WEBVTT

00:00:03.497 --> 00:00:10.497
This is the Convergent Science Network podcast. Leading researchers in the domain

00:00:10.497 --> 00:00:16.777
of neuroscience, brain theory and technology are interviewed by Paul Verschoor and Tony Prescott.

00:00:20.857 --> 00:00:27.937
This is Paul Verschoor with the Convergent Science Network podcast and I'm here with Vincent Hayward.

00:00:28.097 --> 00:00:35.037
Welcome, Vincent. Vincent, who spoke today at our BCBT 2018 summer school.

00:00:36.457 --> 00:00:47.337
So, Vincent, your focus in your research, I mean, also coming from engineering, is very much haptics.

00:00:49.517 --> 00:00:56.737
And do you think haptics is a specialized sense, or do you think it's sort of

00:00:56.737 --> 00:01:01.917
full of the same general principles of organization that you might find in other senses as well?

00:01:04.397 --> 00:01:07.277
Yes, that's a complicated question.

00:01:12.657 --> 00:01:17.257
So my answer is like yes and no. It's a very easy answer.

00:01:18.177 --> 00:01:23.217
In the what I think is the case is that the.

00:01:26.631 --> 00:01:32.731
There must be general principles that are common to the senses. There is no doubt.

00:01:34.411 --> 00:01:41.911
It's not an idea, like the Gestalt held that view, that you had the perceptual principles, and they.

00:01:44.731 --> 00:01:50.211
Spent many years actually finding them, and they do exist.

00:01:51.951 --> 00:01:58.291
This morning we talked about illusions, for example, and quite a number of visual

00:01:58.291 --> 00:02:05.531
illusions have direct tactile counterparts, but many don't.

00:02:05.531 --> 00:02:10.911
And so that indicates that actually the,

00:02:13.451 --> 00:02:21.051
haptic haptic so in the sense of the mechanical relationship we have with the

00:02:21.051 --> 00:02:24.711
world really occupies a specialized,

00:02:27.251 --> 00:02:37.811
niche in our being if only because it's the functions that make us,

00:02:39.851 --> 00:02:40.771
habilis.

00:02:43.411 --> 00:02:51.291
Animals that can manipulate or walk or run efficiently.

00:02:57.851 --> 00:03:08.511
But the haptics is of course not special to humans. uh worms have it even paramecium have it uh,

00:03:14.685 --> 00:03:23.665
It's so unfortunate that actually it's a topic that's so ignored when it's in reality so important.

00:03:25.325 --> 00:03:29.345
But now you are a biased observer of that. Yes.

00:03:29.765 --> 00:03:37.465
From an evolutionary perspective, you could argue that the very first modality is touch.

00:03:37.645 --> 00:03:40.565
It's mechanosensing and chemical sensing. Yes.

00:03:41.045 --> 00:03:47.905
Well, I don't know if we know much about really early animals,

00:03:48.085 --> 00:03:50.185
but we do have them, like paramecium.

00:03:50.285 --> 00:03:52.245
It's been there for a long time, I suppose.

00:03:53.925 --> 00:03:58.185
If you want to radicalize that, you could also say, look, molecular interactions

00:03:58.185 --> 00:04:01.785
are all about shape. Shape interactions are all about shape touching.

00:04:02.605 --> 00:04:09.265
So it's most fundamental level, even at that sort of pre-single-cell organism.

00:04:10.705 --> 00:04:14.325
We talk about touch, shapes colliding. Colliding, yeah.

00:04:14.645 --> 00:04:19.945
So the question should be turned around and we say, well, maybe it's the other

00:04:19.945 --> 00:04:23.285
monalities that have been built on the principles of touch. Could be.

00:04:23.965 --> 00:04:30.865
So would you find that they're useful? But I'm not sure if there's a reason to have a...

00:04:31.725 --> 00:04:36.725
First of all, there's no hierarchy. quirky, like often it is discussed.

00:04:36.885 --> 00:04:39.225
There is no sense that's better than the other.

00:04:39.785 --> 00:04:41.485
They are there for a reason.

00:04:43.115 --> 00:04:49.775
So touch is not better than vision, or it's just touch is touch,

00:04:49.855 --> 00:04:51.815
and vision is vision, audition is audition.

00:04:53.255 --> 00:05:01.375
And in the history of the development of animals, I'm not sure that one came

00:05:01.375 --> 00:05:04.495
before or after. I don't think that's an important question.

00:05:05.035 --> 00:05:08.355
Very early organisms were sensitive to light.

00:05:08.995 --> 00:05:15.975
Plants are sensitive to light. I mean, there's very much an anthropomorphic

00:05:15.975 --> 00:05:18.935
projection we put on having to,

00:05:18.935 --> 00:05:29.575
you know, put things in a sequence as if… Yeah, but also finding a heuristic, right?

00:05:29.675 --> 00:05:36.195
And sort of try to find a way through the complexity of the organization of

00:05:36.195 --> 00:05:38.275
perceptual systems. Yeah.

00:05:38.755 --> 00:05:46.415
So you can say, well, look, touch is like the principle of touch can be generalized,

00:05:46.935 --> 00:05:50.575
to let's say a particular condition. Well, one we discussed earlier is constancy.

00:05:51.855 --> 00:06:01.275
That's a really, really fundamental principle, which is equally important in all modalities.

00:06:01.275 --> 00:06:06.675
I mean, if you didn't have constancy, your whisky would really taste terrible, because,

00:06:07.355 --> 00:06:16.475
you would have all the different chemical reactions, or anything you looked at would be a big blur.

00:06:19.415 --> 00:06:25.075
So that's really a common need. But now, for instance, if you take a worm,

00:06:25.255 --> 00:06:30.035
which will also have a sense of touch, Right?

00:06:30.195 --> 00:06:32.395
I assume. A very good one. Yeah.

00:06:32.695 --> 00:06:36.575
To what extent would its organization generalize to how, for instance,

00:06:36.655 --> 00:06:39.615
our sense of touch, that laminal fingertip, is organized?

00:06:40.315 --> 00:06:46.575
Well, that's a long jump from the world. Yeah. So the primate finger that we

00:06:46.575 --> 00:06:51.295
have is obviously very late in evolution.

00:06:52.175 --> 00:06:59.995
This pad. Of course, many animals have pads, but putting a pad at the end of

00:06:59.995 --> 00:07:04.495
a highly movable organ is really a good invention.

00:07:09.588 --> 00:07:15.568
Ursidae and felines have something like that, or rodents for that matter,

00:07:15.688 --> 00:07:23.548
but the primate has really pushed the sophistication of that organ to a great

00:07:23.548 --> 00:07:25.008
limit. It has incredible.

00:07:27.648 --> 00:07:32.788
Mechanical properties. Like one of them, for example, which is,

00:07:34.668 --> 00:07:42.308
surprising like most mechanical solids they when they are compressed,

00:07:43.708 --> 00:07:49.628
modify their global behavior like if you and tissues are not no exception like

00:07:49.628 --> 00:07:52.368
if you pull on a tendon it becomes stiffer.

00:07:54.468 --> 00:08:02.048
But somehow the fingertip doesn't you can push as many put any load on the fingertip

00:08:02.048 --> 00:08:07.088
it'll regain or keep its same elasticity.

00:08:07.388 --> 00:08:09.748
It's really amazing, actually.

00:08:10.468 --> 00:08:12.828
You can do the test with me, actually.

00:08:13.988 --> 00:08:18.508
You put like a very small load on your finger and you can see how it moves by

00:08:18.508 --> 00:08:19.768
a couple of millimeters.

00:08:20.408 --> 00:08:25.148
And I push always all my might and it moves the same amount.

00:08:26.468 --> 00:08:31.268
Exactly the same. And I span here maybe three orders of magnitude of load difference.

00:08:32.068 --> 00:08:36.668
But the mechanics have essentially remained invariant.

00:08:37.268 --> 00:08:42.588
That's a really good property. So when does that emerge in evolution?

00:08:44.511 --> 00:08:47.871
Do you know of examples of more primitive organisms that would show similar?

00:08:47.991 --> 00:08:51.111
Yeah, I think felines don't have that.

00:08:51.391 --> 00:08:56.331
I looked a little bit at the feline pads, which are really good running machines.

00:08:57.111 --> 00:09:06.431
They don't need that. It looks like a primate sort of feature,

00:09:07.491 --> 00:09:12.231
that really makes the prehensile extremity,

00:09:13.531 --> 00:09:17.671
so good at holding and sensing at the same time.

00:09:19.671 --> 00:09:27.611
So, it's a little bit unclear whether the principles are shared, right? But now, can you.

00:09:30.171 --> 00:09:39.171
Enumerate then the principles of haptics that make it really a highly reliable sense?

00:09:40.691 --> 00:09:47.491
So you would have to reorder the organizational principles of ethical well there's

00:09:47.491 --> 00:09:50.391
something I think it is rather,

00:09:52.411 --> 00:09:58.211
special to touch is the very clear,

00:10:00.431 --> 00:10:01.231
lateral,

00:10:04.131 --> 00:10:12.731
organization which seems to be a a consequence of the fact that.

00:10:15.425 --> 00:10:21.585
The ambient field that gives you a mechanical sensation is not local.

00:10:22.965 --> 00:10:25.825
So that's something very profound. Explain that.

00:10:25.985 --> 00:10:31.465
Yeah. When you look at something or you hear two sounds,

00:10:33.045 --> 00:10:43.185
which have two points in space which are very close, they create a signal on

00:10:43.185 --> 00:10:44.965
your retina that is also very close.

00:10:45.725 --> 00:10:49.725
Up to the diffraction, essentially.

00:10:50.685 --> 00:10:56.665
So it's very, very small. And a bit of also the blurring in the retina,

00:10:56.705 --> 00:11:00.285
but by and large, that's something you can rely on.

00:11:00.585 --> 00:11:05.065
Things that are close in space are close on your retina.

00:11:05.645 --> 00:11:11.865
In the auditory space, sounds that are very close also are very distinct,

00:11:11.865 --> 00:11:14.765
and you can actually take them

00:11:14.765 --> 00:11:21.005
apart even they are very close it's not the case it cannot be the case in haptics

00:11:21.005 --> 00:11:27.445
and the reason is because we are made we are a solid actually viscoelastic solida

00:11:27.445 --> 00:11:33.785
and any change in the boundary condition has far away consequences,

00:11:36.145 --> 00:11:37.945
and so it.

00:11:41.125 --> 00:11:44.025
So, small things are not local.

00:11:45.325 --> 00:11:51.125
A simple example, you read Braille. So, it's a small point. Let me close the door.

00:11:52.165 --> 00:11:53.705
Because the rest is bloody glory.

00:12:10.745 --> 00:12:15.705
Yeah, so let's go back a second. Maybe you'll restore the audio.

00:12:16.965 --> 00:12:23.465
Yeah, let's take an example. Reading Braille, of course, it's not a spontaneous

00:12:23.465 --> 00:12:26.985
activity, but it's something you can do. And...

