WEBVTT

1
00:00:00.000 --> 00:00:00.640
<v Ian Gibbins>Well, good morning, everybody.

2
00:00:01.300 --> 00:00:05.080
<v Ian Gibbins>My name's Ian Gibbons, and I'm a neuroscientist from Flinders University.

3
00:00:06.820 --> 00:00:07.940
<v Ian Gibbins>And we've got...

4
00:00:08.600 --> 00:00:11.460
<v Ian Gibbins>First of all, I'd like to say there's two really great things about this morning.

5
00:00:11.580 --> 00:00:17.040
<v Ian Gibbins>First of all, we've got Mandiyam Srini-Vazan, known as Srini, here to talk to us.

6
00:00:17.100 --> 00:00:21.000
<v Ian Gibbins>And secondly, we've got a fantastic audience on a cold Sunday morning, which is great.

7
00:00:22.560 --> 00:00:25.500
<v Ian Gibbins>I'll just say a few words of introduction for Srini.

8
00:00:25.500 --> 00:00:30.160
<v Ian Gibbins>Srini, he's got an original background in engineering, electrical engineering,

9
00:00:30.960 --> 00:00:34.260
<v Ian Gibbins>and then engineering and applied science at Yale University.

10
00:00:35.140 --> 00:00:40.980
<v Ian Gibbins>But then he sort of made a switch, maybe a subtle switch, maybe a big switch,

11
00:00:41.080 --> 00:00:44.140
<v Ian Gibbins>maybe we'll talk about that, into looking at neuroscience,

12
00:00:44.980 --> 00:00:48.680
<v Ian Gibbins>how nerves work, how the brain works, potentially how the mind works.

13
00:00:49.000 --> 00:00:51.200
<v Ian Gibbins>But not in people, but in bees.

14
00:00:52.220 --> 00:00:55.260
<v Ian Gibbins>And Srini is extremely well known.

15
00:00:55.500 --> 00:01:01.340
<v Ian Gibbins>In the neuroscientific community in Australia, internationally, for his work on bees.

16
00:01:01.540 --> 00:01:08.260
<v Ian Gibbins>He's also been well publicised through the media, through the television shows like Quantum and Compass,

17
00:01:08.440 --> 00:01:09.920
<v Ian Gibbins>its precursor, and ABC.

18
00:01:13.120 --> 00:01:17.300
<v Ian Gibbins>Srini himself has been recognised by his peers and by the scientific community

19
00:01:17.300 --> 00:01:19.440
<v Ian Gibbins>for his contributions to neuroscience.

20
00:01:19.440 --> 00:01:25.120
<v Ian Gibbins>He's published extensively in absolutely the very best scientific journals in the world.

21
00:01:25.120 --> 00:01:28.500
<v Ian Gibbins>Not just those which specialise in neuroscience, but which do all of science.

22
00:01:28.640 --> 00:01:30.760
<v Ian Gibbins>Journals like Science and Nature.

23
00:01:32.280 --> 00:01:36.180
<v Ian Gibbins>That's a real pinnacle of what you can do as a scientist.

24
00:01:36.940 --> 00:01:42.060
<v Ian Gibbins>And he's been recognised with fellowships at the Australian Academy of Science and the Royal Society of London.

25
00:01:42.560 --> 00:01:45.340
<v Ian Gibbins>And this is like getting a knighthood in science.

26
00:01:45.620 --> 00:01:51.000
<v Ian Gibbins>It's like Srini of the bees in a scientific sort of context.

27
00:01:53.960 --> 00:01:54.520
<v Ian Gibbins>The...

28
00:01:54.520 --> 00:01:59.280
<v Ian Gibbins>And most recently, in 2006, he got the Australian Prime Minister's Science Prize,

29
00:01:59.480 --> 00:02:02.520
<v Ian Gibbins>which is, again, a great recognition of his general contributions.

30
00:02:03.760 --> 00:02:06.420
<v Ian Gibbins>There's just a couple of things I'd just like to say about...

31
00:02:06.420 --> 00:02:08.000
<v Ian Gibbins>I'm sure he's going to tell us why bees.

32
00:02:08.380 --> 00:02:11.380
<v Ian Gibbins>But as someone who doesn't work on bees, but who works on the nervous system,

33
00:02:12.340 --> 00:02:13.880
<v Ian Gibbins>there's some really key things.

34
00:02:13.980 --> 00:02:19.340
<v Ian Gibbins>And the basic biology of all animals is pretty much the same when it comes to the crunch.

35
00:02:19.500 --> 00:02:21.040
<v Ian Gibbins>You can see how the individual cells work.

36
00:02:21.180 --> 00:02:23.800
<v Ian Gibbins>It's all the same bits and pieces working in there.

37
00:02:23.920 --> 00:02:24.500
<v Ian Gibbins>But just put...

38
00:02:24.500 --> 00:02:25.760
<v Ian Gibbins>Put together in a slightly different way.

39
00:02:26.040 --> 00:02:29.180
<v Ian Gibbins>And even more recently, I suppose, over the last 20 years,

40
00:02:29.320 --> 00:02:31.540
<v Ian Gibbins>it's even been realised that the way things are put together,

41
00:02:31.600 --> 00:02:33.240
<v Ian Gibbins>the rules that are used to put things together,

42
00:02:33.400 --> 00:02:36.020
<v Ian Gibbins>are not that much different between any sort of animal.

43
00:02:36.440 --> 00:02:40.100
<v Ian Gibbins>And so maybe when Srini started, it might have been with a long shot,

44
00:02:40.160 --> 00:02:42.420
<v Ian Gibbins>saying that looking at bees was going to have something to do

45
00:02:42.420 --> 00:02:44.100
<v Ian Gibbins>with understanding our consciousness, our intelligence.

46
00:02:44.260 --> 00:02:50.660
<v Ian Gibbins>I think by the time he's finished, you will see that it's not that long a path at all.

47
00:02:51.320 --> 00:02:53.420
<v Ian Gibbins>The other thing which Srini's...

48
00:02:54.500 --> 00:02:57.260
<v Ian Gibbins>The other thing which his work has shown us is how pure research

49
00:02:57.260 --> 00:03:01.800
<v Ian Gibbins>can quickly and easily develop into unexpected practical spin-offs.

50
00:03:01.920 --> 00:03:04.360
<v Ian Gibbins>So I'm sure he's going to tell you something about some of that today

51
00:03:04.360 --> 00:03:06.080
<v Ian Gibbins>in terms of engineering, in this case,

52
00:03:07.160 --> 00:03:10.980
<v Ian Gibbins>different ways of looking at how you can make novel practical applications.

53
00:03:11.900 --> 00:03:14.200
<v Ian Gibbins>And the final point I'd like to make is, I suppose,

54
00:03:14.360 --> 00:03:16.240
<v Ian Gibbins>taking advantage of being the chair for a minute,

55
00:03:17.620 --> 00:03:21.280
<v Ian Gibbins>is that his sort of work, working on bees,

56
00:03:21.380 --> 00:03:22.160
<v Ian Gibbins>is a sort...

57
00:03:22.160 --> 00:03:24.420
<v Ian Gibbins>And the way it's moved into all sorts of...

58
00:03:24.500 --> 00:03:25.300
<v Ian Gibbins>other areas of interest

59
00:03:25.300 --> 00:03:28.960
<v Ian Gibbins>is the perfect retort or antidote

60
00:03:28.960 --> 00:03:31.300
<v Ian Gibbins>for some of those sort of characters

61
00:03:31.300 --> 00:03:33.800
<v Ian Gibbins>who seem to be over-represented in politics

62
00:03:34.860 --> 00:03:36.440
<v Ian Gibbins>who'd say, if you looked at a grant application

63
00:03:36.440 --> 00:03:40.360
<v Ian Gibbins>looking at something to do with what the brain of bees,

64
00:03:40.700 --> 00:03:44.780
<v Ian Gibbins>why are we wasting taxpayers' money on something as arcane and as silly as that?

65
00:03:45.280 --> 00:03:47.220
<v Ian Gibbins>As you will see by the end of Srini's talk,

66
00:03:47.340 --> 00:03:48.980
<v Ian Gibbins>it's not arcane, it's not silly.

67
00:03:49.160 --> 00:03:50.400
<v Ian Gibbins>Apart from anything else, it's fascinating

68
00:03:51.060 --> 00:03:53.200
<v Ian Gibbins>and it's going to have all sorts of practical spin-offs.

69
00:03:53.380 --> 00:03:54.280
<v Ian Gibbins>So with great pleasure,

70
00:03:54.320 --> 00:03:54.480
<v Ian Gibbins>I'd like to thank you all for your time.

71
00:03:54.480 --> 00:03:56.360
<v Ian Gibbins>I'd like to welcome Srini to come and give us a talk about

72
00:03:56.360 --> 00:03:58.240
<v Ian Gibbins>the milligram brain of a bee.

73
00:04:03.420 --> 00:04:07.180
<v Mandyam Srinivasan>Thank you, Ian, for that very kind introduction.

74
00:04:08.300 --> 00:04:11.420
<v Mandyam Srinivasan>It is a tremendous pleasure and a tremendous honour

75
00:04:11.420 --> 00:04:13.200
<v Mandyam Srinivasan>to be speaking in front of you folks today.

76
00:04:13.300 --> 00:04:14.700
<v Mandyam Srinivasan>Thank you so much for coming.

77
00:04:16.780 --> 00:04:18.420
<v Mandyam Srinivasan>So that's the title of today's talk,

78
00:04:18.540 --> 00:04:20.200
<v Mandyam Srinivasan>Pushing the Limits of a One Milligram Brain.

79
00:04:20.360 --> 00:04:22.420
<v Mandyam Srinivasan>We're going to be looking at vision and perception

80
00:04:22.420 --> 00:04:23.980
<v Mandyam Srinivasan>and cognition of honeybees.

81
00:04:24.480 --> 00:04:27.420
<v Mandyam Srinivasan>And possibly some applications to new ways

82
00:04:27.420 --> 00:04:29.640
<v Mandyam Srinivasan>of maybe designing flying vehicles,

83
00:04:30.360 --> 00:04:32.160
<v Mandyam Srinivasan>vision systems for flying vehicles anyway.

84
00:04:32.800 --> 00:04:34.780
<v Mandyam Srinivasan>So, well, you know, any one of you who is

85
00:04:37.340 --> 00:04:40.400
<v Mandyam Srinivasan>maybe tried to, you know, swat a fly or watch a fly

86
00:04:40.400 --> 00:04:43.320
<v Mandyam Srinivasan>do a lovely, beautiful landing on the rim of your teacup

87
00:04:43.320 --> 00:04:45.440
<v Mandyam Srinivasan>or marvel at a honeybee heading out from a hive,

88
00:04:45.880 --> 00:04:48.280
<v Mandyam Srinivasan>travelling several kilometres in search of food.

89
00:04:48.400 --> 00:04:50.020
<v Mandyam Srinivasan>It's a meandering route that it does

90
00:04:50.020 --> 00:04:51.500
<v Mandyam Srinivasan>because it doesn't know where to find the food.

91
00:04:51.680 --> 00:04:54.400
<v Mandyam Srinivasan>But when it finds the food, which might be as much as,

92
00:04:54.480 --> 00:04:56.480
<v Mandyam Srinivasan>as many as six or seven kilometres away,

93
00:04:57.020 --> 00:04:58.860
<v Mandyam Srinivasan>it'll make a beeline back home, literally.

94
00:04:59.320 --> 00:05:01.080
<v Mandyam Srinivasan>And the question is, how do they do this?

95
00:05:01.140 --> 00:05:01.760
<v Mandyam Srinivasan>How do they do this?

96
00:05:01.780 --> 00:05:04.640
<v Mandyam Srinivasan>A brain that weighs about the size of a milligram

97
00:05:05.240 --> 00:05:07.660
<v Mandyam Srinivasan>carries far fewer neurons than our own brains.

98
00:05:08.080 --> 00:05:09.060
<v Mandyam Srinivasan>By the way, can you hear me all right?

99
00:05:09.080 --> 00:05:10.120
<v Mandyam Srinivasan>Can you hear me clearly? Loud enough?

100
00:05:11.140 --> 00:05:13.500
<v Mandyam Srinivasan>And what is it that makes them tick and tick so well?

101
00:05:13.600 --> 00:05:16.340
<v Mandyam Srinivasan>So as curious scientists, as curious zoologists,

102
00:05:16.900 --> 00:05:19.160
<v Mandyam Srinivasan>one of the missions in our lab is to try and work out

103
00:05:19.160 --> 00:05:22.080
<v Mandyam Srinivasan>exactly what it is that makes these animals tick and tick so well.

104
00:05:22.760 --> 00:05:24.000
<v Mandyam Srinivasan>The other driving force,

105
00:05:24.480 --> 00:05:26.520
<v Mandyam Srinivasan>because as Ian mentioned, I was trained as an engineer initially,

106
00:05:26.760 --> 00:05:29.780
<v Mandyam Srinivasan>is if we do find something neat and clever that these creatures do,

107
00:05:29.960 --> 00:05:33.900
<v Mandyam Srinivasan>can we put them into, can we devise novel kinds of algorithms

108
00:05:33.900 --> 00:05:37.620
<v Mandyam Srinivasan>for vision that we can put into flying vehicles

109
00:05:37.620 --> 00:05:40.300
<v Mandyam Srinivasan>that will one day hopefully fly autonomously

110
00:05:40.300 --> 00:05:43.540
<v Mandyam Srinivasan>and do things which are slightly dangerous or slightly risky,

111
00:05:43.740 --> 00:05:46.520
<v Mandyam Srinivasan>where you prefer not to put a pilot in there in that scenario

112
00:05:46.520 --> 00:05:48.620
<v Mandyam Srinivasan>and, you know, put him or her into trouble.

113
00:05:48.880 --> 00:05:51.260
<v Mandyam Srinivasan>That's one aspect of the work that I also hope to talk about.

114
00:05:51.480 --> 00:05:54.420
<v Mandyam Srinivasan>The third aspect, which I will come back to in the end,

115
00:05:54.480 --> 00:05:57.380
<v Mandyam Srinivasan>is, again, goes back to the brain of the bee and says,

116
00:05:57.560 --> 00:06:00.660
<v Mandyam Srinivasan>how far can you push the perceptual capacities of a bee?

117
00:06:00.820 --> 00:06:03.420
<v Mandyam Srinivasan>I mean, of a creature with a small brain.

118
00:06:03.900 --> 00:06:06.740
<v Mandyam Srinivasan>There's no doubt that insects will learn and bees will learn too,

119
00:06:06.900 --> 00:06:09.040
<v Mandyam Srinivasan>but when they learn things, will they learn them in a fairly,

120
00:06:09.120 --> 00:06:11.560
<v Mandyam Srinivasan>in a rote automaton-like way,

121
00:06:11.720 --> 00:06:14.820
<v Mandyam Srinivasan>or can they be flexible about the ways they think they learn things?

122
00:06:14.920 --> 00:06:17.820
<v Mandyam Srinivasan>Can they learn, for example, general rules from certain tasks

123
00:06:17.820 --> 00:06:20.500
<v Mandyam Srinivasan>that they've been asked to learn and apply them to novel situations?

124
00:06:21.020 --> 00:06:23.540
<v Mandyam Srinivasan>In other words, can you teach an old bee new tricks?

