the moment.

493
00:57:46,640 --> 00:57:54,380
I a personal suspicion is that this isn't going to be sorted out until we have at least one more conceptual revolution.

494
00:57:54,380 --> 00:58:03,810
And I haven't a clue where that's going to come from. But, you know, I think it may be that trying to sort this all out in terms of current science,

495
00:58:03,810 --> 00:58:10,280
it is just futile because we've got to wait for better understanding of the brain.

496
00:58:10,280 --> 00:58:12,470
Much, much more investigation.

497
00:58:12,470 --> 00:58:22,820
You know, maybe into biochemistry or maybe physics, whatever it won't, I, I expect that, you know, in the next 500 years,

498
00:58:22,820 --> 00:58:32,030
whatever, there may be significant conceptual advances, which will allow us to tackle problems that at present, we can't tackle.

499
00:58:32,030 --> 00:58:38,090
But if I have to make a bet as it were, evolutionary theory is well understood.

500
00:58:38,090 --> 00:58:45,770
Supported by overwhelming empirical evidence, strong scientific evidence trumps armchair speculation.

501
00:58:45,770 --> 00:58:53,360
So there are lots of philosophers speculating in their different ways about consciousness, its role and all the rest.

502
00:58:53,360 --> 00:58:59,480
Where I'm going to plant my flag is I'm confident that consciousness has an

503
00:58:59,480 --> 00:59:08,360
evolutionary role is causally efficacious because of the evolutionary argument.

504
00:59:08,360 --> 00:59:13,100
OK, now we humans are all products of the same evolved biological processes.

505
00:59:13,100 --> 00:59:20,480
Know you were born in much the same way as I was and by similar processes.

506
00:59:20,480 --> 00:59:27,260
Even if we cannot work out how consciousness arises, we have very good reason to attribute it to each other.

507
00:59:27,260 --> 00:59:35,660
That argument doesn't apply at all to computers. So there is no reason for supposing computers to be conscious.

508
00:59:35,660 --> 00:59:36,590
But nevertheless,

509
00:59:36,590 --> 00:59:50,660
I have argued we have strong reason in the wake of Turing and his monumental discoveries or inventions to allow that computers can be intelligent.

510
00:59:50,660 --> 00:59:55,532
And that's it. Thank you.