00:12:30.207 --> 00:12:35.927
The dots are very small, so you would intuitively believe that actually it's

00:12:35.927 --> 00:12:41.347
because the dot makes a very small contact on your finger that you feel it being small.

00:12:42.027 --> 00:12:45.167
But if you actually look at the mechanics of that interaction,

00:12:45.427 --> 00:12:53.087
you realize that the braille dot has mechanical consequences that are like five

00:12:53.087 --> 00:12:54.347
or six millimeters in size.

00:12:54.507 --> 00:13:01.267
So it's a huge patch of skin that is actually disturbed by the very local contact.

00:13:01.847 --> 00:13:06.247
Of course, consciously, all you feel is that tiny little thing.

00:13:06.427 --> 00:13:11.767
But what comes from the periphery is like a really blurred, enormously blurred signal.

00:13:14.067 --> 00:13:21.387
And that's static. Now, if you have dynamics, which is actually the province

00:13:21.387 --> 00:13:24.287
of haptics, things that are dynamic and move,

00:13:25.107 --> 00:13:33.787
then you have dynamical effects in your body, which, to simplify it,

00:13:33.827 --> 00:13:36.827
could be thought of as waves, like propagation of signal.

00:13:37.287 --> 00:13:49.067
And small events in time and space become very large things in your body.

00:13:49.067 --> 00:13:51.767
But i don't so so the yeah

00:13:51.767 --> 00:13:55.507
so the consequence of that is that the neural organization

00:13:55.507 --> 00:13:59.247
has to be coping with that non-locality principle

00:13:59.247 --> 00:14:02.127
neither in time or space to have sufficient

00:14:02.127 --> 00:14:07.927
non-locality because there is topography there's mechanical coupling and there

00:14:07.927 --> 00:14:14.187
are distance relations that are fixed right so if you talk about let's say the

00:14:14.187 --> 00:14:20.887
number some millimeters of spread of a mechanical and what the information due to a braille pixel.

00:14:22.708 --> 00:14:27.608
That's a constant that's defined, that's given by the mechanical properties of the tissue, right?

00:14:27.708 --> 00:14:32.148
So in that sense, it's the same thing as having some sort of retinotopic organization

00:14:32.148 --> 00:14:42.588
that maps distance or proximal locations in visual space to proximal locations on your retina, right?

00:14:42.588 --> 00:14:49.588
So, I thought you were suggesting that the haptic system was not following such

00:14:49.588 --> 00:14:53.568
a topographic relation.

00:14:54.088 --> 00:15:01.048
Ah, no. So, what I said is, in order to achieve sufficient acuity,

00:15:03.948 --> 00:15:11.648
then the neural system has to have inputs from large regions.

00:15:13.928 --> 00:15:23.188
And that has consequences on the way, apparently, the early stages of somatization are organized.

00:15:23.688 --> 00:15:29.368
Then I can argue in the retina, you also see how it compresses the responses

00:15:29.368 --> 00:15:34.188
of many of the receptors into single neural responses.

00:15:34.508 --> 00:15:40.608
Yeah, so that reduction principle you are talking about is obviously shared

00:15:40.608 --> 00:15:46.648
by all the sensors. I mean, what's the...

00:15:51.020 --> 00:15:56.360
You know, the guy who works on the frog retina.

00:16:00.020 --> 00:16:06.080
Is it Maserano maybe? No, no, no, in the 60s. Yeah, anyway.

00:16:09.300 --> 00:16:13.560
The classic work that the frogs, I don't know, that's more the tectum, okay.

00:16:13.560 --> 00:16:24.400
Yeah, the idea that actually the organs detect the most relevant signals and reduce them,

00:16:24.540 --> 00:16:29.260
compress the information essentially to only what is relevant to behavior.

00:16:29.720 --> 00:16:32.520
That's an old idea. Of course, it's true also in touch.

00:16:32.880 --> 00:16:40.020
If you felt everything that your body experienced, then the mechanical world

00:16:40.020 --> 00:16:44.220
would be completely indescribable. It would be an incredible mess.

00:16:44.480 --> 00:16:45.960
It wouldn't be surfaces,

00:16:48.640 --> 00:16:54.820
textures, materials, shapes, weights.

00:16:57.860 --> 00:17:04.620
I mean, the example we gave at lunch, you hold a cup and it has a heaviness.

00:17:04.800 --> 00:17:08.780
You can hold it in a hundred different ways and it's the same heaviness.

00:17:10.200 --> 00:17:14.880
But that's invariance. That's invariance. Yeah, but in order to achieve that

00:17:14.880 --> 00:17:17.880
perceptual performance,

00:17:18.380 --> 00:17:26.740
the only way is that very early in the nervous system that information collected

00:17:26.740 --> 00:17:30.420
for very large regions are integrated.

00:17:32.780 --> 00:17:35.880
And maybe also across modalities. And also across modalities.

00:17:35.880 --> 00:17:40.120
But remember, we started out by trying to rank order the organizational principles

00:17:40.120 --> 00:17:46.480
of haptics. Now we're talking about some sort of topography of the organization

00:17:46.480 --> 00:17:49.600
of the receptors that you feel is unique.

00:17:51.800 --> 00:18:04.580
In terms of the raw periphery, of course the skin is the seat of the largest number of inputs,

00:18:04.800 --> 00:18:09.220
but you should also account for the muscles.

00:18:12.616 --> 00:18:17.816
And also, there is a tendency,

00:18:17.976 --> 00:18:24.776
which is very confusing, and still confuses me, in thinking that actually somatosensation

00:18:24.776 --> 00:18:30.376
is like separate channels, like proprioception and tactile inputs.

00:18:32.256 --> 00:18:35.736
But mechanical and the neural reality is not so simple.

00:18:37.916 --> 00:18:42.936
For example it's something I learned from colleagues in Sweden also like Melanie

00:18:42.936 --> 00:18:50.336
Hedin if you close your hand you do this you feel your fingers moving but the

00:18:50.336 --> 00:18:51.836
neural reality is that you have,

00:18:53.596 --> 00:18:59.236
thousands of spikes coming from the entire hand and from the skin not from the

00:18:59.236 --> 00:19:01.916
muscle that actually tell your fingers where your fingers are,

00:19:03.596 --> 00:19:10.036
but at the same time you can actually touch your hand here and separate out

00:19:10.036 --> 00:19:15.856
the perturbation given by this object and the perturbation given by my own movements.

00:19:17.196 --> 00:19:19.036
So that's actually an old idea.

00:19:20.676 --> 00:19:30.436
But the idea that actually the skin is all tactile and muscles are only proprioception

00:19:30.436 --> 00:19:35.176
is certainly untrue. and and and um.

00:19:38.030 --> 00:19:44.310
And vice versa. That's something I didn't speak about this morning,

00:19:44.350 --> 00:19:48.790
but a study done by another colleague of mine, Jean-Rito Nard, who is a former student,

00:19:50.110 --> 00:19:55.730
who was working with patients with reattached hands,

00:19:57.590 --> 00:20:06.570
after having lost a hand in a trauma, and it's being stitched back.

00:20:06.570 --> 00:20:15.930
And as you know, when the procedure is complete, then the reattached hand has no sensation.

00:20:16.950 --> 00:20:22.150
Even though the nerves may be transmitting signals, they have no significance.

00:20:22.290 --> 00:20:29.550
So you can actually bang on the hand and touch it, and there's no sensation attached to it.

00:20:30.110 --> 00:20:36.450
And then, and then Tona and his students, they tested patients to see how fast they would recover.

00:20:36.890 --> 00:20:43.210
Okay. So they had the, this patient actually ranked textures from coarse to fine.

00:20:44.150 --> 00:20:49.110
And to their astonishment, like days after the operation, they could do it.

00:20:52.770 --> 00:20:56.610
But was there dynamics to the effect? Yeah. So, right. Right,

00:20:56.710 --> 00:21:01.070
so that's why I came in, and we know that was pretty amazing, actually. No skin.

00:21:01.990 --> 00:21:07.430
And then what happens is actually the mechanical signal that,

00:21:08.170 --> 00:21:13.830
mechanics that are elicited by friction, frictional noise, if you want,

00:21:14.410 --> 00:21:18.990
let's call it this way, propagates practically in the entire arm.

00:21:20.970 --> 00:21:24.550
So there's a lot of other sensing options possible.

00:21:26.610 --> 00:21:29.690
And we replicated that study by.

00:21:32.368 --> 00:21:38.268
Anesthetizing the hands of healthy participants and they still could do it and

00:21:38.268 --> 00:21:44.428
so so that's that's really interesting it it really shows that this uh.

00:21:46.348 --> 00:21:51.568
Intuition we have that it's because we touch a place something at a particular

00:21:51.568 --> 00:21:58.968
place that's where the interaction is, is really a naive conception.

00:21:59.968 --> 00:22:05.848
The mechanical reality is many times more distributed.

00:22:06.248 --> 00:22:08.868
There are two problems we have to solve now.

00:22:10.528 --> 00:22:14.148
One, mechanosensing is not limited to the skin.

00:22:15.228 --> 00:22:21.828
This is point number one. That raises the question, what is the realm of the

00:22:21.828 --> 00:22:24.188
mechanosensing, because that's in the end what haptics is about,

00:22:24.848 --> 00:22:26.268
right? That's about the mechanosensing.

00:22:26.468 --> 00:22:29.548
So this might include the stretch receptors in your muscles and so on.

00:22:29.788 --> 00:22:32.028
It does, yeah, most definitely.

00:22:32.548 --> 00:22:39.208
It might also have, let's say, low density of sensors throughout the organs and so on, right?

00:22:40.228 --> 00:22:42.888
Misantery and... So where do you draw the boundary?

00:22:44.068 --> 00:22:48.288
I don't... You don't. Yeah, so that's what I was trying to get at.

00:22:48.428 --> 00:22:52.328
The textbook view likes to draw boundaries.

00:22:53.748 --> 00:23:01.828
You know, proprioception, it is this set of receptors, and tactile sensation, it is that set.

00:23:03.648 --> 00:23:09.248
So if you want to prove that you can, you can always come up with examples where

00:23:09.248 --> 00:23:14.368
that works, but you can also come up with as many counter-examples.

00:23:14.368 --> 00:23:21.948
If we look at the somatosensory cortex, was then the old view of Penfield,

00:23:21.948 --> 00:23:23.428
of the homunculus, also restricted?

00:23:24.428 --> 00:23:28.228
And so if we take the perspective that you now advance, which is more inclusive,

00:23:28.768 --> 00:23:32.208
what would the homunculus look like? Would it change a lot?

00:23:32.548 --> 00:23:36.808
Well, the homunculus… Add dimensions to it. I would add dimensions to it.

00:23:36.808 --> 00:23:45.788
There is no doubt that if you follow the neural pathway from a place in the

00:23:45.788 --> 00:23:53.028
body and through the different neural stages,

00:23:53.228 --> 00:24:00.948
you do observe a correspondence between body regions and cortical regions.