125
00:06:23.760 --> 00:06:24.460
<v Mandyam Srinivasan>That's the kind of thing.

126
00:06:24.480 --> 00:06:26.120
<v Mandyam Srinivasan>We'd like to look at in the third part of the talk.

127
00:06:26.440 --> 00:06:28.380
<v Mandyam Srinivasan>So that's the general sort of outline of the talk.

128
00:06:28.840 --> 00:06:31.560
<v Mandyam Srinivasan>Now, when you take a close look at an insect eye,

129
00:06:32.600 --> 00:06:35.540
<v Mandyam Srinivasan>and as many of you know, insects have compound eyes,

130
00:06:36.700 --> 00:06:39.140
<v Mandyam Srinivasan>as opposed to our eyes, which are called simple eyes.

131
00:06:39.460 --> 00:06:41.300
<v Mandyam Srinivasan>And the thing where a simple eye, of course,

132
00:06:41.380 --> 00:06:43.560
<v Mandyam Srinivasan>has consists of a single lens that focuses light

133
00:06:43.560 --> 00:06:45.340
<v Mandyam Srinivasan>onto a large number of photoreceptors,

134
00:06:45.380 --> 00:06:48.060
<v Mandyam Srinivasan>typically 100 million in our case, in the retina.

135
00:06:48.380 --> 00:06:50.320
<v Mandyam Srinivasan>And so it's more or less like a camera eye.

136
00:06:51.000 --> 00:06:54.140
<v Mandyam Srinivasan>Whereas the compound eye consists of many, many little eyes.

137
00:06:54.480 --> 00:06:56.120
<v Mandyam Srinivasan>And each little eye is called an ommatidium.

138
00:06:56.320 --> 00:06:58.440
<v Mandyam Srinivasan>And this ommatidium consists of a small lens

139
00:06:58.440 --> 00:07:01.620
<v Mandyam Srinivasan>that focuses light onto a small group of light-sensitive cells,

140
00:07:01.740 --> 00:07:04.260
<v Mandyam Srinivasan>typically between six and eight in number, photoreceptors.

141
00:07:04.440 --> 00:07:07.360
<v Mandyam Srinivasan>And each ommatidium looks in a slightly different direction

142
00:07:07.360 --> 00:07:07.760
<v Mandyam Srinivasan>in space.

143
00:07:08.080 --> 00:07:11.540
<v Mandyam Srinivasan>So in the case of a honeybee, you've got about 4,500 ommatidia

144
00:07:11.540 --> 00:07:13.760
<v Mandyam Srinivasan>in one eye and a similar number in the other eye.

145
00:07:13.960 --> 00:07:17.160
<v Mandyam Srinivasan>And together, this collection of 9,000 ommatidia

146
00:07:17.640 --> 00:07:19.220
<v Mandyam Srinivasan>is picking up a representation.

147
00:07:19.500 --> 00:07:20.560
<v Mandyam Srinivasan>It's sampling space.

148
00:07:21.000 --> 00:07:24.300
<v Mandyam Srinivasan>And what is nice is that within each ommatidium, you've got three

149
00:07:24.480 --> 00:07:27.320
<v Mandyam Srinivasan>different kinds of color receptors for light-sensitive receptors.

150
00:07:27.600 --> 00:07:29.660
<v Mandyam Srinivasan>Just like we have trichromatic color vision,

151
00:07:29.840 --> 00:07:32.300
<v Mandyam Srinivasan>turns out honeybees also have trichromatic color vision.

152
00:07:32.380 --> 00:07:33.500
<v Mandyam Srinivasan>So they have three primary colors.

153
00:07:33.780 --> 00:07:36.200
<v Mandyam Srinivasan>And so each ommatidium is picking up a kind

154
00:07:36.200 --> 00:07:39.900
<v Mandyam Srinivasan>of a pointillistic representation, if you will, of the world.

155
00:07:40.000 --> 00:07:42.760
<v Mandyam Srinivasan>It's a panoramic, pointillistic representation of the world

156
00:07:42.760 --> 00:07:44.760
<v Mandyam Srinivasan>that the compound eye is acquiring.

157
00:07:45.540 --> 00:07:48.620
<v Mandyam Srinivasan>So obviously, the optics of the compound eye

158
00:07:49.220 --> 00:07:51.860
<v Mandyam Srinivasan>are a little bit different from the optics of our own eyes.

159
00:07:52.120 --> 00:07:54.460
<v Mandyam Srinivasan>But that by itself is not a great way

160
00:07:54.480 --> 00:07:55.240
<v Mandyam Srinivasan>to see the world.

161
00:07:55.240 --> 00:07:56.460
<v Mandyam Srinivasan>It's not of great consequence.

162
00:07:56.880 --> 00:07:59.360
<v Mandyam Srinivasan>A much more fundamental difference between insect vision

163
00:07:59.980 --> 00:08:02.020
<v Mandyam Srinivasan>and our own vision comes about from the fact

164
00:08:02.020 --> 00:08:04.620
<v Mandyam Srinivasan>that the two eyes of an insect are actually very close together.

165
00:08:05.500 --> 00:08:07.100
<v Mandyam Srinivasan>Our eyes are fairly far apart.

166
00:08:07.300 --> 00:08:08.420
<v Mandyam Srinivasan>They're about 8 centimeters apart.

167
00:08:08.660 --> 00:08:11.540
<v Mandyam Srinivasan>And so we can do stereo vision quite easily.

168
00:08:11.700 --> 00:08:14.060
<v Mandyam Srinivasan>The reason why stereo vision works well in our case

169
00:08:14.060 --> 00:08:16.020
<v Mandyam Srinivasan>is that because it has a fairly far apart,

170
00:08:16.220 --> 00:08:17.360
<v Mandyam Srinivasan>if I look at my finger with one eye

171
00:08:17.360 --> 00:08:19.440
<v Mandyam Srinivasan>and then with the other eye, go on, you can try it yourself,

172
00:08:19.800 --> 00:08:21.540
<v Mandyam Srinivasan>you'll find that the image of that finger

173
00:08:21.540 --> 00:08:24.260
<v Mandyam Srinivasan>is displaced between the two images.

174
00:08:24.480 --> 00:08:25.060
<v Mandyam Srinivasan>It's the retinae, right?

175
00:08:25.200 --> 00:08:27.600
<v Mandyam Srinivasan>And what your brain is doing is it's actually measuring

176
00:08:27.600 --> 00:08:29.960
<v Mandyam Srinivasan>that displacement, or disparity as it's called,

177
00:08:30.080 --> 00:08:32.640
<v Mandyam Srinivasan>and by simple triangulation, it's working out

178
00:08:32.640 --> 00:08:33.900
<v Mandyam Srinivasan>how far away that finger is.

179
00:08:34.200 --> 00:08:34.640
<v Mandyam Srinivasan>Right?

180
00:08:34.680 --> 00:08:35.140
<v Mandyam Srinivasan>Does that make sense?

181
00:08:35.460 --> 00:08:38.160
<v Mandyam Srinivasan>Now, if you take these two eyes and move them progressively

182
00:08:38.160 --> 00:08:40.720
<v Mandyam Srinivasan>closer and closer together, that disparity

183
00:08:40.720 --> 00:08:41.860
<v Mandyam Srinivasan>gets smaller and smaller.

184
00:08:42.300 --> 00:08:44.120
<v Mandyam Srinivasan>Ultimately, when the two eyes are very close together,

185
00:08:44.420 --> 00:08:47.120
<v Mandyam Srinivasan>that disparity is close to zero, because you virtually

186
00:08:47.120 --> 00:08:49.120
<v Mandyam Srinivasan>have the same viewpoint from both eyes.

187
00:08:49.220 --> 00:08:50.640
<v Mandyam Srinivasan>The eyes haven't shifted very much at all.

188
00:08:50.880 --> 00:08:54.220
<v Mandyam Srinivasan>So the stereo cues become very weak, very hard to measure,

189
00:08:54.480 --> 00:08:55.860
<v Mandyam Srinivasan>if your eyes are very close together.

190
00:08:56.160 --> 00:09:00.400
<v Mandyam Srinivasan>It's almost virtually like being a cyclops, rather than a two-eyed

191
00:09:00.400 --> 00:09:00.660
<v Mandyam Srinivasan>creature.

192
00:09:01.080 --> 00:09:02.780
<v Mandyam Srinivasan>So what do insects do?

193
00:09:02.900 --> 00:09:04.500
<v Mandyam Srinivasan>It turns out insects have overcome this problem

194
00:09:04.500 --> 00:09:05.920
<v Mandyam Srinivasan>by using an entirely different way

195
00:09:06.420 --> 00:09:07.500
<v Mandyam Srinivasan>to see the world in three dimensions.

196
00:09:07.720 --> 00:09:11.280
<v Mandyam Srinivasan>And we use that too, but insects use it a lot more.

197
00:09:11.520 --> 00:09:14.420
<v Mandyam Srinivasan>And the idea is that you don't try to use stereo at all.

198
00:09:14.480 --> 00:09:16.760
<v Mandyam Srinivasan>You simply do a much more active way

199
00:09:16.760 --> 00:09:18.080
<v Mandyam Srinivasan>of sensing the world in 3D.

200
00:09:18.140 --> 00:09:19.980
<v Mandyam Srinivasan>And this is by actually moving in the world

201
00:09:19.980 --> 00:09:22.700
<v Mandyam Srinivasan>and seeing how the image of the world changes in response

202
00:09:22.700 --> 00:09:23.580
<v Mandyam Srinivasan>to your motion.

203
00:09:23.580 --> 00:09:25.120
<v Mandyam Srinivasan>So the idea is very simple.

204
00:09:25.260 --> 00:09:26.740
<v Mandyam Srinivasan>Let's say you're flying along in a straight line

205
00:09:26.740 --> 00:09:28.220
<v Mandyam Srinivasan>or driving along a highway.

206
00:09:28.480 --> 00:09:30.440
<v Mandyam Srinivasan>You notice that the trees lining the highway

207
00:09:31.000 --> 00:09:32.260
<v Mandyam Srinivasan>are whizzing by you very rapidly.

208
00:09:32.600 --> 00:09:34.620
<v Mandyam Srinivasan>And this rapidity of image motion

209
00:09:34.620 --> 00:09:36.620
<v Mandyam Srinivasan>tells you that these objects are very close to you.

210
00:09:37.020 --> 00:09:39.480
<v Mandyam Srinivasan>On the other hand, distant objects, like a distant hill,

211
00:09:39.620 --> 00:09:41.920
<v Mandyam Srinivasan>or the clouds in the sky, or the sun,

212
00:09:42.180 --> 00:09:44.140
<v Mandyam Srinivasan>those objects, the images of those objects,

213
00:09:44.220 --> 00:09:46.000
<v Mandyam Srinivasan>don't move much at all, as long as you're

214
00:09:46.000 --> 00:09:46.740
<v Mandyam Srinivasan>moving in a straight line.

215
00:09:47.000 --> 00:09:48.720
<v Mandyam Srinivasan>And that slowness of image motion

216
00:09:48.720 --> 00:09:50.500
<v Mandyam Srinivasan>tells you that those objects are very far away.

217
00:09:51.040 --> 00:09:52.600
<v Mandyam Srinivasan>So insects have learned to translate,

218
00:09:52.600 --> 00:09:58.520
<v Mandyam Srinivasan>or interpret, the speed of image motion as object distance.

219
00:09:58.700 --> 00:10:01.520
<v Mandyam Srinivasan>And that's how they disentangle the world into a 3D world.

220
00:10:01.720 --> 00:10:04.280
<v Mandyam Srinivasan>So they have to move in order to see the world in 3D.

221
00:10:04.380 --> 00:10:06.680
<v Mandyam Srinivasan>If they stop, the whole world collapses

222
00:10:06.680 --> 00:10:08.700
<v Mandyam Srinivasan>to a two-dimensional world.

223
00:10:08.800 --> 00:10:10.400
<v Mandyam Srinivasan>So they have to move in order to literally see.

224
00:10:10.800 --> 00:10:12.780
<v Mandyam Srinivasan>That's the way insect vision seems to work.

225
00:10:13.100 --> 00:10:16.760
<v Mandyam Srinivasan>Now what I'd like to do is spend the next few slides describing

226
00:10:16.760 --> 00:10:18.080
<v Mandyam Srinivasan>to you some of the experiments we've

227
00:10:18.080 --> 00:10:20.420
<v Mandyam Srinivasan>done in our lab to gain these particular insights about how

228
00:10:21.080 --> 00:10:22.240
<v Mandyam Srinivasan>insects see the world in three dimensions.

229
00:10:22.600 --> 00:10:25.140
<v Mandyam Srinivasan>And avoid colliding with objects and so on as they fly through it.

230
00:10:27.000 --> 00:10:31.160
<v Mandyam Srinivasan>Before I do that, has any of the audience actually trained bees?

231
00:10:32.600 --> 00:10:34.580
<v Mandyam Srinivasan>This is what I do for a living most of the time.

232
00:10:34.680 --> 00:10:35.720
<v Mandyam Srinivasan>I train honeybees.

233
00:10:36.240 --> 00:10:37.060
<v Mandyam Srinivasan>It's actually lovely.

234
00:10:37.180 --> 00:10:38.180
<v Mandyam Srinivasan>You can do it in your backyard.

235
00:10:38.320 --> 00:10:40.280
<v Mandyam Srinivasan>It doesn't require any expense.

236
00:10:40.880 --> 00:10:43.100
<v Mandyam Srinivasan>So by the way, this is Dr. Miriam Lehrer,

237
00:10:43.900 --> 00:10:45.280
<v Mandyam Srinivasan>who's taught me everything about bees.

238
00:10:45.420 --> 00:10:47.740
<v Mandyam Srinivasan>She is probably the world's expert in training bees.

239
00:10:47.800 --> 00:10:48.900
<v Mandyam Srinivasan>She's sadly passed away now.

240
00:10:49.120 --> 00:10:52.220
<v Mandyam Srinivasan>But when I joined the University of Zurich, she really was

241
00:10:52.600 --> 00:10:52.700
<v Mandyam Srinivasan>a very good person.

242
00:10:53.300 --> 00:10:55.580
<v Mandyam Srinivasan>She took me by the hand and really showed me how to do this.

243
00:10:55.880 --> 00:10:58.280
<v Mandyam Srinivasan>Here she is setting up a sugar water feeder

244
00:10:58.280 --> 00:11:02.040
<v Mandyam Srinivasan>outside the lab in the balcony in Zurich.

245
00:11:02.300 --> 00:11:06.380
<v Mandyam Srinivasan>And the hives are about 300 or 400 meters away.

246
00:11:06.680 --> 00:11:10.500
<v Mandyam Srinivasan>And if you set this up here on a summer morning,

247
00:11:10.640 --> 00:11:12.180
<v Mandyam Srinivasan>pretty soon bees will notice it.

248
00:11:12.260 --> 00:11:14.800
<v Mandyam Srinivasan>They'll come and land and taste the good sugar water.

249
00:11:14.960 --> 00:11:17.240
<v Mandyam Srinivasan>And then they'll go back home and report it

250
00:11:17.240 --> 00:11:18.720
<v Mandyam Srinivasan>to their dance, which we'll talk about later.

251
00:11:18.840 --> 00:11:22.300
<v Mandyam Srinivasan>Come back and then they collect more and more recruits.

252
00:11:22.300 --> 00:11:23.380
<v Mandyam Srinivasan>Because they're signaling through the dance

253
00:11:23.380 --> 00:11:24.420
<v Mandyam Srinivasan>that there's some good stuff there.

254
00:11:25.620 --> 00:11:27.480
<v Mandyam Srinivasan>So what you do is then, once you've

255
00:11:27.480 --> 00:11:29.940
<v Mandyam Srinivasan>got a few bees coming out here, you suddenly

256
00:11:29.940 --> 00:11:32.340
<v Mandyam Srinivasan>increase the concentration of the sugar solution there.

257
00:11:32.460 --> 00:11:35.340
<v Mandyam Srinivasan>So the food tastes a lot better, much more sweet.

258
00:11:35.380 --> 00:11:36.680
<v Mandyam Srinivasan>That gets the bees a lot more excited.

259
00:11:37.080 --> 00:11:40.420
<v Mandyam Srinivasan>They go back home and dance more vigorously, recruit more bees.

260
00:11:40.640 --> 00:11:42.480
<v Mandyam Srinivasan>And pretty soon, you've got quite a number of bees

261
00:11:42.480 --> 00:11:43.440
<v Mandyam Srinivasan>sort of accumulating here.

262
00:11:43.660 --> 00:11:46.720
<v Mandyam Srinivasan>And then you start to move this feeder step by step

263
00:11:47.280 --> 00:11:49.020
<v Mandyam Srinivasan>into the lab where you want to do your experiments.

264
00:11:49.140 --> 00:11:50.080
<v Mandyam Srinivasan>And of course, the bees will follow.

265
00:11:50.440 --> 00:11:51.320
<v Mandyam Srinivasan>There's no problem there.

266
00:11:52.300 --> 00:11:53.940
<v Mandyam Srinivasan>But you've got to take account of the fact

267
00:11:53.940 --> 00:11:55.580
<v Mandyam Srinivasan>that this number is building up exponentially.

268
00:11:56.120 --> 00:11:58.260
<v Mandyam Srinivasan>And you don't want the whole hive to be visiting your lab.

269
00:11:58.520 --> 00:12:00.920
<v Mandyam Srinivasan>Because typically, you've got about 40,000 bees in a hive.

270
00:12:01.180 --> 00:12:02.100
<v Mandyam Srinivasan>And you don't want that.

271
00:12:02.340 --> 00:12:05.380
<v Mandyam Srinivasan>So what you do is, once you've got the feeder well and truly

272
00:12:05.380 --> 00:12:08.340
<v Mandyam Srinivasan>inside your lab and about to enter your apparatus where

273
00:12:08.340 --> 00:12:10.400
<v Mandyam Srinivasan>you want to do your experiment, you

274
00:12:10.400 --> 00:12:13.100
<v Mandyam Srinivasan>place, in addition, another feeder here

275
00:12:13.560 --> 00:12:14.880
<v Mandyam Srinivasan>with weaker sugar solution.

276
00:12:15.460 --> 00:12:17.100
<v Mandyam Srinivasan>And that acts like a bit of a decoy.

277
00:12:17.440 --> 00:12:19.620
<v Mandyam Srinivasan>So what happens is that these new recruits, which

278
00:12:19.620 --> 00:12:21.840
<v Mandyam Srinivasan>have been alerted by the dance, come in here

279
00:12:21.840 --> 00:12:23.740
<v Mandyam Srinivasan>and land at the most obvious target, which

280
00:12:23.740 --> 00:12:25.940
<v Mandyam Srinivasan>is this outside sugar solution now, which is much weaker

281
00:12:25.940 --> 00:12:26.740
<v Mandyam Srinivasan>than the inside one.

282
00:12:27.080 --> 00:12:28.900
<v Mandyam Srinivasan>And they say, ah, this is a false alarm.

283
00:12:29.040 --> 00:12:30.600
<v Mandyam Srinivasan>This is not as good as it was cracked up to be.

284
00:12:30.980 --> 00:12:33.560
<v Mandyam Srinivasan>So they get disappointed and don't come back again.

285
00:12:33.820 --> 00:12:35.560
<v Mandyam Srinivasan>So it's only the bees that know the secret of where

286
00:12:35.560 --> 00:12:39.100
<v Mandyam Srinivasan>the good stuff is to the original bees that know where to go.

287
00:12:39.360 --> 00:12:40.480
<v Mandyam Srinivasan>And they keep coming inside.

288
00:12:40.940 --> 00:12:42.400
<v Mandyam Srinivasan>And that's when you start your experiments.

289
00:12:42.500 --> 00:12:44.000
<v Mandyam Srinivasan>So you've got control over exactly how many bees

290
00:12:44.000 --> 00:12:45.880
<v Mandyam Srinivasan>you've got, or at least approximately how many bees

291
00:12:45.880 --> 00:12:46.240
<v Mandyam Srinivasan>you've got.

292
00:12:46.480 --> 00:12:49.460
<v Mandyam Srinivasan>It's a manageable number, typically between 6 and 10.

293
00:12:49.900 --> 00:12:51.220
<v Mandyam Srinivasan>And then you can start your experiment.

294
00:12:51.220 --> 00:12:52.440
<v Mandyam Srinivasan>And they'll keep coming again and again.

295
00:12:52.500 --> 00:12:54.520
<v Mandyam Srinivasan>As long as there's no competition from outside food

296
00:12:54.520 --> 00:12:56.600
<v Mandyam Srinivasan>sources, which are much better, they

297
00:12:56.600 --> 00:12:58.640
<v Mandyam Srinivasan>will keep coming to your food source.

298
00:12:59.000 --> 00:13:01.320
<v Mandyam Srinivasan>And by the way, I should mention that these bees are coming to you

299
00:13:01.320 --> 00:13:02.320
<v Mandyam Srinivasan>of their own free will.

300
00:13:02.900 --> 00:13:04.820
<v Mandyam Srinivasan>They're not being forced or coerced in any way.

301
00:13:04.880 --> 00:13:06.840
<v Mandyam Srinivasan>It's just the food that you're offering them, right?

302
00:13:07.180 --> 00:13:08.120
<v Mandyam Srinivasan>That's the only reason.

303
00:13:09.060 --> 00:13:10.460
<v Mandyam Srinivasan>OK, so that's so you can be.

304
00:13:10.640 --> 00:13:12.900
<v Mandyam Srinivasan>By the way, you can then color code these bees individually

305
00:13:13.440 --> 00:13:14.260
<v Mandyam Srinivasan>to keep track of individuals.

306
00:13:14.540 --> 00:13:15.920
<v Mandyam Srinivasan>And by the way, they do have individual personalities.

307
00:13:16.400 --> 00:13:18.480
<v Mandyam Srinivasan>Each bee is really very special.

308
00:13:18.560 --> 00:13:19.600
<v Mandyam Srinivasan>I mean, you can tell by its behavior

309
00:13:19.600 --> 00:13:21.060
<v Mandyam Srinivasan>after you know a particular bee.

310
00:13:21.520 --> 00:13:22.680
<v Mandyam Srinivasan>You can see exactly what it's going to do.

311
00:13:22.860 --> 00:13:23.920
<v Mandyam Srinivasan>They really are individual.

312
00:13:24.120 --> 00:13:24.580
<v Mandyam Srinivasan>I believe that.

313
00:13:24.660 --> 00:13:27.060
<v Mandyam Srinivasan>They're not just creatures of one big collection.

314
00:13:27.700 --> 00:13:29.120
<v Mandyam Srinivasan>So you can color code them individually

315
00:13:29.120 --> 00:13:30.420
<v Mandyam Srinivasan>and follow their individual progress.

316
00:13:30.560 --> 00:13:32.200
<v Mandyam Srinivasan>If you're doing a training experiment, a learning

317
00:13:32.200 --> 00:13:34.920
<v Mandyam Srinivasan>experiment, you can keep track of how many times you've

318
00:13:34.920 --> 00:13:37.560
<v Mandyam Srinivasan>rewarded them and see how their learning curve develops

319
00:13:37.560 --> 00:13:38.640
<v Mandyam Srinivasan>and things like that.

320
00:13:38.700 --> 00:13:40.060
<v Mandyam Srinivasan>You can compare individuals and so on.

321
00:13:40.280 --> 00:13:43.260
<v Mandyam Srinivasan>By the way, just before I forget to mention,

322
00:13:43.420 --> 00:13:45.500
<v Mandyam Srinivasan>all of the bees that come to us are, of course, females.

323
00:13:45.640 --> 00:13:47.080
<v Mandyam Srinivasan>Because it's only the female worker bees

324
00:13:47.080 --> 00:13:48.020
<v Mandyam Srinivasan>that do all the foraging.

325
00:13:48.920 --> 00:13:50.760
<v Mandyam Srinivasan>So all of our work is done typically

326
00:13:50.760 --> 00:13:51.480
<v Mandyam Srinivasan>with females.

327
00:13:51.700 --> 00:13:53.160
<v Mandyam Srinivasan>And so many females do all the hard work.

328
00:13:53.200 --> 00:13:54.280
<v Mandyam Srinivasan>The males don't do any work.

329
00:13:54.400 --> 00:13:55.680
<v Mandyam Srinivasan>They've got only one thing on their mind

330
00:13:55.680 --> 00:13:58.140
<v Mandyam Srinivasan>that certainly isn't foraging or hard work.

331
00:14:00.780 --> 00:14:03.720
<v Mandyam Srinivasan>So many, many years ago, we noticed

332
00:14:03.720 --> 00:14:05.840
<v Mandyam Srinivasan>that when we trained bees, sometimes we

333
00:14:05.840 --> 00:14:08.620
<v Mandyam Srinivasan>trained them to come through a hole in the window like this,

334
00:14:08.680 --> 00:14:09.060
<v Mandyam Srinivasan>for example.

335
00:14:10.360 --> 00:14:12.880
<v Mandyam Srinivasan>We found that when bees flew through an aperture like this,

336
00:14:12.880 --> 00:14:14.740
<v Mandyam Srinivasan>a hole, they flew fairly precisely

337
00:14:14.740 --> 00:14:16.100
<v Mandyam Srinivasan>through the middle of the hole.

338
00:14:16.360 --> 00:14:17.740
<v Mandyam Srinivasan>And the question we asked ourselves

339
00:14:17.740 --> 00:14:20.240
<v Mandyam Srinivasan>is, how do they balance the distances to the left

340
00:14:20.240 --> 00:14:23.300
<v Mandyam Srinivasan>and right-hand rims so well when they fly down the middle,

341
00:14:23.400 --> 00:14:25.080
<v Mandyam Srinivasan>given the fact that they don't have stereo

342
00:14:25.080 --> 00:14:27.480
<v Mandyam Srinivasan>to measure the distances of these edges?

343
00:14:28.040 --> 00:14:29.920
<v Mandyam Srinivasan>So even if they had stereo, it really

344
00:14:29.920 --> 00:14:30.900
<v Mandyam Srinivasan>wouldn't work in this case.

345
00:14:30.960 --> 00:14:33.180
<v Mandyam Srinivasan>Because you see, the right-hand rim of this hole

346
00:14:33.180 --> 00:14:34.860
<v Mandyam Srinivasan>is being viewed only by the right eye.

347
00:14:35.040 --> 00:14:37.200
<v Mandyam Srinivasan>And the left-hand rim is being viewed only by the left eye.

348
00:14:37.320 --> 00:14:38.840
<v Mandyam Srinivasan>So even if you had stereo, you couldn't really

349
00:14:38.840 --> 00:14:41.240
<v Mandyam Srinivasan>use it to solve this particular problem.

350
00:14:41.440 --> 00:14:42.760
<v Mandyam Srinivasan>So how do they do this?

351
00:14:43.260 --> 00:14:45.740
<v Mandyam Srinivasan>Well, we wanted to be doing something far simpler,

352
00:14:45.940 --> 00:14:48.840
<v Mandyam Srinivasan>which is to balance the speeds of motion of the two rims

353
00:14:48.840 --> 00:14:49.780
<v Mandyam Srinivasan>over the eye.

354
00:14:50.240 --> 00:14:51.540
<v Mandyam Srinivasan>The images of these two rims over the eye

355
00:14:51.540 --> 00:14:52.940
<v Mandyam Srinivasan>as they fly through the aperture.

356
00:14:53.220 --> 00:14:55.580
<v Mandyam Srinivasan>If one rim appears to be moving faster than the other,

357
00:14:55.680 --> 00:14:57.580
<v Mandyam Srinivasan>you know you're too close to that rim, right?

358
00:14:57.680 --> 00:15:00.760
<v Mandyam Srinivasan>So you can position yourself to compensate for that.

359
00:15:01.080 --> 00:15:01.880
<v Mandyam Srinivasan>Well, do they do that?

360
00:15:02.260 --> 00:15:03.600
<v Mandyam Srinivasan>They can check that rather easily

361
00:15:03.600 --> 00:15:05.740
<v Mandyam Srinivasan>by training bees to come into a tunnel like this

362
00:15:05.740 --> 00:15:06.460
<v Mandyam Srinivasan>inside your lab.

363
00:15:06.700 --> 00:15:10.280
<v Mandyam Srinivasan>This is a corridor-shaped tunnel with a sugar water feeder

364
00:15:10.280 --> 00:15:12.140
<v Mandyam Srinivasan>at the far end, which I haven't actually shown you here.

365
00:15:12.400 --> 00:15:14.540
<v Mandyam Srinivasan>And you line the inside of the tunnel

366
00:15:14.540 --> 00:15:16.900
<v Mandyam Srinivasan>with the visual texture, for example,

367
00:15:16.900 --> 00:15:19.160
<v Mandyam Srinivasan>vertical black and white stripes or some kind of visual texture.

368
00:15:20.240 --> 00:15:22.100
<v Mandyam Srinivasan>And then you place a video camera on top

369
00:15:22.100 --> 00:15:24.400
<v Mandyam Srinivasan>and film the trajectories of these bees

370
00:15:24.980 --> 00:15:26.520
<v Mandyam Srinivasan>as they're flying to the feeder.

371
00:15:26.900 --> 00:15:29.300
<v Mandyam Srinivasan>And lo and behold, what you find is

372
00:15:29.300 --> 00:15:32.280
<v Mandyam Srinivasan>the bees indeed fly down the middle of this tunnel,

373
00:15:32.320 --> 00:15:33.100
<v Mandyam Srinivasan>pretty much in the middle.

374
00:15:33.180 --> 00:15:35.720
<v Mandyam Srinivasan>These green bars denote the mean and standard deviation

375
00:15:35.720 --> 00:15:38.600
<v Mandyam Srinivasan>of several hundred flights analyzed in this way.

376
00:15:38.780 --> 00:15:41.020
<v Mandyam Srinivasan>There's a bit of slop, as you can see here.

377
00:15:41.080 --> 00:15:43.500
<v Mandyam Srinivasan>They're not precisely dead in the middle.

378
00:15:43.560 --> 00:15:44.900
<v Mandyam Srinivasan>But hey, bees are only human.

379
00:15:45.020 --> 00:15:45.980
<v Mandyam Srinivasan>This is pretty good for a bee.

380
00:15:46.520 --> 00:15:47.240
<v Mandyam Srinivasan>It's not bad.