00:24:01.328 --> 00:24:06.928
There is no doubt about it. but the more modern views have actually shown that

00:24:06.928 --> 00:24:10.368
first of all these cortical regions are many and.

00:24:12.392 --> 00:24:17.532
Second, they can change, whereas your body doesn't.

00:24:18.632 --> 00:24:28.892
Third, that they can also be involved in other things than tactile experience, like movement.

00:24:30.272 --> 00:24:36.092
That they have bilateral, very surprising bilateral representations.

00:24:37.252 --> 00:24:40.432
That's something that I was often ignored.

00:24:40.432 --> 00:24:49.352
It's like if you drive a particular finger and then you illuminate the pits

00:24:49.352 --> 00:24:54.212
of the cortex, you also drive the cutter side.

00:24:55.332 --> 00:25:02.612
Why? I don't know, but it is an empirical finding.

00:25:05.512 --> 00:25:08.012
It's not really complexifying now, right? So basically you're saying,

00:25:08.072 --> 00:25:11.152
look, it's higher dimensional than just the skin.

00:25:11.652 --> 00:25:14.432
Yes. So let's go back to this idea of dimensions.

00:25:16.172 --> 00:25:19.912
The textbook view is the skin has two dimensions. It's a sheet.

00:25:22.152 --> 00:25:27.992
This morning we talked about mechanics. And clearly in mechanics,

00:25:28.112 --> 00:25:30.612
there is no such thing as sheets. There are solids.

00:25:31.332 --> 00:25:37.832
And solids are three dimension. dimension, and their state requires at least

00:25:37.832 --> 00:25:45.552
6 numbers at the most basic level to be represented.

00:25:47.092 --> 00:25:53.832
And worst of all, you need an infinite number of 6 numbers to be able to describe

00:25:53.832 --> 00:25:57.732
the static state of a solid.

00:25:59.532 --> 00:26:10.652
So this idea of two dimension is is is too simplified much too simplified but then,

00:26:12.512 --> 00:26:17.732
okay it's higher dimensional but then we can also add time because silver we're

00:26:17.732 --> 00:26:22.632
only talking about space yes so the time you mean dynamics yeah it means yeah so,

00:26:23.792 --> 00:26:26.312
how would you include dynamics is

00:26:26.312 --> 00:26:32.872
it basically just we are multiplying the multidimensional maps of space,

00:26:33.432 --> 00:26:37.532
and then we have some, let's say, different time windows or different frequency

00:26:37.532 --> 00:26:41.032
responses or what have you. Is this roughly how we should think about that?

00:26:42.572 --> 00:26:47.172
Yeah, you could. But I tend to have a different view.

00:26:49.812 --> 00:26:50.292
That...

00:26:54.118 --> 00:26:58.718
The brain doesn't care much about the skin itself.

00:26:59.238 --> 00:27:05.558
It really only cares about the objects that could possibly interact with your skin.

00:27:08.178 --> 00:27:12.858
So that's really a very different way of thinking.

00:27:13.198 --> 00:27:20.298
And when I say objects, I really mean the sort of mechanical effect they can create.

00:27:20.938 --> 00:27:22.878
A good example are shocks. rocks.

00:27:24.518 --> 00:27:28.378
In the old days, there was a tactile display that was extremely successful,

00:27:28.618 --> 00:27:31.298
but didn't last very long. It was called the Opticon.

00:27:31.638 --> 00:27:34.058
I don't know if you ever felt one.

00:27:34.758 --> 00:27:38.318
So the Opticon was essentially a set of little hammers.

00:27:38.998 --> 00:27:42.918
It didn't touch the skin, actually. They were actually at some distance of the skin.

00:27:43.398 --> 00:27:49.278
And the display principle was actually to make them hit, collide.

00:27:50.218 --> 00:27:55.878
Each time you have a collision, it creates quite an interesting mechanical event.

00:27:58.818 --> 00:28:03.938
But it's of course non-local, it's waves and it's complicated,

00:28:04.238 --> 00:28:09.038
but it's processed beautifully by the nervous system.

00:28:09.198 --> 00:28:13.718
It makes, if you want, quote-unquote, very sharp images.

00:28:13.918 --> 00:28:18.498
You can feel lines and motion very nicely.

00:28:18.838 --> 00:28:21.498
But the mechanical reality is an incredible mess.

00:28:22.918 --> 00:28:28.318
And the explanation for that is because actually your nervous system is tuned

00:28:28.318 --> 00:28:31.158
to collisions. That's something that is really important.

00:28:33.898 --> 00:28:39.038
As a mechanical event. And collisions, of course, have many consequences,

00:28:39.238 --> 00:28:44.758
and it's the consequences that are sensed. It's not because it's there, if you want.

00:28:45.478 --> 00:28:51.778
Right. But now... And there are lots... I was telling you at lunch about the example of a wood stick,

00:28:53.378 --> 00:29:02.378
which has special mechanics, and they have been integrated in people's brains

00:29:02.378 --> 00:29:03.698
completely seamlessly.

00:29:05.258 --> 00:29:08.738
Yeah, but then we're running ahead of it, because we're still trying to sort

00:29:08.738 --> 00:29:10.098
out the basic principles, right?

00:29:10.238 --> 00:29:15.658
And also one thing that you mentioned in your talk is that actually there are

00:29:15.658 --> 00:29:18.158
very few people working on this, right? It's not such a crowded field.

00:29:19.258 --> 00:29:23.198
And there are very few really quantitative assessments of the response properties

00:29:23.198 --> 00:29:26.138
of haptic sensing, right?

00:29:26.198 --> 00:29:34.598
So you spoke of how the scale of stimulation would sort of map to different

00:29:34.598 --> 00:29:36.358
kinds of thresholds of detection.

00:29:37.858 --> 00:29:42.418
So what do we know about that? This relationship between sort of object size

00:29:42.418 --> 00:29:48.118
or the scale of stimulation and the sensitivity we might have to then respond to, to stimuli?

00:29:51.715 --> 00:29:52.795
Well, we don't know a lot.

00:29:55.515 --> 00:30:04.895
But scale is clearly one of these information reduction principles.

00:30:06.655 --> 00:30:12.375
Events that are short in time or localized in space are treated differently from events that are,

00:30:16.035 --> 00:30:26.175
elongated long lasting if you want and occupy large portions like a whole table

00:30:26.175 --> 00:30:29.355
it's very different from a,

00:30:31.255 --> 00:30:36.715
fabric I don't know how to say that more precisely but.

00:30:40.075 --> 00:30:44.815
Well, that's the example of the elephant, right, so to touch a small object

00:30:44.815 --> 00:30:53.535
or elephant… Yeah, the elephant is proverbial.

00:30:54.075 --> 00:30:57.215
Of course. But there's always truth in proverbs.

00:30:59.955 --> 00:31:05.595
And it's much the case that if you want to have the shape of an elephant,

00:31:05.675 --> 00:31:07.695
then you have to touch it with your whole arms.

00:31:09.335 --> 00:31:18.075
And it's really the displacement that gives you the relevant information.

00:31:18.735 --> 00:31:26.555
But if you're manipulating a needle, then it's pretty useless.

00:31:26.875 --> 00:31:29.275
What matters is actually the way the skin is bent.

00:31:32.135 --> 00:31:37.815
But in this way of haptics, what do you feel has been, if you look at the whole

00:31:37.815 --> 00:31:40.875
set of experiments performed in this domain, which was quite a bit,

00:31:40.935 --> 00:31:41.915
but you know all of them. It's quite a bit, yeah.

00:31:43.055 --> 00:31:48.935
Which of these have shed the most light on the organization of this system?

00:31:49.355 --> 00:31:53.515
Which is really the crucial experiment that was a breakthrough,

00:31:53.615 --> 00:31:55.355
if you want, in understanding of haptics?

00:31:57.495 --> 00:32:04.535
I will have a rather arrogant opinion. Let me guess. in believing that there's

00:32:04.535 --> 00:32:10.855
been none so far okay really important uh that will you know why is that,

00:32:12.510 --> 00:32:16.290
Because it implies that you have some expectations. Yeah, exactly.

00:32:18.190 --> 00:32:24.210
The reason for that is actually a lot of the so-called basic understanding of

00:32:24.210 --> 00:32:29.810
touch were actually drawn from findings in vision.

00:32:31.330 --> 00:32:34.830
But as I've tried to explain,

00:32:34.970 --> 00:32:48.230
the domain is so different that you can always twist your brain into believing

00:32:48.230 --> 00:32:51.070
that touch is like vision, it has many things in common,

00:32:51.770 --> 00:32:59.330
but the number of cut examples is so large that I'm convinced now it's actually

00:32:59.330 --> 00:33:05.210
a very misleading intellectual approach.

00:33:05.210 --> 00:33:08.010
But this is also annoying, right?

00:33:08.050 --> 00:33:12.950
Because there are loads of experiments that have been performed and published

00:33:12.950 --> 00:33:19.550
in haptics, and there are all these micro-observations around something that

00:33:19.550 --> 00:33:21.270
we don't really comprehend yet.

00:33:21.410 --> 00:33:23.810
It's almost like a random sample. Yeah, yeah, yeah, certainly.

00:33:25.170 --> 00:33:28.050
So what are we missing then, in your opinion? what's what uh

00:33:28.050 --> 00:33:30.890
uh yeah i did write a

00:33:30.890 --> 00:33:33.570
grant on this topic uh that was

00:33:33.570 --> 00:33:42.130
about 10 years ago now almost and and and the uh basic argument is that in in

00:33:42.130 --> 00:33:49.030
um in somatosensational touch there was nothing like like like in vision in

00:33:49.030 --> 00:33:51.450
the sense like uh kondring and mar,

00:33:52.410 --> 00:34:01.810
are outlined, and then the neural coalescence that were found from the theoretical consideration.

00:34:05.066 --> 00:34:08.866
And so I promised to do that, actually.

00:34:09.166 --> 00:34:12.066
Yeah, but that's a knowing argument, right? Because you said,

00:34:12.286 --> 00:34:16.486
we don't want to compare to vision. And now you say, well, we've made a bunch of progress in vision.

00:34:16.946 --> 00:34:21.926
Yeah, so we have to do the same in audition. Well, in audition too, actually.

00:34:22.246 --> 00:34:27.566
But maybe the progress in vision is more illusory than real.

00:34:27.886 --> 00:34:32.486
Maybe. A lot of people are very busy. But if you really ask,

00:34:32.666 --> 00:34:34.506
like, okay, what are the...

00:34:35.066 --> 00:34:38.546
The fundamental principles that really allow us to comprehend the system and

00:34:38.546 --> 00:34:41.366
how this works? Well, there's a litmus test, actually.

00:34:42.786 --> 00:34:54.726
After all these centuries of research, there are pretty good artificial vision systems.

00:34:56.206 --> 00:35:03.646
Whether they are biologically relevant or not is another question, but they do function.

00:35:05.066 --> 00:35:13.986
And many of them actually have drawn, consciously or not, ideas from how natural vision operates.

00:35:15.666 --> 00:35:17.426
I disagree, but go ahead.