381
00:15:47.880 --> 00:15:49.860
<v Mandyam Srinivasan>However, what's interesting now is that if you take one

382
00:15:49.860 --> 00:15:53.620
<v Mandyam Srinivasan>of these walls and move it in the same direction as the bee,

383
00:15:53.680 --> 00:15:55.280
<v Mandyam Srinivasan>and you can do this because the walls are actually

384
00:15:56.120 --> 00:15:59.900
<v Mandyam Srinivasan>belts that are mounted on a spinning two-wheeled conveyor

385
00:15:59.900 --> 00:16:00.060
<v Mandyam Srinivasan>belt.

386
00:16:00.440 --> 00:16:03.420
<v Mandyam Srinivasan>And of course, you've got Perspex walls in the middle

387
00:16:03.420 --> 00:16:05.080
<v Mandyam Srinivasan>to make sure that there's no air currents that

388
00:16:05.080 --> 00:16:06.100
<v Mandyam Srinivasan>are affecting the flight of the bees.

389
00:16:06.200 --> 00:16:07.680
<v Mandyam Srinivasan>It's only the visual stimulus that affects them.

390
00:16:07.900 --> 00:16:09.320
<v Mandyam Srinivasan>So if you take one of these walls now

391
00:16:09.320 --> 00:16:11.720
<v Mandyam Srinivasan>and move it in the same direction as the flying bee,

392
00:16:12.080 --> 00:16:15.060
<v Mandyam Srinivasan>you find the bee flies a lot closer to the moving wall.

393
00:16:15.820 --> 00:16:17.920
<v Mandyam Srinivasan>So what's happening here is because the bee and the wall

394
00:16:17.920 --> 00:16:19.240
<v Mandyam Srinivasan>are moving in the same direction.

395
00:16:19.860 --> 00:16:23.100
<v Mandyam Srinivasan>The image velocity as seen by that left eye is lower.

396
00:16:23.180 --> 00:16:23.620
<v Mandyam Srinivasan>It's reduced.

397
00:16:24.040 --> 00:16:26.040
<v Mandyam Srinivasan>So the bee thinks that wall is much further away.

398
00:16:26.760 --> 00:16:29.060
<v Mandyam Srinivasan>So she moves closer to that wall to compensate.

399
00:16:30.080 --> 00:16:30.960
<v Mandyam Srinivasan>Does that make sense?

400
00:16:31.800 --> 00:16:33.360
<v Mandyam Srinivasan>Now, the opposite thing happens if you

401
00:16:33.360 --> 00:16:36.020
<v Mandyam Srinivasan>move this wall in the opposite direction to the bee's flight.

402
00:16:36.100 --> 00:16:39.000
<v Mandyam Srinivasan>Now the bee flies a lot further away from that moving wall

403
00:16:39.000 --> 00:16:40.460
<v Mandyam Srinivasan>because now the bee is moving one way.

404
00:16:40.540 --> 00:16:41.760
<v Mandyam Srinivasan>The wall is moving the other way.

405
00:16:41.820 --> 00:16:44.200
<v Mandyam Srinivasan>So there's a huge image velocity seen by this eye.

406
00:16:44.280 --> 00:16:46.280
<v Mandyam Srinivasan>So the bee says, whoa, there's something dangerously close

407
00:16:46.280 --> 00:16:46.500
<v Mandyam Srinivasan>here.

408
00:16:46.660 --> 00:16:47.860
<v Mandyam Srinivasan>I better move away from that.

409
00:16:48.300 --> 00:16:49.720
<v Mandyam Srinivasan>And so that's what's happening here.

410
00:16:49.860 --> 00:16:51.260
<v Mandyam Srinivasan>So this simple technique actually

411
00:16:51.260 --> 00:16:53.800
<v Mandyam Srinivasan>tells us that bees actually negotiate narrow passages

412
00:16:54.540 --> 00:16:55.700
<v Mandyam Srinivasan>without bumping into the side walls,

413
00:16:55.900 --> 00:16:57.800
<v Mandyam Srinivasan>not by using complicated stereo mechanisms,

414
00:16:58.020 --> 00:17:00.560
<v Mandyam Srinivasan>but by actually balancing the optic flow, as it's called,

415
00:17:00.680 --> 00:17:01.880
<v Mandyam Srinivasan>on the two sides.

416
00:17:02.660 --> 00:17:05.440
<v Mandyam Srinivasan>And since we did this technique and published this paper,

417
00:17:06.040 --> 00:17:09.000
<v Mandyam Srinivasan>people have built actually robots, corridor navigating

418
00:17:09.000 --> 00:17:10.580
<v Mandyam Srinivasan>robots that use the same principle.

419
00:17:10.680 --> 00:17:12.460
<v Mandyam Srinivasan>And they find that computationally, it's

420
00:17:12.460 --> 00:17:14.400
<v Mandyam Srinivasan>a lot simpler to do it that way rather than

421
00:17:14.400 --> 00:17:17.840
<v Mandyam Srinivasan>to build stereo vision into a computer or into a robot.

422
00:17:18.600 --> 00:17:19.680
<v Mandyam Srinivasan>Also, it turns out that

423
00:17:19.860 --> 00:17:21.660
<v Mandyam Srinivasan>someone also did experiments with human beings

424
00:17:21.660 --> 00:17:24.500
<v Mandyam Srinivasan>after we did this work on bees and had a human being walking

425
00:17:24.500 --> 00:17:28.140
<v Mandyam Srinivasan>on a treadmill viewing a virtual reality setup where he could

426
00:17:28.140 --> 00:17:30.060
<v Mandyam Srinivasan>assimilate it walking along a corridor.

427
00:17:30.260 --> 00:17:32.080
<v Mandyam Srinivasan>And there you could manipulate the apparent speeds

428
00:17:32.080 --> 00:17:32.800
<v Mandyam Srinivasan>of the two walls.

429
00:17:33.000 --> 00:17:34.820
<v Mandyam Srinivasan>And they found that humans did the same thing too.

430
00:17:34.880 --> 00:17:38.380
<v Mandyam Srinivasan>So if you moved one wall in the virtual reality world,

431
00:17:38.540 --> 00:17:40.460
<v Mandyam Srinivasan>the person would veer off to one side or the other,

432
00:17:40.520 --> 00:17:42.060
<v Mandyam Srinivasan>depending upon what you did to the wall.

433
00:17:42.220 --> 00:17:45.180
<v Mandyam Srinivasan>So even though we have stereo vision, subconsciously,

434
00:17:45.180 --> 00:17:48.660
<v Mandyam Srinivasan>we're also using these optic flow cues to navigate quite often.

435
00:17:48.720 --> 00:17:49.840
<v Mandyam Srinivasan>So that's probably why you can do this.

436
00:17:49.840 --> 00:17:51.320
<v Mandyam Srinivasan>You can do a lot of things with, with, with one eye,

437
00:17:51.800 --> 00:17:53.220
<v Mandyam Srinivasan>even though you don't have stereo.

438
00:17:53.540 --> 00:17:54.920
<v Mandyam Srinivasan>We're using motion cues a lot.

439
00:17:56.500 --> 00:17:58.860
<v Mandyam Srinivasan>This just shows you a view of our setup.

440
00:17:59.100 --> 00:18:01.940
<v Mandyam Srinivasan>This is one of our earlier setups at the Australian National University.

441
00:18:03.480 --> 00:18:06.340
<v Mandyam Srinivasan>Where we do experiments with bees all year round pretty much.

442
00:18:06.440 --> 00:18:09.680
<v Mandyam Srinivasan>This is, we call it the all-weather bee flight facility, AWBFF.

443
00:18:10.300 --> 00:18:13.020
<v Mandyam Srinivasan>It's just a glorified greenhouse which maintains, you know,

444
00:18:13.020 --> 00:18:17.060
<v Mandyam Srinivasan>24 degrees during the day, 17 degrees in the evening throughout the, the year.

445
00:18:17.160 --> 00:18:19.580
<v Mandyam Srinivasan>So you can sort of, it's like a, simulates a balmy summer.

446
00:18:19.840 --> 00:18:22.160
<v Mandyam Srinivasan>Sort of a situation to them all year round.

447
00:18:22.340 --> 00:18:24.180
<v Mandyam Srinivasan>And so they happily fly indoors here.

448
00:18:24.480 --> 00:18:27.780
<v Mandyam Srinivasan>The, the, the hives are actually placed against the walls.

449
00:18:28.040 --> 00:18:29.740
<v Mandyam Srinivasan>And they have an opening to the outside.

450
00:18:30.040 --> 00:18:33.700
<v Mandyam Srinivasan>So they can fly outdoors and take care of the nutritional requirements.

451
00:18:33.900 --> 00:18:34.920
<v Mandyam Srinivasan>We don't even need to feed them.

452
00:18:35.020 --> 00:18:38.980
<v Mandyam Srinivasan>They, they live, live a pretty normal life as far as the outdoors is concerned.

453
00:18:39.280 --> 00:18:43.840
<v Mandyam Srinivasan>But in some of these experimental hives, there's another hole opening to the inside.

454
00:18:43.920 --> 00:18:47.200
<v Mandyam Srinivasan>A small hole through which we can train bees to actually come in and

455
00:18:47.200 --> 00:18:49.740
<v Mandyam Srinivasan>participate in whatever experiment we have planned for them.

456
00:18:49.840 --> 00:18:53.660
<v Mandyam Srinivasan>So we just do that by placing a sugar water feeder at the entrance that collects

457
00:18:53.660 --> 00:18:55.920
<v Mandyam Srinivasan>a few bees, then we gradually move that feeder step by step.

458
00:18:56.180 --> 00:18:59.700
<v Mandyam Srinivasan>In this case into this tunnel here, and then up to the far end.

459
00:19:00.140 --> 00:19:03.960
<v Mandyam Srinivasan>And here we have these conveyor belts with polka dots on this thing at this time.

460
00:19:04.200 --> 00:19:06.460
<v Mandyam Srinivasan>And we've got the camera overhead filming these things.

461
00:19:06.720 --> 00:19:07.840
<v Mandyam Srinivasan>So it's very simple.

462
00:19:09.480 --> 00:19:12.300
<v Mandyam Srinivasan>Now how, some of the, one of the other things we've been looking at

463
00:19:12.740 --> 00:19:15.160
<v Mandyam Srinivasan>relatively recently is how, how do insects land?

464
00:19:15.620 --> 00:19:17.740
<v Mandyam Srinivasan>Now landing, as you know, is a tricky business.

465
00:19:18.980 --> 00:19:19.820
<v Mandyam Srinivasan>It's, it's fairly quick.

466
00:19:19.840 --> 00:19:22.340
<v Mandyam Srinivasan>It's a coordinated exercise you have to do because it's quite a dangerous thing.

467
00:19:22.440 --> 00:19:25.980
<v Mandyam Srinivasan>As you approach the ground, as every pilot will know, you will have to decelerate

468
00:19:26.500 --> 00:19:29.160
<v Mandyam Srinivasan>appropriately so that we're moving slow enough when you're close to the ground

469
00:19:29.160 --> 00:19:31.920
<v Mandyam Srinivasan>that you don't bang into the ground and hurt yourself, obviously, right?

470
00:19:32.260 --> 00:19:36.640
<v Mandyam Srinivasan>Not only that, you should not decelerate too rapidly because then you come to a

471
00:19:36.640 --> 00:19:39.820
<v Mandyam Srinivasan>sort of premature halt or hover above the ground and that can be embarrassing.

472
00:19:40.720 --> 00:19:42.120
<v Mandyam Srinivasan>You've got to decide what to do next.

473
00:19:42.820 --> 00:19:44.460
<v Mandyam Srinivasan>So you want to orchestrate this landing.

474
00:19:44.580 --> 00:19:47.060
<v Mandyam Srinivasan>It's a very beautifully orchestrated timed sort of maneuver.

475
00:19:47.360 --> 00:19:49.820
<v Mandyam Srinivasan>And we asked ourselves, how do, how do flying insects do this?

476
00:19:49.820 --> 00:19:50.760
<v Mandyam Srinivasan>How do they do these nice landings?

477
00:19:51.400 --> 00:19:55.520
<v Mandyam Srinivasan>So to look at this, we, we, we train bees to come and actually land on a, on a,

478
00:19:55.560 --> 00:19:59.360
<v Mandyam Srinivasan>on a surface, a visually textured surface, just by placing a drop of sugar water on it.

479
00:19:59.480 --> 00:20:01.920
<v Mandyam Srinivasan>So a bee will come back again and again as long as the food is there.

480
00:20:02.080 --> 00:20:04.540
<v Mandyam Srinivasan>And from time to time, you shift the position of this drop so

481
00:20:04.540 --> 00:20:07.100
<v Mandyam Srinivasan>the bee doesn't know where to expect to find the drop.

482
00:20:07.320 --> 00:20:10.100
<v Mandyam Srinivasan>She has to come, look for it, and then do a landing.

483
00:20:10.560 --> 00:20:12.020
<v Mandyam Srinivasan>And then we film the landing.

484
00:20:12.140 --> 00:20:14.620
<v Mandyam Srinivasan>We're in three dimensions this time, using two cameras, or

485
00:20:14.620 --> 00:20:15.580
<v Mandyam Srinivasan>there's several ways to do it.

486
00:20:15.600 --> 00:20:16.640
<v Mandyam Srinivasan>I won't worry about the details.

487
00:20:17.100 --> 00:20:19.640
<v Mandyam Srinivasan>And, and we look at the, how the landing actually occurs.

488
00:20:19.820 --> 00:20:22.800
<v Mandyam Srinivasan>And measuring the following parameters of the landing trajectory.

489
00:20:23.040 --> 00:20:26.820
<v Mandyam Srinivasan>So as, as the bee comes in to land, you can measure its forward speed.

490
00:20:27.540 --> 00:20:31.180
<v Mandyam Srinivasan>See how that changes as, with time as the bee lands.

491
00:20:31.380 --> 00:20:32.340
<v Mandyam Srinivasan>The descent speed.

492
00:20:32.800 --> 00:20:35.040
<v Mandyam Srinivasan>See how that changes as the bee lands.

493
00:20:35.160 --> 00:20:36.340
<v Mandyam Srinivasan>The height above the ground.

494
00:20:36.540 --> 00:20:39.540
<v Mandyam Srinivasan>And see how that decreases as the bee lands, and so on.

495
00:20:39.820 --> 00:20:44.640
<v Mandyam Srinivasan>And the most sort of consistent and compelling finding to, to come out of

496
00:20:44.640 --> 00:20:48.380
<v Mandyam Srinivasan>these analyses, and I won't bore you with all the details, is that if you plot

497
00:20:48.380 --> 00:20:49.620
<v Mandyam Srinivasan>the horizontal flight speed.

498
00:20:49.820 --> 00:20:52.940
<v Mandyam Srinivasan>So this is the speed of horizontal, the horizontal component of flight speed.

499
00:20:53.360 --> 00:20:57.500
<v Mandyam Srinivasan>If you see how that varies with the height above the ground,

500
00:20:58.420 --> 00:20:59.980
<v Mandyam Srinivasan>you get a fairly straight line.

501
00:21:00.700 --> 00:21:04.020
<v Mandyam Srinivasan>There's four different examples where you have a fairly linear relationship

502
00:21:04.700 --> 00:21:08.860
<v Mandyam Srinivasan>between the speed of flight and the height above the ground.

503
00:21:09.180 --> 00:21:12.800
<v Mandyam Srinivasan>So the higher the bee, the faster it flies.

504
00:21:13.140 --> 00:21:15.560
<v Mandyam Srinivasan>The lower the bee, the slower it flies, right?

505
00:21:15.980 --> 00:21:19.240
<v Mandyam Srinivasan>So, so what this means is that,

506
00:21:20.060 --> 00:21:22.740
<v Mandyam Srinivasan>so obviously as it's coming down to the ground, it's flying slower and slower.

507
00:21:22.900 --> 00:21:26.640
<v Mandyam Srinivasan>But this linear relationship now tells you that it's actually keeping the image

508
00:21:26.640 --> 00:21:30.240
<v Mandyam Srinivasan>of the speed of the ground, speed of the image of the ground constant as it's

509
00:21:30.240 --> 00:21:30.940
<v Mandyam Srinivasan>coming in to land.

510
00:21:31.400 --> 00:21:33.900
<v Mandyam Srinivasan>So let's say you land on an aircraft and the pilot brings you down,

511
00:21:33.940 --> 00:21:35.080
<v Mandyam Srinivasan>you look at the side through the window.

512
00:21:35.260 --> 00:21:37.240
<v Mandyam Srinivasan>You find that as you get closer and closer to the ground,

513
00:21:37.380 --> 00:21:39.540
<v Mandyam Srinivasan>the ground appears to be moving faster and faster, right?

514
00:21:40.340 --> 00:21:43.740
<v Mandyam Srinivasan>Now, if you don't allow the ground speed to get faster and faster, but

515
00:21:43.740 --> 00:21:46.780
<v Mandyam Srinivasan>you adjust your speed so the ground always appears to be moving at the same speed

516
00:21:46.780 --> 00:21:49.800
<v Mandyam Srinivasan>in your eye as you come down to land, then you'll automatically, you'll get a lot of speed.

517
00:21:49.800 --> 00:21:52.180
<v Mandyam Srinivasan>And you'll automatically fly slower and slower as you come close,

518
00:21:52.180 --> 00:21:53.180
<v Mandyam Srinivasan>closer and closer to the ground.

519
00:21:53.380 --> 00:21:56.400
<v Mandyam Srinivasan>So it's a beautiful way to make a beautiful smooth touchdown because when

520
00:21:56.400 --> 00:21:58.580
<v Mandyam Srinivasan>you're close to the ground, you're now moving at very low speed so

521
00:21:58.580 --> 00:22:00.080
<v Mandyam Srinivasan>you don't burn your feet when you make contact.

522
00:22:00.520 --> 00:22:03.680
<v Mandyam Srinivasan>So it's, it's a lovely way to land because if you think about it,

523
00:22:03.700 --> 00:22:06.260
<v Mandyam Srinivasan>you don't need to know how, how you are by the ground with this technique.

524
00:22:06.500 --> 00:22:09.000
<v Mandyam Srinivasan>You don't need to know how rapidly you're approaching the ground.

525
00:22:09.360 --> 00:22:11.800
<v Mandyam Srinivasan>All you have to do is to keep the image velocity constant,

526
00:22:11.940 --> 00:22:14.940
<v Mandyam Srinivasan>and it brings you down to the ground like a magical autopilot.

527
00:22:15.360 --> 00:22:19.100
<v Mandyam Srinivasan>And this is something you wouldn't have realized until you actually observed bees

528
00:22:19.100 --> 00:22:19.640
<v Mandyam Srinivasan>and filmed them.

529
00:22:19.800 --> 00:22:20.680
<v Mandyam Srinivasan>So, so how they landed.

530
00:22:20.760 --> 00:22:23.260
<v Mandyam Srinivasan>So this is something we're trying to put into aircraft now to make them land

531
00:22:24.080 --> 00:22:24.440
<v Mandyam Srinivasan>autonomously.

532
00:22:26.520 --> 00:22:30.180
<v Mandyam Srinivasan>Okay, just to go quickly into some of the applications oriented aspect of our work.

533
00:22:32.040 --> 00:22:35.100
<v Mandyam Srinivasan>Of course, you'd like to have, to start with, to have a, a, a,

534
00:22:35.100 --> 00:22:39.480
<v Mandyam Srinivasan>a machine that has all round panoramic vision, or as wide angle as possible,

535
00:22:39.620 --> 00:22:41.380
<v Mandyam Srinivasan>to emulate the vision of an insect.

536
00:22:42.320 --> 00:22:46.260
<v Mandyam Srinivasan>Insects have lots of advantages by being able to look all around them.

537
00:22:46.360 --> 00:22:49.380
<v Mandyam Srinivasan>First of all, they can look for predators and prey everywhere around them.

538
00:22:49.800 --> 00:22:52.820
<v Mandyam Srinivasan>It also turns out that if you're using vision to rely on how you navigate,

539
00:22:53.280 --> 00:22:56.740
<v Mandyam Srinivasan>it helps a lot to be able to look all around you rather than look just in the

540
00:22:56.740 --> 00:22:57.420
<v Mandyam Srinivasan>front, for example.

541
00:22:57.420 --> 00:23:00.260
<v Mandyam Srinivasan>Because you can tell by looking at the front and the back, for example,

542
00:23:00.360 --> 00:23:03.000
<v Mandyam Srinivasan>whether you are translating or, or rotating.

543
00:23:03.520 --> 00:23:06.480
<v Mandyam Srinivasan>So separating your emotions and working out exactly what you're doing in space

544
00:23:07.080 --> 00:23:08.680
<v Mandyam Srinivasan>becomes a lot easier if you look all around you.

545
00:23:08.700 --> 00:23:11.720
<v Mandyam Srinivasan>So panoramic vision is actually a good thing to have if you're a flying,

546
00:23:11.800 --> 00:23:12.840
<v Mandyam Srinivasan>flying, flying creature.

547
00:23:13.860 --> 00:23:17.060
<v Mandyam Srinivasan>So we'd like to replicate panoramic vision into a vision system.

548
00:23:17.200 --> 00:23:18.560
<v Mandyam Srinivasan>And you can do that in several ways.

549
00:23:18.640 --> 00:23:19.780
<v Mandyam Srinivasan>You can, for example, build a compound system.

550
00:23:19.780 --> 00:23:23.880
<v Mandyam Srinivasan>eye and that's something a lot of people are starting to do now but we're not experts in that

551
00:23:23.880 --> 00:23:28.560
<v Mandyam Srinivasan>kind of optoelectronic we don't have that kind of technology know-how to do that so we instead

552
00:23:28.560 --> 00:23:34.580
<v Mandyam Srinivasan>sort of copy or emulate panoramic vision by using a simple dirty off-the-shelf kind of trick where

553
00:23:34.580 --> 00:23:40.880
<v Mandyam Srinivasan>we take a standard video camera and point it at a specially shaped shiny surface like a hubcap you

554
00:23:40.880 --> 00:23:46.960
<v Mandyam Srinivasan>know it's got a special shape though it's not quite a circle or a parabola it's a special shape

555
00:23:46.960 --> 00:23:51.620
<v Mandyam Srinivasan>and basically you point it at the shiny shape and then you can take this image and just digitally

556
00:23:51.620 --> 00:23:58.300
<v Mandyam Srinivasan>unwarp it you know make a cut here and stretch it out and then this is what you get here so what you

557
00:23:58.300 --> 00:24:02.960
<v Mandyam Srinivasan>see is virtually virtually the whole world well not entirely the whole world but a good good part

558
00:24:02.960 --> 00:24:08.900
<v Mandyam Srinivasan>of it so going from here to here represents going from zero to all the way to 360 degrees in azimuth

559
00:24:08.900 --> 00:24:14.420
<v Mandyam Srinivasan>and going from the top to the bottom is roughly viewing angle going from about there to about

560
00:24:14.420 --> 00:24:16.840
<v Mandyam Srinivasan>there so pretty much the whole world although

561
00:24:16.840 --> 00:24:20.160
<v Mandyam Srinivasan>not all of it there's a few blind zones in the system which you can overcome by doing other

562
00:24:20.160 --> 00:24:23.800
<v Mandyam Srinivasan>techniques but I won't bore you with the details so this is a panoramic vision system that we use

563
00:24:24.400 --> 00:24:29.760
<v Mandyam Srinivasan>quite routinely in in our aircraft so here's a test of this panoramic vision system before

564
00:24:29.760 --> 00:24:33.500
<v Mandyam Srinivasan>putting it onto a real aircraft you want to make sure that your vision algorithms actually working

565
00:24:33.500 --> 00:24:38.200
<v Mandyam Srinivasan>well first of all you want to test to see whether your understanding of how insects do it is

566
00:24:38.200 --> 00:24:41.900
<v Mandyam Srinivasan>actually correct because you never know and the only way to actually do put it to a test is to

567
00:24:41.900 --> 00:24:46.740
<v Mandyam Srinivasan>actually put into a some kind of a semi realistic system at least and test its performance to see

568
00:24:46.740 --> 00:24:46.820
<v Mandyam Srinivasan>how it works and how it works and how it works and how it works and how it works and how it works

569
00:24:46.820 --> 00:24:50.060
<v Mandyam Srinivasan>and how it works if it doesn't work well you know you've done something wrong with your with your

570
00:24:50.060 --> 00:24:54.380
<v Mandyam Srinivasan>concepts you don't understand the system very well so you go back to the drawing board and design

571
00:24:54.380 --> 00:24:58.500
<v Mandyam Srinivasan>more experiments to find out exactly what how the bees are doing it so here's one simple test and

572
00:24:58.500 --> 00:25:02.120
<v Mandyam Srinivasan>the idea is to put put this vision system that I mentioned to you so here's that shiny surface

573
00:25:03.780 --> 00:25:10.200
<v Mandyam Srinivasan>which is being viewed by a by a camera here and this is capturing a an all-round view of the

574
00:25:10.200 --> 00:25:15.680
<v Mandyam Srinivasan>environment here and this is made to navigate inside an artificial paper mache terrain that

575
00:25:15.680 --> 00:25:16.540
<v Mandyam Srinivasan>we've set up in the lab

576
00:25:16.540 --> 00:25:20.660
<v Mandyam Srinivasan>it's sort of some simulation of maybe the red center somewhere in the middle of Australia or

577
00:25:20.660 --> 00:25:28.020
<v Mandyam Srinivasan>maybe terrain in Mars it's just an unknown terrain and this thing is actually meant to navigate on

578
00:25:28.020 --> 00:25:33.620
<v Mandyam Srinivasan>its own from a specified point a in the terrain to another specified point B in the terrain now

579
00:25:33.620 --> 00:25:38.540
<v Mandyam Srinivasan>the interesting thing about this exercise is that the the system does not have a map of the terrain

580
00:25:38.540 --> 00:25:44.460
<v Mandyam Srinivasan>it has no knowledge of the topography or no knowledge of the map it's meant to build it up

581
00:25:44.460 --> 00:25:45.060
<v Mandyam Srinivasan>as it goes along

582
00:25:46.540 --> 00:25:54.520
<v Mandyam Srinivasan>five minutes to 25 whoo we're going okay so it's meant to go from there to there okay and it's

583
00:25:54.520 --> 00:25:59.680
<v Mandyam Srinivasan>totally unknown terrain and the principle is that actually makes small steps along the way at every

584
00:25:59.680 --> 00:26:04.960
<v Mandyam Srinivasan>point it measures the image motion all around it and based on the image motion it works out how far

585
00:26:04.960 --> 00:26:09.540
<v Mandyam Srinivasan>away the surfaces are and if they're dangerously close to colliding it goes away from them of

586
00:26:09.540 --> 00:26:14.780
<v Mandyam Srinivasan>course and it's meant to steer down the middle of this ravine here and then get to the final

587
00:26:14.780 --> 00:26:16.520
<v Mandyam Srinivasan>finishing point here it's also meant to fly out of the way of the topography and it's meant to go

588
00:26:16.540 --> 00:26:21.880
<v Mandyam Srinivasan>fairly concentrated by the ground close to the ground because as you might have guessed this is a

589
00:26:21.880 --> 00:26:26.900
<v Mandyam Srinivasan>project funded by the Air Force and they would like to have aircraft follow the terrain stay as

590
00:26:26.900 --> 00:26:32.640
<v Mandyam Srinivasan>close as possible to the ground to escape detection by enemy radar and things like that okay so here's

591
00:26:32.640 --> 00:26:38.000
<v Mandyam Srinivasan>the system doing its thing starting point specified finishing point specified and back to a

592
00:26:38.000 --> 00:26:44.460
<v Mandyam Srinivasan>starting point there we go now as I said this thing has no knowledge of its of its world it's

593
00:26:44.460 --> 00:26:45.780
<v Mandyam Srinivasan>building it up as it goes along

594
00:26:46.540 --> 00:26:53.100
<v Mandyam Srinivasan>and that's what it's doing this is being played back by the way at five times normal speed so in reality

595
00:26:53.100 --> 00:26:58.180
<v Mandyam Srinivasan>we're moving five times slower when it has to come back it is something that surprised us took a different route

596
00:27:02.820 --> 00:27:11.240
<v Mandyam Srinivasan>so it was actually smarter than us so you know I mean we always learn something new with these

597
00:27:11.240 --> 00:27:15.940
<v Mandyam Srinivasan>experiments this is why we do these experiments as it turns out in retrospect it turns out it is the right thing to do

598
00:27:16.540 --> 00:27:40.540
<v Mandyam Srinivasan>if

599
00:27:40.540 --> 00:27:44.400
<v Mandyam Srinivasan>large length and similar, you know, rotor blade diameter.

600
00:27:44.480 --> 00:27:48.320
<v Mandyam Srinivasan>This is sort of a high-end model helicopter, the Japanese

601
00:27:48.320 --> 00:27:52.100
<v Mandyam Srinivasan>helicopter, maybe wealthy Japanese business executive used to fly

602
00:27:52.100 --> 00:27:56.420
<v Mandyam Srinivasan>on weekends just for fun. So we take this helicopter and

603
00:27:56.420 --> 00:28:00.320
<v Mandyam Srinivasan>equip it with our own vision system, a special purpose design. So here's the camera looking

604
00:28:00.320 --> 00:28:04.520
<v Mandyam Srinivasan>down at the ground, a flight computer, there are the inertial sensors,

605
00:28:05.140 --> 00:28:08.100
<v Mandyam Srinivasan>gyroscopes and accelerometers which measure the physical motion of the aircraft

606
00:28:08.100 --> 00:28:11.260
<v Mandyam Srinivasan>and all the signals actually, in this particular, this is a rather

607
00:28:11.260 --> 00:28:15.960
<v Mandyam Srinivasan>earlier version of our efforts, the signals are sent down to a ground station

608
00:28:15.960 --> 00:28:20.060
<v Mandyam Srinivasan>through a television camera link and they're processed

609
00:28:20.060 --> 00:28:23.760
<v Mandyam Srinivasan>in the ground station and the appropriate control commands are sent back via

610
00:28:23.760 --> 00:28:27.980
<v Mandyam Srinivasan>the radio control system that the pilot would normally have if he or she

611
00:28:27.980 --> 00:28:31.920
<v Mandyam Srinivasan>was actually flying the aircraft manually. So the system is in close

612
00:28:31.920 --> 00:28:35.540
<v Mandyam Srinivasan>loop with the ground via this wireless link and it's trying to fly

613
00:28:35.540 --> 00:28:37.920
<v Mandyam Srinivasan>autonomously or semi-autonomously through

614
00:28:37.920 --> 00:28:38.080
<v Mandyam Srinivasan>the ground.