00:35:19.206 --> 00:35:21.326
I didn't say all, some.

00:35:23.386 --> 00:35:27.046
But when you look at the state of the art of artificial touch,

00:35:28.926 --> 00:35:31.146
it's really a terrible situation.

00:35:32.586 --> 00:35:34.746
Sure. There's no comparison.

00:35:37.006 --> 00:35:40.646
That's tricky because you can, of course, also say like, well,

00:35:40.786 --> 00:35:44.966
we made great progress with building microprocessors, right?

00:35:45.266 --> 00:35:46.766
Yeah. So we know about cognition.

00:35:47.086 --> 00:35:49.606
No, that's not what I mean.

00:35:50.026 --> 00:35:54.746
There are just no market drivers for advanced haptic systems.

00:35:55.546 --> 00:35:59.766
There are lots of them. There are market drivers for cheap vision.

00:36:00.246 --> 00:36:04.266
Yeah. I'm not sure I find it a very convincing argument.

00:36:05.466 --> 00:36:14.966
No, it's not an argument. It's an observation that the computer vision actually exists,

00:36:16.066 --> 00:36:21.306
but computer touch is essentially non-existent.

00:36:21.466 --> 00:36:26.186
And if you look at most of the activity in that domain, I mean,

00:36:26.186 --> 00:36:31.526
actually, it's a survival copy of a visual.

00:36:32.346 --> 00:36:36.586
Right. Okay, so the bottom line would definitely be that you're saying,

00:36:36.686 --> 00:36:39.966
well, we're actually really in the dark. We're in the dark about.

00:36:42.002 --> 00:36:45.002
Yeah i would uh would agree with that yeah then where

00:36:45.002 --> 00:36:48.162
where should we start because what you present one experiment that's

00:36:48.162 --> 00:36:51.542
always extremely interesting where you had these these actuation arrays

00:36:51.542 --> 00:36:54.322
that you either stimulated in uh

00:36:54.322 --> 00:36:57.622
in sync so they're oscillating at some frequency i don't remember which one

00:36:57.622 --> 00:37:04.182
but either they were they were synchronized or they were um not synchronized

00:37:04.182 --> 00:37:11.882
right so you said actually they were out of phase right yeah which led to two very different,

00:37:12.562 --> 00:37:14.742
experiences from the touch perspective, right?

00:37:14.782 --> 00:37:19.002
So one, the first one, synchronized, it's like a solid or like a solid surface.

00:37:19.322 --> 00:37:24.402
And the other one feels like there's some viscous media. It's the other way, but...

00:37:25.202 --> 00:37:29.342
This is very interesting. But now we bring in dynamics, right?

00:37:29.382 --> 00:37:33.362
Now we have time, we have movement, change. Right.

00:37:34.102 --> 00:37:38.222
So is that the thing that we've been missing? Because also, in vision,

00:37:38.282 --> 00:37:42.882
you see the same thing. Vision is very much analyzed in terms of spatial maps,

00:37:43.182 --> 00:37:44.762
hierarchies of spatial maps.

00:37:44.862 --> 00:37:49.162
If you look at therapy, that's what it's all about. How can I learn hierarchies of spatial maps?

00:37:49.342 --> 00:37:53.942
And I have lots of wires between the maps, going back to Rosenblatt and then,

00:37:54.602 --> 00:37:55.542
Ptolemaic and whatever.

00:37:55.982 --> 00:38:00.302
And then, of course, we know anatomically, actually the wires between layers

00:38:00.302 --> 00:38:05.862
in cortex or regions in cortex, V1, V2, and so on, is rather limited. with it.

00:38:06.042 --> 00:38:09.142
It's less than 3% of your synaptic volume.

00:38:09.562 --> 00:38:13.142
So when there you see that the approach taken is sort of orthogonal to what

00:38:13.142 --> 00:38:14.502
we understand of the biology.

00:38:15.102 --> 00:38:21.322
But is this maybe a key insight at which you say, well, maybe we have not opened

00:38:21.322 --> 00:38:25.362
up enough to rethink what that haptic code should be.

00:38:25.522 --> 00:38:29.262
And maybe the haptic code is much more temporal than spatial.

00:38:29.702 --> 00:38:33.302
Would that be an entry point for you? Yeah, that would be correct.

00:38:33.302 --> 00:38:38.082
The temporal performance of touch is excellent.

00:38:38.582 --> 00:38:41.082
It's almost a driver's edition.

00:38:43.299 --> 00:38:54.559
Actually. And evidence is accumulating that if you want, the property of objects that are,

00:38:55.459 --> 00:39:02.119
in contact with your body are acquired through timing.

00:39:05.639 --> 00:39:07.919
I'll give you a simple example of that.

00:39:10.519 --> 00:39:17.299
Actually, it was Flanagan and Johansson, they put the hypothesis that actually,

00:39:17.359 --> 00:39:21.259
how do you know that you're actually touching a table?

00:39:22.479 --> 00:39:26.359
Well the code is simply because all the afferents in ad regions are responding

00:39:26.359 --> 00:39:30.019
at the same time, from the first spike. like. It's a great idea.

00:39:31.239 --> 00:39:35.939
And it's not important that they are all from the same place,

00:39:35.959 --> 00:39:37.439
it's because they are at the same time.

00:39:39.659 --> 00:39:45.019
And it's probably true. I'm pretty sure that… Let's follow this,

00:39:45.059 --> 00:39:46.419
right? So now I'm tapping the table.

00:39:47.099 --> 00:39:50.139
Okay, so I have a synchronized response in my fingertip.

00:39:51.119 --> 00:39:55.519
This thing goes up to my enthalomus and from there into the cortex.

00:39:56.139 --> 00:39:57.779
How many processing steps?

00:39:58.659 --> 00:40:03.279
How many? How many processing steps? Before you feel it? Before I'm going to say, oh, it's stable.

00:40:05.559 --> 00:40:09.739
One in the brainstem, which we commented

00:40:09.739 --> 00:40:12.999
this morning is probably quite sophisticated. Cuneus, right? Yeah.

00:40:14.239 --> 00:40:20.859
One in the thalamus. Nobody knows. There's almost no study on the uh...

00:40:25.123 --> 00:40:30.163
Function of the thalamus in somatosensation, almost none. But it should be like

00:40:30.163 --> 00:40:31.523
the others, more specialized.

00:40:32.863 --> 00:40:35.183
Yeah, there should be something like the superior colliculus,

00:40:35.363 --> 00:40:37.843
but for touch, somewhere. It's called VPN.

00:40:38.563 --> 00:40:43.903
Yeah, it has a name, of course, because the anatomists have identified it.

00:40:44.263 --> 00:40:50.323
But the function is a complete mystery, or a terra cognita.

00:40:51.363 --> 00:40:57.603
And then you have the cortex, And then you have the primary area,

00:40:57.963 --> 00:41:05.023
S1, which actually has the subdivision, and then you have S2,

00:41:05.223 --> 00:41:13.103
and then maybe something, I think there's something like S3,

00:41:13.243 --> 00:41:16.023
I'm not quite sure, and then it becomes Pareto.

00:41:17.223 --> 00:41:23.603
And that's probably where the integration really takes place.

00:41:24.723 --> 00:41:30.203
So, if you can count, no, we can go back and count, it's quite a few stages, actually.

00:41:30.863 --> 00:41:35.843
But based on what you're saying, then, the real experience sits in the parietal

00:41:35.843 --> 00:41:38.143
area. Oh, I don't know. No, no, no, that's where I went.

00:41:38.583 --> 00:41:42.483
And from there, we might then speculate, because that's what this is all about,

00:41:42.683 --> 00:41:45.743
that it's right then also indeed a multimodal construct. For sure.

00:41:46.083 --> 00:41:50.943
Yeah, lots of examples of that, including some of my studies.

00:41:52.363 --> 00:41:57.903
Do we know anything about the latency between this touch and 20 milliseconds yeah.

00:41:59.764 --> 00:42:03.884
The experience? Yeah. In the parietal cortex?

00:42:05.264 --> 00:42:09.964
Probably longer. Yeah. It's 20 milliseconds to the primary areas. Okay.

00:42:10.144 --> 00:42:15.124
And then I don't know. I could know. I don't have the number in my head now.

00:42:15.484 --> 00:42:20.644
But I have a former colleague who has actually done some really interesting

00:42:20.644 --> 00:42:26.584
studies recently on that using high-performance EEGs. Okay.

00:42:27.164 --> 00:42:31.644
So now then, the other thing you pointed to, and still, again,

00:42:31.704 --> 00:42:35.664
now we go all the way back to the periphery, is how already the periphery is

00:42:35.664 --> 00:42:36.764
being really manipulated,

00:42:37.164 --> 00:42:45.804
how the viscosity or the rigidity of the skin itself at the fingertip is of direct relevance.

00:42:46.304 --> 00:42:52.964
Direct, yeah. To what you feel. Both the control of grasp and the touch sensation.

00:42:53.404 --> 00:42:58.344
Yeah. So, what you were saying is that depending on how liquid or how moist

00:42:58.344 --> 00:43:00.784
the skin is, of course, slippage will be different.

00:43:01.004 --> 00:43:10.124
But for some reason, that grass is automatically tuned to the wetness of the skin. Yes, it has to.

00:43:11.544 --> 00:43:14.764
So, why does it have to be like that?

00:43:15.084 --> 00:43:20.744
Well, I don't know if it has to be like that, but we're living tissues.

00:43:21.604 --> 00:43:23.304
So, we have to have water.

00:43:24.964 --> 00:43:30.164
And the repertory of materials is not so great.

00:43:30.284 --> 00:43:34.104
I mean, you have collagen fibers, and you have keratin, and you have a certain

00:43:34.104 --> 00:43:37.284
set of possible materials you use.

00:43:38.969 --> 00:43:45.649
That actually have the structural properties for making an animal.

00:43:46.569 --> 00:43:49.549
So that's what you have. And they are good or bad, I don't know,

00:43:49.589 --> 00:43:50.609
but they are what they are.

00:43:56.749 --> 00:44:04.209
And it's likely that the neural system co-evolves to deal with that because you don't have a choice.

00:44:05.429 --> 00:44:08.229
Well, but look, it sounds good. it and then

00:44:08.229 --> 00:44:12.949
also you brought robust control right but this

00:44:12.949 --> 00:44:15.669
is based now on a number of assumptions that i think we

00:44:15.669 --> 00:44:19.809
should be clear about because in some sense now we get complexification because

00:44:19.809 --> 00:44:25.009
you're saying well i i only have touch and touch is mapped to grass and for

00:44:25.009 --> 00:44:31.709
some reason i adjust the grass force to the friction force i have on my skin

00:44:31.709 --> 00:44:34.769
yeah which in turn depends on how wet my skin is.

00:44:34.949 --> 00:44:40.509
And not only on the wetness, but the material. Of course. If it's glass or wood.

00:44:41.869 --> 00:44:46.949
So that loop is closed. Apparently, it's not a loop.

00:44:47.789 --> 00:44:50.449
Much of it is open.

00:44:51.849 --> 00:44:56.769
Even worse. Now your trouble is even greater.

00:44:57.189 --> 00:45:05.209
Because to make that work, I must sense how slippery That's where my skin is. Right.

00:45:05.689 --> 00:45:08.329
So you look at what you're...

00:45:09.509 --> 00:45:17.609
And if you don't look, then you have this over-powerful grip.

00:45:18.969 --> 00:45:24.769
So it means the controller here is basically lacking any further information.

00:45:25.269 --> 00:45:29.949
It's the worst case, yeah. Maximum. Or something slippery, yeah.

00:45:30.649 --> 00:45:35.149
Like if your life depends on it, that's uh you know you're grabbing a handle

00:45:35.149 --> 00:45:40.069
you're not going to group it for you just go and right,

00:45:40.849 --> 00:45:44.949
okay so you're saying this fine tuning of control is learned that because it

00:45:44.949 --> 00:45:50.909
depends on other modalities and would depend on yeah it it is learned um i mean

00:45:50.909 --> 00:45:54.989
i don't think it's innate i mean some of it is innate as far as i can tell okay

00:45:54.989 --> 00:45:58.329
yeah but you went in a direction that that,

00:46:00.161 --> 00:46:02.701
Actually, that would be a good place to look for innate.

00:46:05.241 --> 00:46:11.161
What I was thinking, it actually creates great opportunities for control because

00:46:11.161 --> 00:46:15.921
I can also say, okay, if I have to slip, if I want to control the slip of an

00:46:15.921 --> 00:46:18.221
object, because we don't always hold everything in a rigid way,

00:46:18.801 --> 00:46:21.501
we also have to slip to also reduce damage. Not only we...

00:46:24.161 --> 00:46:28.061
Actually, it's a very interesting point nobody discussed. because most of the

00:46:28.061 --> 00:46:34.781
time you don't want to have your stable contact unless you are holding a cup or eating something.

00:46:35.141 --> 00:46:43.581
But many of the cases, you don't want to actually sleep or being quite mobile.

00:46:44.061 --> 00:46:48.101
But the mechanical relationship, as we saw this morning, is incredibly varied.

00:46:48.101 --> 00:46:58.321
So the motor loops have to be able to deal with that very large variability

00:46:58.321 --> 00:47:01.761
and make sure that sleep happens.

00:47:02.301 --> 00:47:06.721
Exactly. The other thing that we didn't talk about, in other words you didn't

00:47:06.721 --> 00:47:12.381
mention, if I'm over-conservative I'm damaging my skin.

00:47:12.621 --> 00:47:15.141
Oh, you do, yeah. Which is a huge cost. Which is pretty bad,

00:47:15.221 --> 00:47:19.521
yeah. So it's actually not even that the error is on one side only,

00:47:19.601 --> 00:47:23.181
like the open drops, the catastrophic failure. You can get cuts.

00:47:24.201 --> 00:47:28.381
Also, there's an error at the other side, where it leads to self-damage,

00:47:28.481 --> 00:47:31.221
which would actually maybe be a higher cost.

00:47:31.701 --> 00:47:39.181
Yeah, self-damage happens, actually, mechanically speaking, in two possible ways.

00:47:39.181 --> 00:47:43.281
Either you have abrasion so it means that you have.

00:47:44.941 --> 00:47:51.841
Surfaces that are sharp edges and you slip against them actually it's the only

00:47:51.841 --> 00:47:57.601
case or you can have an edge which is basically a simplified version of abrasion,

00:47:58.874 --> 00:48:07.254
And so these are really extreme cases. In a natural object, you have thorns,

00:48:07.314 --> 00:48:09.874
of course, which have been designed for that.

00:48:10.634 --> 00:48:21.334
But most objects are innocuous, like stones and wood, unless they have thorns.

00:48:22.514 --> 00:48:27.234
And of course, in the modern world, you have metal sharp and,

00:48:28.334 --> 00:48:32.934
Well, it leads to, also, if we compare it in all other modalities,

00:48:33.274 --> 00:48:40.214
no, not all, yes, well maybe not all, there is always an automatic calibration to intensity.

00:48:41.154 --> 00:48:44.354
Yes, that's a tricky question.

00:48:44.654 --> 00:48:46.074
That's maybe the realm we're in,

00:48:46.234 --> 00:48:50.534
because do you want to normalize friction forces that you're exposed to?

00:48:50.534 --> 00:48:55.414
And do you want to do that by sensing the actual friction forces or actually sensing,

00:48:56.894 --> 00:49:02.474
the moistness of the skin right to make sure you sort of recalibrate continuously

00:49:02.474 --> 00:49:07.414
when you say sensing you said something very.

00:49:08.914 --> 00:49:12.994
Precise you mean you had a sensor you mean for the first time but you don't

00:49:12.994 --> 00:49:18.834
have any sensor for moistness or you have no sensor for friction this is where I disagree with you,

00:49:19.714 --> 00:49:24.894
For instance, we do know that the sweat correlates with resistivity.

00:49:26.314 --> 00:49:29.874
This makes it therefore, in theory, possible to measure resistivity.

00:49:30.014 --> 00:49:32.654
How? Which receptor would do that?

00:49:34.145 --> 00:49:39.865
Well, the resistivity of the skin would have electrical consequences, right?

00:49:40.185 --> 00:49:44.785
Oh, you mean new properties of ion flows might be affected.

00:49:45.105 --> 00:49:48.145
Might the speed to which I can respond will be affected. Why not?

00:49:48.225 --> 00:49:52.285
Wow, that's a pretty daring hypothesis.

00:49:57.385 --> 00:49:57.905
Moreover,

00:50:00.065 --> 00:50:04.725
the field is only two people by itself. it's worth thinking about,

00:50:06.325 --> 00:50:12.665
actually because moistness is definitely a sensation it has it's like,

00:50:13.685 --> 00:50:19.105
you know it's like a kind of a color if you want it has it's and it's really

00:50:19.105 --> 00:50:21.405
actually interesting to,

00:50:22.745 --> 00:50:28.785
see how we actually experience wetness or presence of a film of water between

00:50:28.785 --> 00:50:29.885
your hand and the object object.

00:50:31.205 --> 00:50:34.465
It seems like there are two ways to do it.

00:50:34.645 --> 00:50:40.045
Either the object is impermeable, like we said this morning,

00:50:40.105 --> 00:50:43.745
like a glass surface, where you have some kind of lubrification.

00:50:44.865 --> 00:50:51.105
And you have another class of objects which is very common, they are porous,

00:50:51.145 --> 00:50:55.445
so they actually diffuse water.

00:50:57.025 --> 00:51:03.045
And in the two cases, they are correlated to thermal behavior.

00:51:05.085 --> 00:51:09.505
So the thermal bridge, which is made by an impermeable surface,

00:51:09.645 --> 00:51:13.465
is totally different from the thermal bridge made by a fabric.

00:51:14.125 --> 00:51:24.485
And so it's this combination of mechanics and thermal input that creates the sensation of wetness.

00:51:26.425 --> 00:51:35.585
Also. Yeah. Actually, in VR, it's possible to make things that feel wet if you

00:51:35.585 --> 00:51:40.225
do recreate artificially the thermal profile.

00:51:41.585 --> 00:51:44.505
But maybe the answer sits in the keratin itself.

00:51:44.565 --> 00:51:50.145
You gave this example where you say keratin changes its viscosity.

00:51:50.885 --> 00:51:58.245
Yes. well yeah it's a stiffness it becomes very stiff very soft but um.

00:52:00.071 --> 00:52:03.671
This will again depend on how much moisture there is.

00:52:03.831 --> 00:52:09.671
So the speed to which this changes will tell you implicitly about moisture.

00:52:09.991 --> 00:52:13.431
Would you agree with that? Completely. So didn't we say the same thing about

00:52:13.431 --> 00:52:17.071
that? Completely, yeah. Because then… That's one aspect.

00:52:17.631 --> 00:52:23.891
Yeah, so that would be the tribology, the way the skin slips and drags on the object.

00:52:24.011 --> 00:52:29.771
And the second is the thermal behavior. So you have two, and the two correlated

00:52:29.771 --> 00:52:37.751
will give you a possibility to estimate the water content.

00:52:39.271 --> 00:52:42.751
And as I said, you can do it in VR quite easily.

00:52:42.911 --> 00:52:50.471
I mean, relatively easily by modifying the temperature of a surface according to certain profiles.

00:52:50.711 --> 00:52:56.211
And then you do get pretty nice. It's interesting, because in all other modalities,

00:52:56.331 --> 00:52:58.691
issues of gain control are very central.

00:52:59.271 --> 00:53:02.111
People spend a lot of time... Oh yeah, I know what you mean.

00:53:02.251 --> 00:53:04.251
Like in Audition, it's tremendous.

00:53:07.491 --> 00:53:13.351
It's like five orders of magnitude of gating mechanically and neurally.

00:53:14.711 --> 00:53:18.131
It's interesting for hefty... The gain control...

00:53:18.751 --> 00:53:27.271
Because that's what we're talking about. Yeah, so the dynamics in touch is about,

00:53:27.411 --> 00:53:29.951
if I recall, five orders of magnitude.

00:53:31.031 --> 00:53:34.791
So from the smallest load to the biggest load, if you want.

00:53:35.591 --> 00:53:41.731
And it does have this log profile, if you want. Otherwise, you would not have

00:53:41.731 --> 00:53:43.511
all these orders of magnitude. Right.

00:53:46.211 --> 00:53:50.431
But the notion of intensity is a slippery one in touch. actually.

00:53:53.871 --> 00:53:57.751
You can talk about the intensity of a vibration, you can. But why not?

00:53:57.891 --> 00:54:01.391
Well, it can be slippage or force. Yeah.

00:54:01.991 --> 00:54:05.871
Well, force is a very misleading idea.

00:54:06.091 --> 00:54:11.131
Force only made any sense when you have a point mass.

00:54:11.591 --> 00:54:14.511
There's no point mass, as far as I can tell, in this room.

00:54:15.349 --> 00:54:18.369
One of those examples, you were pressing on the table. So what you have,

00:54:18.689 --> 00:54:24.669
in mechanical, in theoretical mechanics, you talk about, you don't talk about

00:54:24.669 --> 00:54:26.569
force, you talk about load.

00:54:27.329 --> 00:54:30.129
So it's a very different idea. It's also measured in Newton.

00:54:30.309 --> 00:54:36.729
But load is something which happens on a domain, okay, not at a point.

00:54:37.829 --> 00:54:42.809
And so the load, and I don't think the force has any representation.

00:54:42.809 --> 00:54:45.609
In this is tricky because

00:54:45.609 --> 00:54:52.669
if we if we now exert force on the tabletop yeah I have muscles generating force

00:54:52.669 --> 00:54:57.869
load activity yeah but the muscles themselves are contracting no they are they

00:54:57.869 --> 00:55:02.389
are contracting and they are viscoelastic yeah and they change length and they

00:55:02.389 --> 00:55:04.729
create load yes they don't create force,

00:55:05.529 --> 00:55:11.869
I could imagine this as force yeah aha yeah but you use a horrible trick do

00:55:11.869 --> 00:55:19.449
I yeah Yeah, you put your finger on an object that is completely rigid. Yes.