615
00:28:38.100 --> 00:28:39.140
<v Mandyam Srinivasan>This entire closed loop.

616
00:28:39.800 --> 00:28:43.680
<v Mandyam Srinivasan>Okay, so here's just a block diagram of this cruise control system of the aircraft,

617
00:28:43.940 --> 00:28:48.060
<v Mandyam Srinivasan>which the details I won't bother you about, but here's a nice flight cruise test.

618
00:28:48.720 --> 00:28:51.900
<v Mandyam Srinivasan>So that's the helicopter. That's the ute on which the computer

619
00:28:51.900 --> 00:28:55.900
<v Mandyam Srinivasan>is based and of course you have to drive the ute at about the same speed as the

620
00:28:55.900 --> 00:28:59.700
<v Mandyam Srinivasan>helicopter, otherwise you lose radio contact with it and that's a disaster, that's

621
00:28:59.700 --> 00:29:04.060
<v Mandyam Srinivasan>an expensive exercise. So, yeah, and this thing is traveling

622
00:29:04.060 --> 00:29:07.900
<v Mandyam Srinivasan>at about 40 or 50 kilometers an hour, so it's actually quite an exciting exercise and of course

623
00:29:07.900 --> 00:29:08.000
<v Mandyam Srinivasan>the

624
00:29:08.100 --> 00:29:13.120
<v Mandyam Srinivasan>engineer boys love doing this kind of thing. Now this forward-looking camera is

625
00:29:13.120 --> 00:29:17.400
<v Mandyam Srinivasan>not really doing anything useful particularly, it's just giving us a few thrills. But there's

626
00:29:17.400 --> 00:29:20.920
<v Mandyam Srinivasan>another camera, a downward-looking camera, which is the sort of business end of the whole

627
00:29:20.920 --> 00:29:25.380
<v Mandyam Srinivasan>system, and what this is doing is it's measuring the optic flow, the image motion produced

628
00:29:25.380 --> 00:29:30.360
<v Mandyam Srinivasan>by the ground, and holding that at a constant level to keep the height of the aircraft at

629
00:29:30.360 --> 00:29:34.820
<v Mandyam Srinivasan>a constant height by the ground. So the idea is that if the image velocity increases, it

630
00:29:34.820 --> 00:29:37.860
<v Mandyam Srinivasan>means you've gone down, you've lost altitude, so you've pulled back on the

631
00:29:37.860 --> 00:29:42.000
<v Mandyam Srinivasan>strip to increase your altitude. If the image velocity decreases, you mean you've gone too

632
00:29:42.000 --> 00:29:46.740
<v Mandyam Srinivasan>far, too high up, and you push down to bring it back down. So here's the sort of the result

633
00:29:46.740 --> 00:29:50.060
<v Mandyam Srinivasan>of trying to get this aircraft to maintain a constant height above the ground as it's

634
00:29:50.060 --> 00:29:53.920
<v Mandyam Srinivasan>flying using just this vision system. So the advantage of using this vision system and

635
00:29:53.920 --> 00:29:58.840
<v Mandyam Srinivasan>not radar or sonar is that you're not emitting any radiation, you're not giving yourself

636
00:29:58.840 --> 00:30:03.300
<v Mandyam Srinivasan>away to the enemy, and also it's very light. Video cameras now are so light and cheap and

637
00:30:03.300 --> 00:30:07.580
<v Mandyam Srinivasan>inexpensive, so you can put them on a small aircraft and help the soldier, for example,

638
00:30:07.580 --> 00:30:11.440
<v Mandyam Srinivasan>you know, in the battlefield by, the idea is the person would have a few of these things

639
00:30:11.440 --> 00:30:16.300
<v Mandyam Srinivasan>in this backpack and he would launch them to go over a hill and see if there's any danger

640
00:30:16.300 --> 00:30:21.240
<v Mandyam Srinivasan>on the other side, rather than him going out there and then exposing himself to fire. Okay,

641
00:30:21.300 --> 00:30:25.460
<v Mandyam Srinivasan>so that's that. Now if I may have, Ian, if I may have just one second, five minutes,

642
00:30:25.560 --> 00:30:30.380
<v Mandyam Srinivasan>I'll just, if you've got the patience, I'll just, five minutes, I'll finish. This is going

643
00:30:30.380 --> 00:30:34.780
<v Mandyam Srinivasan>on to the last part of the talk, which is the cognitive part. How far can you push the

644
00:30:34.780 --> 00:30:36.560
<v Mandyam Srinivasan>perception of a creature with a tiny brain?

645
00:30:37.580 --> 00:30:40.980
<v Mandyam Srinivasan>Well, we all know that rats can learn to navigate through mazes and so on, and they've

646
00:30:40.980 --> 00:30:46.080
<v Mandyam Srinivasan>been, those studies have been really beaten to death, but no one's looked to see whether

647
00:30:46.080 --> 00:30:50.280
<v Mandyam Srinivasan>insects really can navigate through mazes. So here's an attempt to train an insect to

648
00:30:50.280 --> 00:30:55.100
<v Mandyam Srinivasan>go through a complicated labyrinth like this. Several boxes here, some of these boxes have

649
00:30:55.100 --> 00:30:59.220
<v Mandyam Srinivasan>more than one exit, and the bee has to learn to find the correct exit, wherever there are

650
00:30:59.220 --> 00:31:02.460
<v Mandyam Srinivasan>multiple exits, to find the reward at the far end, right?

651
00:31:02.660 --> 00:31:07.040
<v Mandyam Srinivasan>Now in this case, we make the task a little bit easier for the bee by training it to follow

652
00:31:07.580 --> 00:31:13.680
<v Mandyam Srinivasan>a green tag that marks the correct exit, wherever there's more than one exit. So it has to follow,

653
00:31:13.720 --> 00:31:16.880
<v Mandyam Srinivasan>learn to follow the green tag, and you do that fairly simply by putting a green tag

654
00:31:16.880 --> 00:31:20.820
<v Mandyam Srinivasan>on a sugar water feeder and taking it step by step through the maze. You don't have to

655
00:31:20.820 --> 00:31:24.080
<v Mandyam Srinivasan>take it through the entire maze, just two or three steps, two or three boxes through

656
00:31:24.080 --> 00:31:28.760
<v Mandyam Srinivasan>the maze, and by then it's got the hang of, you know, the rule, which is to learn to follow

657
00:31:28.760 --> 00:31:31.880
<v Mandyam Srinivasan>the green tag, and then you can take, you know, to remove the feeder and put it at the

658
00:31:31.880 --> 00:31:35.540
<v Mandyam Srinivasan>far end, and the bee will just follow right through. And the performance is actually very

659
00:31:35.540 --> 00:31:37.320
<v Mandyam Srinivasan>good, and they can learn this thing in about a half a day.

660
00:31:38.260 --> 00:31:41.720
<v Mandyam Srinivasan>Here's a trained bee flying through one of these mazes. This happens very fast, so you'll

661
00:31:41.720 --> 00:31:50.940
<v Mandyam Srinivasan>have to be alert. So here she comes, green tag, green tag, green tag, green tag, there

662
00:31:51.980 --> 00:31:55.620
<v Mandyam Srinivasan>she goes, and then finally she comes, and that's the far, that's the destination where

663
00:31:55.620 --> 00:31:58.940
<v Mandyam Srinivasan>the reward is. So once you've trained bees to go through a maze like this, you can actually

664
00:31:58.940 --> 00:32:02.360
<v Mandyam Srinivasan>do, make, take them through any maze by just changing the arrangement of tags, and you

665
00:32:02.360 --> 00:32:05.860
<v Mandyam Srinivasan>can even make them go round and round in circles if you want to, but we'd never do that, I

666
00:32:05.860 --> 00:32:06.480
<v Mandyam Srinivasan>mean, we'd never do that.

667
00:32:08.500 --> 00:32:11.840
<v Mandyam Srinivasan>But that, that was simply following a green tag, right? But here's something a little

668
00:32:11.840 --> 00:32:15.700
<v Mandyam Srinivasan>more challenging, a little more symbolic. Now here, the idea is the bee is not simply

669
00:32:16.100 --> 00:32:20.960
<v Mandyam Srinivasan>dumbly following a green tag. They're given symbolic instructions. So whenever it enters

670
00:32:20.960 --> 00:32:25.180
<v Mandyam Srinivasan>one of these chambers, and the back wall happens to be blue in color, it means the bee has

671
00:32:25.180 --> 00:32:29.860
<v Mandyam Srinivasan>to turn right. If the black wall is yellow in color, she has to turn left. So this is

672
00:32:29.860 --> 00:32:33.880
<v Mandyam Srinivasan>a symbolic direction, sort of, that's been given to them, and they can learn this, surprisingly.

673
00:32:34.180 --> 00:32:37.560
<v Mandyam Srinivasan>And just as quickly, and just as well, as the simple rule of generalization, I'm telling

674
00:32:37.560 --> 00:32:40.940
<v Mandyam Srinivasan>just following a tag, which is always quite surprising for a bee.

675
00:32:41.060 --> 00:32:43.680
<v Mandyam Srinivasan>Unfortunately, I don't have a video to show you this, but that's the case.

676
00:32:44.160 --> 00:32:48.500
<v Mandyam Srinivasan>And of course, this final thing is the unmarked maze,

677
00:32:48.620 --> 00:32:49.560
<v Mandyam Srinivasan>which is the ultimate challenge.

678
00:32:49.640 --> 00:32:51.340
<v Mandyam Srinivasan>There's no cues or no hints anywhere.

679
00:32:51.640 --> 00:32:53.680
<v Mandyam Srinivasan>They have to learn the route, and they will learn that too.

680
00:32:53.960 --> 00:32:57.460
<v Mandyam Srinivasan>Not as well as the other ones, of course, because they've got cues in there.

681
00:32:57.540 --> 00:32:59.140
<v Mandyam Srinivasan>There's this association that the performance is not as good,

682
00:32:59.220 --> 00:33:03.200
<v Mandyam Srinivasan>but it's certainly significantly better than just randomly flying through the maze.

683
00:33:03.520 --> 00:33:07.420
<v Mandyam Srinivasan>And it turns out it's only these bees that actually learn the route through the maze.

684
00:33:07.880 --> 00:33:10.080
<v Mandyam Srinivasan>These other bees, which have been given these hints,

685
00:33:10.860 --> 00:33:12.120
<v Mandyam Srinivasan>never learn the route through the maze.

686
00:33:12.200 --> 00:33:13.620
<v Mandyam Srinivasan>They simply learn the dumb rule of following,

687
00:33:14.120 --> 00:33:16.340
<v Mandyam Srinivasan>whereas it's only these bees that actually learn the route through the maze.

688
00:33:16.420 --> 00:33:19.340
<v Mandyam Srinivasan>And that, I find, is very human, because when you're driving somewhere

689
00:33:19.340 --> 00:33:22.100
<v Mandyam Srinivasan>and your passenger is navigating, you never learn the route, do you?

690
00:33:23.240 --> 00:33:25.700
<v Mandyam Srinivasan>It's only when you have to learn it, you force learn it yourself,

691
00:33:25.780 --> 00:33:26.580
<v Mandyam Srinivasan>that you actually learn.

692
00:33:27.340 --> 00:33:27.860
<v Mandyam Srinivasan>Remarkably human.

693
00:33:28.100 --> 00:33:31.380
<v Mandyam Srinivasan>Okay, here's one more example of something which we think is fairly human,

694
00:33:31.480 --> 00:33:33.260
<v Mandyam Srinivasan>but we find in bees too.

695
00:33:33.740 --> 00:33:34.300
<v Mandyam Srinivasan>Cognitive thing.

696
00:33:34.580 --> 00:33:36.880
<v Mandyam Srinivasan>If you look at this image, and maybe some of you have seen this already,

697
00:33:36.880 --> 00:33:39.180
<v Mandyam Srinivasan>but those of you who haven't, what do you see here?

698
00:33:39.280 --> 00:33:40.780
<v Mandyam Srinivasan>You see this young lady here.

699
00:33:41.440 --> 00:33:44.100
<v Mandyam Srinivasan>I suppose you see these birds.

700
00:33:44.320 --> 00:33:45.860
<v Mandyam Srinivasan>You see these mountains in the distance.

701
00:33:46.120 --> 00:33:48.080
<v Mandyam Srinivasan>But if you look a little more carefully, maybe after a while,

702
00:33:48.140 --> 00:33:49.340
<v Mandyam Srinivasan>you might see these fingers here.

703
00:33:49.780 --> 00:33:51.120
<v Mandyam Srinivasan>And you might see this flute.

704
00:33:51.920 --> 00:33:54.040
<v Mandyam Srinivasan>And eventually you might see the face here.

705
00:33:55.580 --> 00:33:56.640
<v Mandyam Srinivasan>Does everybody see it?

706
00:33:58.040 --> 00:34:02.560
<v Mandyam Srinivasan>So once you've looked at this image and picked out all the camouflaged hidden objects,

707
00:34:03.000 --> 00:34:04.960
<v Mandyam Srinivasan>you'll never look at this image in the same way again.

708
00:34:05.180 --> 00:34:06.300
<v Mandyam Srinivasan>So every time you see this image,

709
00:34:06.300 --> 00:34:08.120
<v Mandyam Srinivasan>these hidden objects will just leap out at you.

710
00:34:08.400 --> 00:34:11.740
<v Mandyam Srinivasan>So our brains have got this amazing thing called top-down processing,

711
00:34:11.900 --> 00:34:13.960
<v Mandyam Srinivasan>where once you know what you're looking for,

712
00:34:14.080 --> 00:34:17.880
<v Mandyam Srinivasan>the brain has this wonderful capacity to reach down and pull out that signal,

713
00:34:17.920 --> 00:34:22.300
<v Mandyam Srinivasan>even though it's buried in noise, and extract whatever there is to be extracted.

714
00:34:22.600 --> 00:34:25.520
<v Mandyam Srinivasan>So we wondered whether lowly bees also have something akin

715
00:34:25.520 --> 00:34:27.740
<v Mandyam Srinivasan>to top-down processing, the kind of thing that we have.

716
00:34:28.040 --> 00:34:30.060
<v Mandyam Srinivasan>So a very simple experiment here.

717
00:34:30.120 --> 00:34:33.000
<v Mandyam Srinivasan>We trained bees to try to see if we can train bees to distinguish

718
00:34:33.000 --> 00:34:35.020
<v Mandyam Srinivasan>between two different kinds of camouflaged objects.

719
00:34:35.360 --> 00:34:36.140
<v Mandyam Srinivasan>So here's.

720
00:34:36.140 --> 00:34:36.720
<v Mandyam Srinivasan>What happens?

721
00:34:36.820 --> 00:34:39.220
<v Mandyam Srinivasan>So you've got a textured background in each wall here,

722
00:34:39.320 --> 00:34:40.780
<v Mandyam Srinivasan>which is randomly textured.

723
00:34:40.920 --> 00:34:42.900
<v Mandyam Srinivasan>And this is what you see here as a background view.