00:55:19.709 --> 00:55:24.889
And then you average out all the interaction and you call it a force.

00:55:25.229 --> 00:55:28.609
Isn't that what experiments are about? Controlled conditions?

00:55:29.909 --> 00:55:36.529
Yes, for sure. But it doesn't mean that it's useful information for your… Well,

00:55:36.529 --> 00:55:41.529
at least now we can measure what I can say now. Now, okay, I push on this surface.

00:55:43.389 --> 00:55:50.349
There is an initial load exerted on the fingertip. Yeah. But across my whole arm.

00:55:50.529 --> 00:55:54.709
Yes. The whole thing is firing like crazy.

00:55:54.829 --> 00:55:59.709
Exactly. There are forces being generated in the sense that the contractile

00:55:59.709 --> 00:56:02.049
properties of my muscles. Gold organs.

00:56:02.489 --> 00:56:04.589
I can measure as forces.

00:56:05.569 --> 00:56:11.009
Oh, there are loads there, actually. Actually, so the way I look at it is purely geometrical.

00:56:12.936 --> 00:56:17.796
And actually, in theoretical mechanics, that's all you need.

00:56:18.776 --> 00:56:22.176
Well, apparently not, because we haven't explained haptics yet, but go ahead. Yeah.

00:56:22.676 --> 00:56:28.996
All you need to know is how the bodies change shape.

00:56:30.356 --> 00:56:38.676
And the notion of force is almost an artifact.

00:56:38.756 --> 00:56:42.076
It's like a convenience.

00:56:43.016 --> 00:56:48.756
But it's not contradictory because I could say a set of forces adds up to load.

00:56:49.956 --> 00:56:54.756
Yeah, but it's an infinite set. But still, if I have the controlled conditions

00:56:54.756 --> 00:56:58.656
now, define the boundary conditions. Okay, I'll give you a counter example to this argument.

00:56:58.956 --> 00:57:05.456
I push with a flat plate with one Newton on your fingertip and then you have this experience.

00:57:05.936 --> 00:57:11.036
Now I take a sharp needle and I do 1 newton, you'll have a very different experience.

00:57:13.136 --> 00:57:17.176
But from a mechanical perspective... It's a very different situation, same force.

00:57:18.976 --> 00:57:25.776
Sure, okay, that's fair enough. But maybe what we're looking at here is a microscopic reductionist.

00:57:26.376 --> 00:57:28.596
No, it's really what happens, you bleed.

00:57:29.496 --> 00:57:34.216
It's not a reductionist. It's a consequence of... But if I just...

00:57:34.216 --> 00:57:36.856
I'm talking about me pushing the tabletop, right?

00:57:37.996 --> 00:57:44.656
Is it useful to think about the muscle fibers, because now I have millions of

00:57:44.656 --> 00:57:48.856
muscle fibers working together, contracting or not, or relaxing,

00:57:49.036 --> 00:57:51.616
to generate this behavior?

00:57:52.216 --> 00:57:59.116
Yeah, so it will load. Is it useful to look at every muscle fiber from the perspective of load?

00:58:00.296 --> 00:58:05.916
Yeah, yeah, definitely. Yeah, actually, we have a pretty nice experiment going on.

00:58:06.156 --> 00:58:09.476
I could show it to you right now. Do you have a spoon here?

00:58:14.536 --> 00:58:18.116
Plastic spoon? No. Anyway.

00:58:19.916 --> 00:58:26.116
No. Yeah, exactly. So the point is that from a geometrical perspective,

00:58:27.016 --> 00:58:32.156
if you forget this idea of forces and stuff, It's all about how the body is

00:58:32.156 --> 00:58:42.076
in track and how some measure of the mechanical state,

00:58:42.916 --> 00:58:50.536
of your body can inform you about the useful properties about what you touch.

00:58:51.016 --> 00:58:53.696
An example is actually elasticity.

00:58:54.496 --> 00:58:57.916
Now you have a hard surface and you have a soft surface here.

00:59:01.316 --> 00:59:05.316
So if you look at it from a theoretical mechanics viewpoint,

00:59:05.496 --> 00:59:07.636
the only way you can do it, if

00:59:07.636 --> 00:59:11.236
we had the force, you look at the relative deformation of the two bodies.

00:59:13.666 --> 00:59:17.166
So it's purely relative. It's only displacement. There is no load involved.

00:59:18.266 --> 00:59:22.066
And so you apply the laws of contact mechanics, and you realize,

00:59:22.326 --> 00:59:28.566
actually, that the information that is available to the brain is completely ambiguous.

00:59:33.126 --> 00:59:36.046
Confounded between shape and elasticity.

00:59:38.046 --> 00:59:42.786
And you never realize it in real life because you have a prior,

00:59:45.566 --> 00:59:49.486
assumption about the shape of what you touch, like an avocado when you test it.

00:59:50.006 --> 00:59:56.966
But if you organize a lab experiment where there is no such information available,

00:59:57.286 --> 01:00:00.966
you have a perfect confound between shape and elasticity.

01:00:01.406 --> 01:00:06.526
And that falls directly off the contact mechanics principles.

01:00:07.846 --> 01:00:09.186
No loads in.

01:00:11.806 --> 01:00:19.386
Involved there. Okay, I got it. But in some sense, we haven't made that much

01:00:19.386 --> 01:00:21.206
progress yet, right? Because now we're,

01:00:23.686 --> 01:00:28.366
still the periphery of the whole haptic interaction with the world.

01:00:28.966 --> 01:00:36.326
And there, you also alluded towards the end a little bit to how is the brain really sensing?

01:00:36.626 --> 01:00:40.406
How is the brain sensing these kinds of of properties of the world.

01:00:40.426 --> 01:00:42.786
So how many... How many properties?

01:00:42.966 --> 01:00:45.506
How many distinct sensors should we worry about?

01:00:45.986 --> 01:00:48.826
And what do they do, right?

01:00:50.546 --> 01:00:56.766
Oh, the sensors themselves? Yeah, as I said, you have a population of rapidly

01:00:56.766 --> 01:01:02.086
adaptive sensors, and then you have populations of slowly adaptive sensors.

01:01:03.166 --> 01:01:07.386
The way I think of it is that it also goes back to mechanics.

01:01:07.566 --> 01:01:11.026
You have to have So mechanics is mostly dynamics.

01:01:11.546 --> 01:01:15.426
The simplest dynamics are first order, or I mean second order,

01:01:15.466 --> 01:01:19.446
I mean. You need three terms in the equation. So you have two states.

01:01:20.406 --> 01:01:25.306
And so if you have to have a good access to the mechanical state of an object

01:01:25.306 --> 01:01:27.166
that has two states, you need two sensors.

01:01:27.646 --> 01:01:32.746
And that is why, and one is rate sensor and the other one is a static sensor.

01:01:33.246 --> 01:01:39.346
That's really fundamental. metal and and and then that's an explanation why

01:01:39.346 --> 01:01:44.966
you would have these two categories and these are these uh virginity cells uh

01:01:44.966 --> 01:01:49.886
or the patchini is now they are kind of a a world on their own,

01:01:50.466 --> 01:01:56.626
they are uh more in in tissues than in the skin there are some on the skin but

01:01:56.626 --> 01:02:02.366
mostly in the in the tissues uh sort of randomly opportunistically distributed

01:02:02.366 --> 01:02:06.106
in the hand in the entire body you You have lots of them in the mise-en-trie.

01:02:10.668 --> 01:02:14.908
So essentially, I think of them as systemic sensors.

01:02:15.168 --> 01:02:19.968
So they detect waves that are zipping by.

01:02:20.148 --> 01:02:23.688
And by the way, the waves in soft tissues are very slow.

01:02:24.688 --> 01:02:29.728
They move at seven meters per second and go over very long distances.

01:02:30.468 --> 01:02:40.468
And so that would be a very good portrayal of why these sensors are there.

01:02:40.668 --> 01:02:46.588
It's basically, they collect mechanical information propagating. Yeah.

01:02:47.208 --> 01:02:50.848
They are shear waves, actually, so they move slowly.

01:02:52.488 --> 01:02:58.788
We have new work, actually, not me, but a former collaborator of mine,

01:02:58.808 --> 01:03:05.668
who is a prof at UCSB, has made beautiful measurements of those waves in the anatomy.

01:03:07.748 --> 01:03:12.188
How far do they travel, these waves? Well,

01:03:12.468 --> 01:03:22.028
I could show you a movie, but I'll do a… Yeah, if you do this,

01:03:22.268 --> 01:03:26.148
they go very clearly all the way to the wrist.

01:03:26.448 --> 01:03:29.388
Okay, so from the fingertip to the wrist.

01:03:29.528 --> 01:03:33.988
Yeah, so if you put an accelerometer here, you get a very big signal here,

01:03:34.768 --> 01:03:36.248
and another one, and another one.

01:03:37.568 --> 01:03:40.788
And the wavelength is 3 centimeters.

01:03:42.068 --> 01:03:48.288
And the modes, actually, we've recently had a new result. There are very few modes.

01:03:48.588 --> 01:03:54.208
So it's a very compact representation, actually, of the dynamics of the hand.

01:03:54.988 --> 01:03:59.968
It would seem that, actually, you'd need a function basis of 8,

01:04:00.108 --> 01:04:05.928
a dimension 8, to completely represent the dynamics of the hand. That's cool.

01:04:05.988 --> 01:04:08.588
So actually for my wrist, I could sense what my fingers are doing.

01:04:08.968 --> 01:04:14.028
Yes. Well, as I said, it's a medical observation, actually.

01:04:15.468 --> 01:04:18.948
Do you think the brain is using that? I'm convinced it does. Okay.

01:04:20.068 --> 01:04:25.428
Actually, it was one of the projects that's ongoing is actually to have.

01:04:30.748 --> 01:04:35.948
Phantom locations, the dynamics. So I'll show you how you can elicit them.

01:04:36.348 --> 01:04:41.408
You do this, you slide your finger, and you record on the tabletop. It's good, actually.

01:04:42.128 --> 01:04:44.088
You record the signal on the nail here.

01:04:45.480 --> 01:04:50.660
Then you put it in a recording. And then as you slide your finger,

01:04:50.760 --> 01:04:53.180
and you put the transducer that you glue to that skin here.

01:04:53.380 --> 01:04:56.920
Okay. So if you hold your hand like this, you feel... Glue it onto the same

01:04:56.920 --> 01:04:59.180
finger? Same finger, but a different phallus.

01:04:59.420 --> 01:05:01.820
Yeah, close to the palm of the hand. Yeah, here, yeah.

01:05:02.120 --> 01:05:07.920
And you go like this. Oh, yeah, so it's here. You feel the texture going here.

01:05:08.940 --> 01:05:15.700
And if you glue it here, you feel it here. Now you organize the situation.

01:05:15.840 --> 01:05:21.020
Actually, when the signal starts going, the minute you touch and you slide on

01:05:21.020 --> 01:05:26.480
a flat surface, and then the sensation travels back to the fingertip,

01:05:27.100 --> 01:05:27.980
the textural sensation.

01:05:28.860 --> 01:05:33.840
Ah, okay. That's cool. Yeah, but it makes total sense.

01:05:34.280 --> 01:05:37.060
Could you put a transducer anywhere?

01:05:37.660 --> 01:05:41.820
Yes, I think so. So that's what we're going to do for that new project.

01:05:41.880 --> 01:05:43.640
Some mechanical transformation.

01:05:44.220 --> 01:05:47.740
Yeah, exactly. Oh, that's really cool. But now, okay.

01:05:47.840 --> 01:05:53.940
So that's why it goes back to what I was saying, that the neural architecture

01:05:53.940 --> 01:05:58.000
at the most basic level has to be very lateral because of that.

01:05:58.560 --> 01:06:02.080
That's what I meant by non-locality. But then also you're saying that haptic

01:06:02.080 --> 01:06:05.200
is much like audition because I must pick up also the resonances.

01:06:05.360 --> 01:06:08.720
Yes, yes. in my biomechanics.

01:06:08.900 --> 01:06:14.440
I mean, to put an extreme argument, which is wrong but interesting,

01:06:14.720 --> 01:06:19.580
it's like your whole body is like a basilar membrane. Right, exactly.

01:06:20.920 --> 01:06:24.220
Which is interesting, right? Because also the basilar membrane and audition,

01:06:24.520 --> 01:06:27.840
that's interesting features, like it extracts pitch or frequency.

01:06:28.200 --> 01:06:33.100
Yes, but as far as I can tell, no one has ever found any somatotopy,

01:06:33.200 --> 01:06:36.100
I mean, somatotopy in touch.

01:06:36.820 --> 01:06:42.300
Yeah, but pitch is also never really found, it's quite a discussion, but it's an invariant.

01:06:42.420 --> 01:06:45.320
You look at an invariant of the resonance frequency.

01:06:45.740 --> 01:06:49.920
Tautotopy is still a very important organizational principle in audition.

01:06:50.460 --> 01:06:54.160
Yeah, but I want to provoke you even further than that, because pitch is an

01:06:54.160 --> 01:06:56.520
invariant, a subjective invariant for here.

01:06:56.960 --> 01:07:02.120
If we now said the whole body is a metal membrane, what would be the equivalent

01:07:02.120 --> 01:07:04.680
of pitch for the haptic system?

01:07:06.087 --> 01:07:09.147
Would it be something like left hand is touching stuff? Yeah,

01:07:09.147 --> 01:07:12.867
you could say that. My foot, I can touch this end of my foot. Oh, it's the same thing.

01:07:14.327 --> 01:07:25.607
I would tend to think that the elements of sensation in touch are really like objects, are things,

01:07:26.427 --> 01:07:30.667
like pieces of wood or tables, people. That's what they are.

01:07:32.507 --> 01:07:38.227
A good argument. There's zillions of papers on roughness. Oh, yeah, right. Okay?

01:07:39.307 --> 01:07:43.507
But if you think of it a minute, if you take a piece of leather,

01:07:44.007 --> 01:07:47.827
it has a certain roughness, and you can rank leathers.

01:07:48.647 --> 01:07:54.327
Now you take, I don't know, concrete, okay? You have a different...

01:07:54.907 --> 01:08:00.467
But it's not the same roughness. So the notion of roughness depends on the object.

01:08:00.947 --> 01:08:06.627
And woods, you know, and people, Now you have smooth skins and rough skins.

01:08:07.087 --> 01:08:10.587
So it's not a universal concept like pitch.

01:08:11.947 --> 01:08:15.907
Which actually could argue is not universal. It really depends on the source, actually.

01:08:17.147 --> 01:08:22.187
And so you could make the argument that actually all roughnesses are actually

01:08:22.187 --> 01:08:23.767
qualities of materials.

01:08:24.307 --> 01:08:29.387
Well, that's more timber then. Yes. Yeah, but timber, that's also...

01:08:30.567 --> 01:08:37.667
I used to dab a little bit in psychophysics. And so the joke is like, what is timber?

01:08:37.887 --> 01:08:41.227
So you remove intensity, duration, pitch.

01:08:43.187 --> 01:08:45.907
And dynamics, and what is left is timber.

01:08:48.207 --> 01:08:53.787
But this is an interesting thought experiment to say, okay, if we move away

01:08:53.787 --> 01:08:58.367
from vision and think more about the dynamics of an auditory system,

01:08:58.827 --> 01:09:02.507
then now this starts to have interesting questions that we can pose for the

01:09:02.507 --> 01:09:04.667
haptic system that we hadn't thought of before.

01:09:04.867 --> 01:09:10.527
If we have purely spatial terms of the visual system, maybe that has been a misleading analogy.

01:09:12.107 --> 01:09:17.787
I'm completely in tune with that line. You resonate with that.

01:09:18.167 --> 01:09:22.807
Yeah. Touch can do it, you know, but pretty badly.

01:09:22.967 --> 01:09:29.887
You have all these illusions, and because it's certainly something you need,

01:09:30.067 --> 01:09:32.567
you know, when there's an insect, you have to whack it.

01:09:33.327 --> 01:09:37.107
But it's interesting what you're saying, right, because that would mean that

01:09:37.107 --> 01:09:41.027
also because but because of the constraints of experimentation,

01:09:41.747 --> 01:09:45.307
you end up in a very extreme part of the state space of haptics.

01:09:45.427 --> 01:09:48.947
Yeah. That sits very close to a vision issue. Yeah. But it might be really misleading.

01:09:49.227 --> 01:09:53.027
Very misleading, yeah. Yeah, you like to have two-dimensional things because

01:09:53.027 --> 01:09:56.107
it's easy to write papers about. Exactly.

01:09:56.907 --> 01:10:01.047
But then you also make a plug for the nucleus cuneus, right,

01:10:01.107 --> 01:10:06.067
which in some sense I could say you sort of co-discovered maybe with your talent.

01:10:06.067 --> 01:10:08.627
Well, it's, yeah. Big John Powell. Big John Powell, yeah.

01:10:09.667 --> 01:10:12.587
Because it seems to be such an essential organ, huh?

01:10:13.874 --> 01:10:19.674
But tell me, of the whole nucleus, that no one bothered about so far?

01:10:20.014 --> 01:10:26.254
Yeah, because the dogma, there is a dogma, which actually held that...

01:10:27.974 --> 01:10:34.974
Actually, I might offend quite a few people on this one, but the dogma is that,

01:10:35.934 --> 01:10:41.234
touch is very much explainable by label line theory.

01:10:43.874 --> 01:10:48.454
And that is actually written in the textbook that you have on your shelf.

01:10:50.674 --> 01:10:54.854
But the same holds for additionism, it's a dominant dogma. Yeah,

01:10:54.854 --> 01:10:56.374
exactly. But in touch, you see.

01:10:56.574 --> 01:11:04.074
So the consequence is that the physiological properties of receptors are actually

01:11:04.074 --> 01:11:07.094
reflected in the cortex,

01:11:07.094 --> 01:11:11.934
I mean,

01:11:13.334 --> 01:11:20.174
you can organize experiments where that is the case, but they are so contrived

01:11:20.174 --> 01:11:26.094
that they have no relation with normal functioning.

01:11:26.414 --> 01:11:29.834
There's a very good example, a second point also that I'd like to make since

01:11:29.834 --> 01:11:31.094
we're here on this topic.

01:11:32.894 --> 01:11:40.034
This label line thing idea in touch also would like to propose that the different

01:11:40.034 --> 01:11:44.414
type of receptors we have are frequency tuned from low frequency to high frequency

01:11:44.414 --> 01:11:48.834
which very much comes back to the auditory example,

01:11:49.634 --> 01:11:54.194
and so there's a paper that's quoted hundreds of times that shows that actually

01:11:54.194 --> 01:12:00.794
you do have populations of receptors that respond to certain frequency bands and that explains touch,

01:12:01.574 --> 01:12:06.054
But what people don't realize is that if you read the caption of the figure.

01:12:09.513 --> 01:12:19.933
The curves are plotted at threshold, which is like the smallest detectable stimulus.

01:12:20.313 --> 01:12:26.333
No, not the line at the threshold, which is precisely the domain where you don't

01:12:26.333 --> 01:12:30.073
use your sensors, because they are useless at threshold. I just want to kick

01:12:30.073 --> 01:12:31.153
in, right? Yeah, exactly. Exactly.

01:12:31.673 --> 01:12:36.873
And if you redraw this curve at the above threshold, you get something completely different.

01:12:37.073 --> 01:12:40.133
There's no tuning anymore. Yes.

01:12:41.673 --> 01:12:45.853
That's problematic, yes. So, yeah, all these ideas that, you know,

01:12:45.893 --> 01:12:49.253
you have this broad tuning. Neatly organized, yes.

01:12:49.533 --> 01:12:55.193
But then, how did you end up finding this nucleus cuneus?

01:12:55.753 --> 01:12:59.213
Because, um, uh, Henrik Jontel has, uh,

01:12:59.533 --> 01:13:03.533
um, uh, spent his career on, on this, his,

01:13:03.713 --> 01:13:07.533
uh, um, supervisor did, and, and I think the supervisor,

01:13:07.713 --> 01:13:14.353
uh, his supervisor also, so, so his lab has a profound knowledge of,

01:13:14.513 --> 01:13:18.993
of that organ and, and working with, uh, uh, uh, Henrik is, uh,

01:13:18.993 --> 01:13:25.353
an incredible pleasure because it's like his apartment, you know,

01:13:25.373 --> 01:13:28.673
he knows the place and where to put the needles.

01:13:28.873 --> 01:13:30.693
He knows. He's got an amazing physiology.

01:13:31.053 --> 01:13:37.373
Yeah. And so it goes very fast because of this, you know, half a century of

01:13:37.373 --> 01:13:38.433
accumulated knowledge.

01:13:39.873 --> 01:13:47.313
So and it's, though there's no other place I thought I can tell that can redo that fine,

01:13:48.553 --> 01:13:53.713
So what you're saying is, this nucleus cuneus is really like almost a vessel,

01:13:53.833 --> 01:13:55.933
a membrane or a retina. It's a metaphor.

01:13:57.253 --> 01:14:01.693
Yeah, it's a metaphor, but it's probably more true.

01:14:02.073 --> 01:14:05.813
It's really a neurophysiological fact.

01:14:06.313 --> 01:14:09.533
It's first order. How many neurons does it contain?

01:14:17.313 --> 01:14:17.893
Um...

01:14:22.093 --> 01:14:25.033
I don't know in i don't think anybody counted them for

01:14:25.033 --> 01:14:28.013
uh humans okay in cats it's

01:14:28.013 --> 01:14:36.253
by the millions okay and and and uh um connectivity is by the thousands okay

01:14:36.253 --> 01:14:40.933
so it's it's really a powerful uh and so you see the whole periphery being mapped

01:14:40.933 --> 01:14:45.353
to that to that sheet yeah and then there was a

01:14:45.473 --> 01:14:48.033
certain topography to it, right? It has certain domains.