724
00:34:43.100 --> 00:34:45.100
<v Mandyam Srinivasan>And in front of that, you have a textured object.

725
00:34:45.360 --> 00:34:49.840
<v Mandyam Srinivasan>So you have a textured ring over here, which you can barely see because it's camouflaged.

726
00:34:50.260 --> 00:34:52.880
<v Mandyam Srinivasan>And then here you have a textured disk, which again,

727
00:34:52.880 --> 00:34:54.380
<v Mandyam Srinivasan>you can barely see because it's camouflaged.

728
00:34:54.620 --> 00:34:58.860
<v Mandyam Srinivasan>So you know, we tried to train bees to distinguish between the ring and the disk

729
00:34:58.860 --> 00:35:02.900
<v Mandyam Srinivasan>by training, by rewarding them on the ring.

730
00:35:03.460 --> 00:35:04.760
<v Mandyam Srinivasan>And they can't learn it.

731
00:35:04.860 --> 00:35:06.120
<v Mandyam Srinivasan>They simply cannot learn it.

732
00:35:06.140 --> 00:35:07.580
<v Mandyam Srinivasan>You can try to train them for five days.

733
00:35:08.080 --> 00:35:09.880
<v Mandyam Srinivasan>Five days is a long time in the life of a bee.

734
00:35:09.900 --> 00:35:10.880
<v Mandyam Srinivasan>They simply cannot learn it.

735
00:35:11.260 --> 00:35:15.640
<v Mandyam Srinivasan>However, if you start by training them on un-camouflaged versions of the same shapes, right?

736
00:35:15.920 --> 00:35:18.680
<v Mandyam Srinivasan>So a ring, black ring versus black disk.

737
00:35:19.020 --> 00:35:20.040
<v Mandyam Srinivasan>They learn that task.

738
00:35:20.200 --> 00:35:22.660
<v Mandyam Srinivasan>And then if you give them the hidden camouflaged objects,

739
00:35:22.840 --> 00:35:25.100
<v Mandyam Srinivasan>they will immediately pick out the correct one.

740
00:35:25.620 --> 00:35:28.000
<v Mandyam Srinivasan>So it's as though once they've got this hint of what to expect,

741
00:35:28.100 --> 00:35:32.100
<v Mandyam Srinivasan>this prior knowledge allows them to seek out the hidden object and see it.

742
00:35:32.220 --> 00:35:35.980
<v Mandyam Srinivasan>Not only that, once they've learned this thing, these trained bees can be exposed

743
00:35:35.980 --> 00:35:40.520
<v Mandyam Srinivasan>to novel hidden objects and they will learn to distinguish them with training.

744
00:35:40.700 --> 00:35:44.000
<v Mandyam Srinivasan>So you don't have to pre-train them with the un-camouflaged objects anymore.

745
00:35:44.100 --> 00:35:45.600
<v Mandyam Srinivasan>This, for example, is a horizontal bar.

746
00:35:46.040 --> 00:35:47.000
<v Mandyam Srinivasan>That's a vertical bar.

747
00:35:47.340 --> 00:35:50.400
<v Mandyam Srinivasan>So the same bees, once they've been trained to come this far,

748
00:35:50.640 --> 00:35:52.860
<v Mandyam Srinivasan>they don't need to be pre-trained on these bars.

749
00:35:53.000 --> 00:35:54.820
<v Mandyam Srinivasan>They just give it to them and they will learn it.

750
00:35:54.880 --> 00:35:59.780
<v Mandyam Srinivasan>So they've actually basically learned that this motion parallax that you get when you move from side to side,

751
00:35:59.880 --> 00:36:03.380
<v Mandyam Srinivasan>you know, it helps you disentangle or pull out the object against the background.

752
00:36:03.660 --> 00:36:05.960
<v Mandyam Srinivasan>And they've learned that that's a relevant cue to make the object move.

753
00:36:05.960 --> 00:36:07.460
<v Mandyam Srinivasan>It helps these objects pop out or be visible.

754
00:36:07.820 --> 00:36:12.460
<v Mandyam Srinivasan>And they're now seeing the world in new ways, in ways that they normally would not learn.

755
00:36:13.100 --> 00:36:17.160
<v Mandyam Srinivasan>So that's quite not bad for a bee, I think, the way they can train a bee to look at new things.

756
00:36:18.120 --> 00:36:18.960
<v Mandyam Srinivasan>OK, finish off.

757
00:36:19.220 --> 00:36:20.800
<v Mandyam Srinivasan>Very last thing, very last thing in.

758
00:36:22.120 --> 00:36:22.680
<v Mandyam Srinivasan>Associated recall.

759
00:36:22.760 --> 00:36:27.040
<v Mandyam Srinivasan>Now this, I don't know if this works for you, but if I look at this piece of thing, this image,

760
00:36:27.120 --> 00:36:27.980
<v Mandyam Srinivasan>I can almost smell lemon.

761
00:36:29.680 --> 00:36:31.020
<v Mandyam Srinivasan>Associated recall is a very powerful thing.

762
00:36:31.080 --> 00:36:35.280
<v Mandyam Srinivasan>If you listen to a piece of music, for example, it could remind you of something you were doing a long time ago as a child.

763
00:36:35.960 --> 00:36:39.800
<v Mandyam Srinivasan>Or if you get a whiff of perfume or cologne, it might remind you of someone you knew a long time ago,

764
00:36:39.980 --> 00:36:42.000
<v Mandyam Srinivasan>but don't necessarily want to remember anymore.

765
00:36:42.700 --> 00:36:44.660
<v Mandyam Srinivasan>It's a very powerful, powerful phenomenon, right?

766
00:36:45.020 --> 00:36:47.500
<v Mandyam Srinivasan>So we wondered whether associated recall is there in bees too.

767
00:36:47.860 --> 00:36:49.840
<v Mandyam Srinivasan>And whether it can trigger a sort of form of memories.

768
00:36:50.100 --> 00:36:51.060
<v Mandyam Srinivasan>So here's the experiment.

769
00:36:51.180 --> 00:36:57.120
<v Mandyam Srinivasan>The idea is we want to see whether smells, scents can trigger memories of previously visited food sources.

770
00:36:57.360 --> 00:36:58.600
<v Mandyam Srinivasan>Or flowers, for example.

771
00:36:58.880 --> 00:36:59.900
<v Mandyam Srinivasan>So here's the experiment.

772
00:37:00.040 --> 00:37:03.120
<v Mandyam Srinivasan>We train bees to go from a hive to visit a particular feeder.

773
00:37:03.240 --> 00:37:05.740
<v Mandyam Srinivasan>This is a sugar water feeder.

774
00:37:05.780 --> 00:37:05.940
<v Mandyam Srinivasan>We train bees to go from a hive to visit a particular feeder.

775
00:37:05.960 --> 00:37:07.200
<v Mandyam Srinivasan>Which is laced with the scent of rose.

776
00:37:07.480 --> 00:37:11.300
<v Mandyam Srinivasan>So bees, individually marked bees, are trained to visit that feeder for a while.

777
00:37:11.600 --> 00:37:14.760
<v Mandyam Srinivasan>Then we take away that rose scented feeder and place a different feeder,

778
00:37:14.920 --> 00:37:17.240
<v Mandyam Srinivasan>this time now carrying the scent of lemon,

779
00:37:17.420 --> 00:37:20.140
<v Mandyam Srinivasan>in a different location within the same landscape.

780
00:37:20.620 --> 00:37:22.580
<v Mandyam Srinivasan>And the same bees are now trained to visit that.

781
00:37:23.020 --> 00:37:23.980
<v Mandyam Srinivasan>Okay? For a few times.

782
00:37:24.160 --> 00:37:25.360
<v Mandyam Srinivasan>And then we go back and forth.

783
00:37:25.620 --> 00:37:27.980
<v Mandyam Srinivasan>Whenever the feeder's over here, in this formal position,

784
00:37:28.360 --> 00:37:29.360
<v Mandyam Srinivasan>it's got the scent of rose.

785
00:37:29.640 --> 00:37:31.420
<v Mandyam Srinivasan>Whenever it's over there, it's got the scent of lemon.

786
00:37:31.660 --> 00:37:33.920
<v Mandyam Srinivasan>Okay, we train them for a couple of days using this procedure.

787
00:37:33.920 --> 00:37:35.080
<v Mandyam Srinivasan>Then we go away.

788
00:37:35.220 --> 00:37:35.720
<v Mandyam Srinivasan>We can...

789
00:37:36.380 --> 00:37:37.660
<v Mandyam Srinivasan>They can do whatever they want.

790
00:37:37.740 --> 00:37:38.580
<v Mandyam Srinivasan>We do whatever we want.

791
00:37:38.680 --> 00:37:39.940
<v Mandyam Srinivasan>And then come back on Monday morning.

792
00:37:40.840 --> 00:37:42.460
<v Mandyam Srinivasan>It's fairly quiet, early in the morning.

793
00:37:42.720 --> 00:37:45.420
<v Mandyam Srinivasan>So we place two dummy feeders now.

794
00:37:45.720 --> 00:37:47.640
<v Mandyam Srinivasan>These are just empty feeders.

795
00:37:47.640 --> 00:37:48.400
<v Mandyam Srinivasan>There's no sugar water.

796
00:37:48.540 --> 00:37:49.120
<v Mandyam Srinivasan>There's no scent.

797
00:37:49.400 --> 00:37:52.980
<v Mandyam Srinivasan>One feeder where the former rose scented feeder used to be.

798
00:37:53.140 --> 00:37:55.800
<v Mandyam Srinivasan>And one dummy feeder where the former lemon scented feeder used to be.

799
00:37:55.920 --> 00:37:59.940
<v Mandyam Srinivasan>And we simply blow these scents into the hive and see what happens.

800
00:38:00.800 --> 00:38:03.080
<v Mandyam Srinivasan>So if you now blow rose scent into the hive,

801
00:38:03.080 --> 00:38:04.620
<v Mandyam Srinivasan>it's very quiet in the morning.

802
00:38:04.700 --> 00:38:05.660
<v Mandyam Srinivasan>There's hardly any bee traffic.

803
00:38:05.860 --> 00:38:07.960
<v Mandyam Srinivasan>But the moment you blow this rose scented feeder, these bees appear.

804
00:38:08.820 --> 00:38:12.560
<v Mandyam Srinivasan>And most of them visit or look for food in the former rose scented feeder.

805
00:38:12.660 --> 00:38:15.240
<v Mandyam Srinivasan>Whereas if you blow lemon scented thing,

806
00:38:15.360 --> 00:38:17.520
<v Mandyam Srinivasan>most of the bees now visit the...

807
00:38:17.520 --> 00:38:19.440
<v Mandyam Srinivasan>The same bees now go to the lemon scented feeder.

808
00:38:19.640 --> 00:38:20.620
<v Mandyam Srinivasan>So the fact...

809
00:38:20.620 --> 00:38:23.400
<v Mandyam Srinivasan>The interesting thing here is that there's no scent now anymore at the targets.

810
00:38:23.980 --> 00:38:26.540
<v Mandyam Srinivasan>The scent that they're injecting and we're injecting in the hive

811
00:38:26.540 --> 00:38:27.900
<v Mandyam Srinivasan>is simply triggering these memories.

812
00:38:28.580 --> 00:38:31.940
<v Mandyam Srinivasan>It's almost like a deja vu in these bees

813
00:38:31.940 --> 00:38:33.060
<v Mandyam Srinivasan>and that's causing them to...

814
00:38:33.060 --> 00:38:34.680
<v Mandyam Srinivasan>It's causing them to go and actually seek out these sources again.

815
00:38:34.880 --> 00:38:36.980
<v Mandyam Srinivasan>So they do seem to have this associative complex,

816
00:38:37.220 --> 00:38:39.440
<v Mandyam Srinivasan>associative recall that we seem to have.

817
00:38:39.800 --> 00:38:42.960
<v Mandyam Srinivasan>So just a glimpse of cognitive capacities in bees

818
00:38:42.960 --> 00:38:45.260
<v Mandyam Srinivasan>and I really won't go on any further.

819
00:38:45.400 --> 00:38:47.460
<v Mandyam Srinivasan>I'd like to thank my colleagues, Dr. Zhengxia Wu,

820
00:38:47.660 --> 00:38:50.260
<v Mandyam Srinivasan>who for many, many years has collaborated with me

821
00:38:50.260 --> 00:38:52.740
<v Mandyam Srinivasan>and spearheaded the effort on working with honeybees.

822
00:38:52.840 --> 00:38:55.500
<v Mandyam Srinivasan>Judith Reinhardt, who does lovely experiments on associative scent recall

823
00:38:55.500 --> 00:38:56.300
<v Mandyam Srinivasan>that I just described.

824
00:38:56.500 --> 00:38:59.020
<v Mandyam Srinivasan>Pinar Letzkus, whose work I didn't have a chance to describe,

825
00:38:59.140 --> 00:39:01.720
<v Mandyam Srinivasan>but she finds that bees are right-handed,

826
00:39:01.720 --> 00:39:04.200
<v Mandyam Srinivasan>just like we are right-eyed and most of us are right-handed,

827
00:39:04.400 --> 00:39:06.400
<v Mandyam Srinivasan>bees also are handed in their behavior.

828
00:39:06.580 --> 00:39:08.460
<v Mandyam Srinivasan>So they see better with their right eyes,

829
00:39:08.600 --> 00:39:09.820
<v Mandyam Srinivasan>they learn things better with their right eyes,

830
00:39:09.880 --> 00:39:11.820
<v Mandyam Srinivasan>they learn smells better with their right antennae.

831
00:39:12.460 --> 00:39:15.120
<v Mandyam Srinivasan>Jemman Chal, who spearheaded the robotics effort until recently,

832
00:39:15.280 --> 00:39:16.580
<v Mandyam Srinivasan>has moved to DSTU in Adelaide now.

833
00:39:17.320 --> 00:39:21.740
<v Mandyam Srinivasan>Dean Sokol, who is our pilot and mechatronics engineer,

834
00:39:22.200 --> 00:39:24.380
<v Mandyam Srinivasan>and Saul Tarragood, who is our computer engineer.

835
00:39:25.260 --> 00:39:26.820
<v Mandyam Srinivasan>So, yeah, thank you very much.

836
00:39:36.400 --> 00:39:37.120
<v Ian Gibbins>Thank you very much.

837
00:39:37.120 --> 00:39:38.240
<v Ian Gibbins>So thank you very much, Shireen.

838
00:39:38.520 --> 00:39:39.620
<v Ian Gibbins>Sorry for going over it.

839
00:39:39.660 --> 00:39:41.560
<v Ian Gibbins>No, no, I think we would have listened to it all day.

840
00:39:41.560 --> 00:39:42.280
<v Ian Gibbins>At least I would have, anyway.

841
00:39:42.960 --> 00:39:46.180
<v Ian Gibbins>Okay, we've got time for, we've got five minutes for questions.

842
00:39:46.520 --> 00:39:47.540
<v Ian Gibbins>We've got a roving microphone.

843
00:39:47.800 --> 00:39:48.420
<v Ian Gibbins>Ned, use the microphone.

844
00:39:49.540 --> 00:39:51.280
<v Ian Gibbins>So we've got a hand up here for the first one.

845
00:39:51.760 --> 00:39:53.540
<v Ian Gibbins>Just keep your questions brief if you can, and we'll get more through.

846
00:39:53.560 --> 00:39:57.300
<v Audience>Quick comment that we have racism in the human realm

847
00:39:57.300 --> 00:39:59.160
<v Audience>with the obvious meaning known by all.

848
00:39:59.560 --> 00:40:01.700
<v Audience>Perhaps a wordsmith should come up for a word.

849
00:40:01.700 --> 00:40:08.360
<v Audience>That describes the intellectual arrogance we apply to the animal kingdom

850
00:40:08.360 --> 00:40:10.660
<v Audience>that carries the same weight as racism.

851
00:40:11.020 --> 00:40:13.800
<v Audience>But anyway, is smell the dominant sense?

852
00:40:14.020 --> 00:40:17.380
<v Audience>I mean, you've described sight, but is smell the dominant sense in birds?

853
00:40:17.560 --> 00:40:20.420
<v Mandyam Srinivasan>Smell, it seems to work, at least at close range,

854
00:40:20.540 --> 00:40:23.260
<v Mandyam Srinivasan>smell seems to be the most powerful sort of cue.

855
00:40:24.000 --> 00:40:26.700
<v Mandyam Srinivasan>From afar, bees will certainly learn.

856
00:40:26.740 --> 00:40:28.160
<v Mandyam Srinivasan>I didn't have time to talk about all that.

857
00:40:28.240 --> 00:40:31.180
<v Mandyam Srinivasan>Bees will use information on the distance to the food source

858
00:40:31.180 --> 00:40:31.680
<v Mandyam Srinivasan>and the direction of the food.

859
00:40:31.700 --> 00:40:32.700
<v Mandyam Srinivasan>The direction in which to fly to it.

860
00:40:32.760 --> 00:40:34.700
<v Mandyam Srinivasan>All that's given from the dance.

861
00:40:35.120 --> 00:40:40.180
<v Mandyam Srinivasan>So a new recruit will use that information imparted by another bee

862
00:40:40.180 --> 00:40:42.080
<v Mandyam Srinivasan>to get close enough to the food source.

863
00:40:42.440 --> 00:40:44.420
<v Mandyam Srinivasan>And then, of course, when it gets close enough,

864
00:40:44.540 --> 00:40:46.520
<v Mandyam Srinivasan>if it's got a sample of nectar from the other bee,

865
00:40:46.580 --> 00:40:48.320
<v Mandyam Srinivasan>which also happens, they give them scent samples,

866
00:40:48.480 --> 00:40:50.620
<v Mandyam Srinivasan>it can home in using the scent

867
00:40:50.620 --> 00:40:52.420
<v Mandyam Srinivasan>when it's close enough to the actual food source.

868
00:40:53.060 --> 00:40:56.040
<v Mandyam Srinivasan>And then an experienced bee that learns a particular food,

869
00:40:56.260 --> 00:40:57.380
<v Mandyam Srinivasan>well, a particular plant,

870
00:40:57.500 --> 00:41:00.220
<v Mandyam Srinivasan>will learn to recognize the color of the flowers.

871
00:41:00.220 --> 00:41:01.640
<v Mandyam Srinivasan>Color learning is very important to them.

872
00:41:01.720 --> 00:41:02.920
<v Mandyam Srinivasan>Also the shape of the flowers.

873
00:41:03.200 --> 00:41:04.400
<v Mandyam Srinivasan>So they will learn things.

874
00:41:04.520 --> 00:41:05.960
<v Mandyam Srinivasan>They can also learn very abstract things

875
00:41:05.960 --> 00:41:08.080
<v Mandyam Srinivasan>like symmetry versus asymmetry.

876
00:41:08.780 --> 00:41:12.240
<v Mandyam Srinivasan>You can train bees to learn sort of fairly abstract

877
00:41:12.240 --> 00:41:13.480
<v Mandyam Srinivasan>sort of concepts like that.

878
00:41:14.400 --> 00:41:16.900
<v Mandyam Srinivasan>But at close range, smell is pretty important,

879
00:41:17.060 --> 00:41:17.860
<v Mandyam Srinivasan>to answer your question.

880
00:41:18.660 --> 00:41:19.760
<v Ian Gibbins>Question from Middle Heat.

881
00:41:20.100 --> 00:41:22.240
<v Audience>My understanding of your visual description

882
00:41:22.240 --> 00:41:24.460
<v Audience>is that the human eye, each human eye,

883
00:41:26.140 --> 00:41:28.420
<v Audience>processes one linear image

884
00:41:28.420 --> 00:41:31.340
<v Audience>and then arrives at a stereo overlap.

885
00:41:31.680 --> 00:41:35.420
<v Audience>And that the bee has 4,500 receptors in each eye

886
00:41:35.420 --> 00:41:41.120
<v Audience>and it is processing 40 or 9,000 receptor images

887
00:41:41.120 --> 00:41:44.760
<v Audience>through its brain constantly to arrive at one image.

888
00:41:45.400 --> 00:41:49.540
<v Audience>And their brains are doing it in a much smaller...

889
00:41:49.540 --> 00:41:50.440
<v Mandyam Srinivasan>Yeah, yeah.

890
00:41:50.540 --> 00:41:53.260
<v Mandyam Srinivasan>I mean, the big challenge with all visual systems, I guess,

891
00:41:53.280 --> 00:41:54.340
<v Mandyam Srinivasan>is you've got this,

892
00:41:54.360 --> 00:41:56.140
<v Mandyam Srinivasan>you're being bombarded with the sensory information.

893
00:41:56.600 --> 00:41:58.400
<v Mandyam Srinivasan>In our case, we've got 100 million receptors.

894
00:41:58.420 --> 00:42:01.420
<v Mandyam Srinivasan>In each eye, a lot of information going in

895
00:42:01.420 --> 00:42:02.920
<v Mandyam Srinivasan>and the brain has to be very selective

896
00:42:02.920 --> 00:42:05.760
<v Mandyam Srinivasan>and very good at picking out what is relevant

897
00:42:05.760 --> 00:42:08.000
<v Mandyam Srinivasan>and discarding things that are not relevant.

898
00:42:08.340 --> 00:42:10.900
<v Mandyam Srinivasan>And this is, again, a big challenge for machine vision experts.

899
00:42:11.120 --> 00:42:12.960
<v Mandyam Srinivasan>They really don't know how the brain does it

900
00:42:12.960 --> 00:42:17.980
<v Mandyam Srinivasan>and they're still fighting with algorithms that will do that job.

901
00:42:18.240 --> 00:42:19.100
<v Mandyam Srinivasan>It's not easy.

902
00:42:19.460 --> 00:42:20.420
<v Mandyam Srinivasan>Yeah, absolutely.

903
00:42:21.300 --> 00:42:22.100
<v Ian Gibbins>Question up the back.

904
00:42:22.420 --> 00:42:26.160
<v Audience>You did say that the bees that you trained

905
00:42:26.160 --> 00:42:27.160
<v Audience>had different personalities.

906
00:42:27.160 --> 00:42:29.080
<v Audience>Could you just describe that a little bit?

907
00:42:29.100 --> 00:42:29.120
<v Mandyam Srinivasan>Oh.

908
00:42:30.360 --> 00:42:34.140
<v Mandyam Srinivasan>Well, you know, especially in these training with Y-meses and so on

909
00:42:34.140 --> 00:42:35.100
<v Mandyam Srinivasan>where they have two choices

910
00:42:35.100 --> 00:42:37.000
<v Mandyam Srinivasan>and you're trying to train them to pick the correct stimulus,

911
00:42:37.220 --> 00:42:40.380
<v Mandyam Srinivasan>you find certain bees that will shoot through.

912
00:42:40.640 --> 00:42:43.100
<v Mandyam Srinivasan>They won't linger in that decision chamber.

913
00:42:43.180 --> 00:42:45.060
<v Mandyam Srinivasan>They will just make a very quick decision and shoot through.

914
00:42:46.580 --> 00:42:48.460
<v Mandyam Srinivasan>And sometimes get it right, sometimes get it wrong.

915
00:42:48.980 --> 00:42:50.780
<v Mandyam Srinivasan>But there are other bees that will hesitate.

916
00:42:50.940 --> 00:42:54.020
<v Mandyam Srinivasan>They'll vacillate, you know, for what seems like an eternity

917
00:42:54.020 --> 00:42:56.080
<v Mandyam Srinivasan>between the two choices and never make up their mind.

918
00:42:56.080 --> 00:42:58.240
<v Mandyam Srinivasan>In fact, when I was first training bees,

919
00:42:58.440 --> 00:43:00.600
<v Mandyam Srinivasan>my wife, Jascha, was helping me train them

920
00:43:00.600 --> 00:43:02.260
<v Mandyam Srinivasan>and we were doing this and there was one particular bee

921
00:43:02.260 --> 00:43:04.660
<v Mandyam Srinivasan>which would spend forever trying to make a decision.

922
00:43:04.860 --> 00:43:06.680
<v Mandyam Srinivasan>And ultimately, when it made a decision,

923
00:43:06.960 --> 00:43:08.480
<v Mandyam Srinivasan>it was usually the wrong one.

924
00:43:08.820 --> 00:43:11.700
<v Mandyam Srinivasan>So my wife called it Srini.

925
00:43:15.880 --> 00:43:17.480
<v Mandyam Srinivasan>So they really have individual personalities.

926
00:43:18.540 --> 00:43:20.200
<v Ian Gibbins>It's probably the last question up here, I think.

927
00:43:21.240 --> 00:43:23.100
<v Audience>My question relates to that.

928
00:43:23.160 --> 00:43:26.060
<v Audience>I think I was going to ask you what you made of the bees.

929
00:43:26.060 --> 00:43:29.380
<v Audience>Are there bees that actually get it wrong with the scent test?

930
00:43:31.980 --> 00:43:33.340
<v Mandyam Srinivasan>Oh, that's a good point.

931
00:43:33.460 --> 00:43:34.280
<v Mandyam Srinivasan>I don't know.

932
00:43:34.380 --> 00:43:36.020
<v Mandyam Srinivasan>There's always a few bees that...

933
00:43:36.860 --> 00:43:41.740
<v Mandyam Srinivasan>By the way, the percentage of bees that get it wrong

934
00:43:41.740 --> 00:43:43.780
<v Mandyam Srinivasan>seems to depend upon the complexity of the test.

935
00:43:43.980 --> 00:43:46.000
<v Mandyam Srinivasan>If it's a simple test like learning colours,

936
00:43:46.100 --> 00:43:48.520
<v Mandyam Srinivasan>bees will learn to discriminate two colours,

937
00:43:48.620 --> 00:43:50.940
<v Mandyam Srinivasan>blue versus green, very well.

938
00:43:51.080 --> 00:43:52.960
<v Mandyam Srinivasan>So 95% of the bees will get it right.

939
00:43:53.460 --> 00:43:56.040
<v Mandyam Srinivasan>If it's a more complicated test like going through, for example,

940
00:43:56.060 --> 00:43:58.840
<v Mandyam Srinivasan>an unmarked labyrinth or something much more complex

941
00:43:58.840 --> 00:44:01.160
<v Mandyam Srinivasan>like the associated recall that you mentioned,

942
00:44:03.060 --> 00:44:04.880
<v Mandyam Srinivasan>the success rate is not that big.

943
00:44:05.180 --> 00:44:06.880
<v Mandyam Srinivasan>And it's a good question.

944
00:44:06.940 --> 00:44:08.940
<v Mandyam Srinivasan>What is different about those bees that get it wrong?

945
00:44:09.080 --> 00:44:09.980
<v Mandyam Srinivasan>We really don't know.

946
00:44:10.240 --> 00:44:13.280
<v Mandyam Srinivasan>And is it always the same bees that get it wrong?

947
00:44:13.780 --> 00:44:16.220
<v Mandyam Srinivasan>Or is it just every bee making an error?

948
00:44:16.620 --> 00:44:18.620
<v Mandyam Srinivasan>We don't know. It would be a very interesting question.

949
00:44:18.740 --> 00:44:19.660
<v Mandyam Srinivasan>We'll have to do a detailed analysis.

950
00:44:19.800 --> 00:44:21.880
<v Mandyam Srinivasan>But I think there certainly are bees that are good learners,

951
00:44:22.300 --> 00:44:24.260
<v Mandyam Srinivasan>supposed to be bees that are not so good learners.

952
00:44:24.400 --> 00:44:26.040
<v Mandyam Srinivasan>And it's not clear whether it's because of...

953
00:44:26.060 --> 00:44:29.060
<v Mandyam Srinivasan>differential experiences before in the environment

954
00:44:29.060 --> 00:44:31.180
<v Mandyam Srinivasan>or whether there's something genetic in there too.

955
00:44:31.280 --> 00:44:33.840
<v Mandyam Srinivasan>Because as you know, these bees could come from different fathers.

956
00:44:34.600 --> 00:44:36.120
<v Mandyam Srinivasan>Because when the queen bee goes and mates,

957
00:44:36.220 --> 00:44:38.400
<v Mandyam Srinivasan>she mates with several different males and she carries that sperm

958
00:44:38.400 --> 00:44:41.920
<v Mandyam Srinivasan>and she can produce offspring that are full sisters or half sisters,

959
00:44:42.100 --> 00:44:44.320
<v Mandyam Srinivasan>depending upon whether they come from the same father or different fathers.

960
00:44:44.500 --> 00:44:47.160
<v Mandyam Srinivasan>There could be a genetic component to this behaviour too.

961
00:44:48.560 --> 00:44:49.880
<v Ian Gibbins>OK, well, we're out of time.

962
00:44:49.920 --> 00:44:51.220
<v Ian Gibbins>We need to stick to time pretty tightly.

963
00:44:51.640 --> 00:44:53.240
<v Ian Gibbins>There's actually a good point to finish on.

964
00:44:54.080 --> 00:44:55.980
<v Ian Gibbins>Right in the middle of the nature versus nurture debate,

965
00:44:56.040 --> 00:44:58.880
<v Ian Gibbins>and the bees may help us through that as well,

966
00:44:58.940 --> 00:45:00.780
<v Ian Gibbins>which would be even more marvellous.

967
00:45:00.840 --> 00:45:03.660
<v Ian Gibbins>So we've been privileged to have a really fantastic talk today.

968
00:45:05.160 --> 00:45:08.540
<v Ian Gibbins>If you're interested in other aspects of brain function,

969
00:45:08.600 --> 00:45:12.040
<v Ian Gibbins>the next talk in half an hour's time at the Radford Auditorium

970
00:45:12.580 --> 00:45:15.400
<v Ian Gibbins>is Natasha Mitchell, the host of the ABC's All In The Mind program.

971
00:45:15.760 --> 00:45:17.420
<v Ian Gibbins>She's talking about, I think,

972
00:45:17.480 --> 00:45:20.300
<v Ian Gibbins>You Are Not Your Brain, Brain Scan, or something like that,

973
00:45:20.340 --> 00:45:21.320
<v Ian Gibbins>which will be equally fascinating.

974
00:45:21.480 --> 00:45:23.580
<v Ian Gibbins>But anyway, for now, thank you, Cherenie, for a wonderful talk.

975
00:45:23.700 --> 00:45:25.660
<v Ian Gibbins>And just look after your bees, please.

976
00:45:26.040 --> 00:45:26.160
<v Ian Gibbins>Thank you.