01:14:49.013 --> 01:14:52.453
Or not. You can talk about six domains, if I'm correct.

01:14:53.793 --> 01:14:59.653
Yeah, well, you have body parts, like the pads and the regions.

01:15:00.653 --> 01:15:04.813
But the beauty of this architecture is that if you take one single of these...

01:15:05.793 --> 01:15:11.853
No, you take one particular region on the anatomy, then you will have hundreds,

01:15:11.973 --> 01:15:15.433
if not thousands of neurons scrutinizing the same place.

01:15:17.013 --> 01:15:21.553
So it's a stacked organization. See what I mean?

01:15:21.993 --> 01:15:31.613
It's like you have the receptive fields overlapped by hundreds or thousands.

01:15:32.013 --> 01:15:35.893
I would say in a human that's probably the case. You have a massive population

01:15:35.893 --> 01:15:39.333
response for a very local stimulus.

01:15:39.553 --> 01:15:42.733
And they're very large and then you have many stacked on top.

01:15:42.733 --> 01:15:44.613
And is the response temporarily structured?

01:15:45.733 --> 01:15:49.053
Highly, for sure. Yeah. Okay. For sure.

01:15:49.113 --> 01:15:55.193
And it does seem that actually the spiking train coming from the periphery,

01:15:56.233 --> 01:16:01.253
just the order of spikes is sorted out by that network. Yeah. Okay.

01:16:01.773 --> 01:16:06.893
Which would explain things like Flanagan and Johansson hypothesized.

01:16:06.893 --> 01:16:10.593
Like, I do this and I feel a contact because they all come at the same moment.

01:16:10.833 --> 01:16:16.773
Right, exactly. But so would you speculate that in the temporal structure of

01:16:16.773 --> 01:16:21.893
the population response, you are largely encoding whatever happens at the periphery?

01:16:23.759 --> 01:16:27.239
Yes. So it's not a landline, right? Yeah, exactly.

01:16:27.399 --> 01:16:34.739
The specificity of receptors is completely lost because these first-order neurons

01:16:34.739 --> 01:16:39.459
collect everything that comes from all the different types of subtypes,

01:16:39.499 --> 01:16:40.979
as called sub-modalities.

01:16:41.659 --> 01:16:44.619
Including all these details we discussed earlier. Yeah, exactly.

01:16:45.099 --> 01:16:48.059
So there is no segregation.

01:16:48.779 --> 01:16:53.679
What there is is segregation of inputs, not segregation of neural signals.

01:16:54.019 --> 01:16:57.719
Right. It's actually a conversion. So that's also the compression stage,

01:16:57.899 --> 01:16:59.659
right? Yeah. You're compressing a lot of stuff.

01:16:59.819 --> 01:17:03.839
It has to be. Yeah, it has to be like the retina. And it also suggests that

01:17:03.839 --> 01:17:08.159
then this population response is invariant to the actual location where the

01:17:08.159 --> 01:17:11.479
stimulus occurs, if the stimulus is driving this whole set of sensors.

01:17:11.639 --> 01:17:14.179
Yeah, exactly. In a comparable way. Yeah. So, yeah.

01:17:14.499 --> 01:17:18.639
And that's a consequence of the mechanics I was speaking of.

01:17:18.759 --> 01:17:22.519
It's beautiful. And that is why all animals have it, even a crocodile,

01:17:22.739 --> 01:17:28.139
it has all these scales and they have to be integrated into one single prey.

01:17:28.959 --> 01:17:34.259
So would you say that's the most primitive representation of the periphery of the body as the nucleus?

01:17:34.419 --> 01:17:42.099
Yeah, I would say the body is not represented, what is represented is the object that can.

01:17:44.122 --> 01:17:47.042
Right, it's the interaction of the body with the outside. It's really the possible

01:17:47.042 --> 01:17:48.862
interactions, yeah. Okay, I understand.

01:17:50.062 --> 01:17:54.742
Like sharp things or small things. Including with itself, the body touches itself. Yeah, for sure.

01:17:55.162 --> 01:18:00.242
Which actually I can demonstrate right now. There's beautiful illusions about

01:18:00.242 --> 01:18:02.862
this. Okay, so you give me this hand here.

01:18:03.582 --> 01:18:08.882
And you put it in this way. I do this. Now you take the other hand. You do a pinch.

01:18:09.302 --> 01:18:12.682
And then you touch here. And then you rub.

01:18:15.442 --> 01:18:16.722
Yeah, exactly.

01:18:19.362 --> 01:18:23.602
So, that particular input is very unusual. You never had it before.

01:18:24.182 --> 01:18:31.522
And you probably caused it. I mean, it made you illuminate your face immediately.

01:18:33.642 --> 01:18:38.782
So, face illumination is a surprise. It's something that you never knew before.

01:18:40.162 --> 01:18:49.062
Of course, the tactile input was completely normal, but that's something abnormal,

01:18:49.402 --> 01:18:51.582
actually, in its structure.

01:18:53.062 --> 01:19:00.882
What's the future of this research on the desnucleus cuneus? I don't know.

01:19:00.982 --> 01:19:06.022
We applied to different programs unsuccessfully.

01:19:06.422 --> 01:19:17.682
And for now, there are some groups of neuroscientists in the US that have started

01:19:17.682 --> 01:19:21.142
putting multiple electrodes.

01:19:21.142 --> 01:19:23.182
Truth. So that's interesting.

01:19:23.842 --> 01:19:28.402
But you don't get the same information as the single cell.

01:19:29.822 --> 01:19:36.042
And what Henrik can do is a patch clamp, which is even more fine and informative.

01:19:38.802 --> 01:19:42.202
Okay, so it's unclear how much more we're going to learn about that structure

01:19:42.202 --> 01:19:43.322
in the short term. Before, yeah.

01:19:44.402 --> 01:19:47.002
It would be very... But,

01:19:49.389 --> 01:19:55.649
Maybe it could be resuscitated if we could make a good argument that it has health consequences,

01:19:58.429 --> 01:20:04.389
diabetes, or I don't know, something like neuropathies. It seemed to have,

01:20:05.069 --> 01:20:14.069
it seemed like Henrik was talking about it's a possible link to Parkinson's,

01:20:14.069 --> 01:20:18.209
which is not implausible because it's a direct,

01:20:19.949 --> 01:20:22.669
provider of information to the basal ganglia.

01:20:23.989 --> 01:20:25.129
Really direct.

01:20:28.829 --> 01:20:36.489
So now philosophers over the last decades are more and more enthusiastic about

01:20:36.489 --> 01:20:39.789
sort of embodied perspectives on everything.

01:20:40.529 --> 01:20:48.369
And you read these really beautiful stories about it But in these analyses,

01:20:48.549 --> 01:20:53.789
you often get the feeling that the body is taken in a very literal sense.

01:20:54.109 --> 01:20:58.369
The body is basically bounded by the skin.

01:21:00.169 --> 01:21:06.349
So Madan, your research, where you look really at the body as defined through

01:21:06.349 --> 01:21:12.729
the sensors that are at this periphery. Right, so how strictly would you draw

01:21:12.729 --> 01:21:14.789
this boundary of the body?

01:21:15.069 --> 01:21:20.449
Well, it does seem actually that this mental picture we have of a boundary,

01:21:21.049 --> 01:21:29.549
you know, from, if you want, our abstract understanding of physical object,

01:21:29.809 --> 01:21:35.289
doesn't have much of an equivalent in cognition, actually.

01:21:36.589 --> 01:21:41.469
With another former postdoc, we have this really interesting experiment going

01:21:41.469 --> 01:21:45.129
about counting the number of sides of the skin.

01:21:46.529 --> 01:21:50.789
So the skin normally should have two, right? In and out. Well,

01:21:50.809 --> 01:21:54.249
it turns out that actually the number of sides of the skin is smaller than two.

01:21:55.949 --> 01:22:01.389
What does that mean? Well, I'll give you the phenomenology, actually.

01:22:01.389 --> 01:22:07.949
If you have an object that's rotating on the skin, it has an orientation,

01:22:08.229 --> 01:22:10.389
and that's unique to the side.

01:22:10.529 --> 01:22:13.849
So if you look at the skin from this side, it has a certain orientation.

01:22:14.269 --> 01:22:17.809
If you look at it from the other side, the same stimulus will have another orientation.

01:22:18.409 --> 01:22:27.049
Well, it does seem that actually, if you turn your hand, then the skin has only one side. Yeah.

01:22:31.565 --> 01:22:34.605
So, then it has some sort of more global reference. Exactly,

01:22:34.765 --> 01:22:36.265
yeah. What's the global reference?

01:22:37.505 --> 01:22:41.025
Well, good, yeah. So, there are several hypotheses.

01:22:42.165 --> 01:22:45.865
So, we have a north and a south, do we? Yeah, it could be visual.

01:22:46.085 --> 01:22:49.685
It could be the visual world. That's one possibility.

01:22:50.725 --> 01:23:00.425
It could be completely object-specified, which I think is probably a good way

01:23:00.425 --> 01:23:05.065
to think about it really has nothing to do with the skin it's about the type

01:23:05.065 --> 01:23:06.325
of object that can create,

01:23:07.105 --> 01:23:17.905
these symptoms the periphery of the body on its own is defined through this

01:23:17.905 --> 01:23:19.085
whole conglomerate of sensors,

01:23:19.705 --> 01:23:23.005
and then resonances that then are talking

01:23:23.005 --> 01:23:28.805
to the nucleus cuneus and whatever I

01:23:28.805 --> 01:23:32.525
do that is sort of system systematically driving those

01:23:32.525 --> 01:23:36.165
sensors the body seems to redefine

01:23:36.165 --> 01:23:39.185
itself as just incorporating in what then this physical

01:23:39.185 --> 01:23:41.905
self is yeah yeah right so if i take a

01:23:41.905 --> 01:23:45.565
stick and as you describe it things with that stick it

01:23:45.565 --> 01:23:48.665
will just drive the same resonance system for sure to described earlier

01:23:48.665 --> 01:23:53.085
yeah and then the curious and everything that follows says okay that's me yeah

01:23:53.085 --> 01:23:57.385
physically this yeah that's me yeah yeah so so so physically but is this physical

01:23:57.385 --> 01:24:02.945
me as you showed in your more recent experiment we will publish soon you showed

01:24:02.945 --> 01:24:06.765
that people can have very accurate predictions about what the physical configuration

01:24:06.765 --> 01:24:08.605
is of this extension of this.

01:24:09.285 --> 01:24:15.445
But now can we push it one step further and say I can also consider physical

01:24:15.445 --> 01:24:17.165
me as being discontinuous.

01:24:18.645 --> 01:24:24.185
Could I have discontinuous in the mechanical sense like having two as long as

01:24:24.185 --> 01:24:29.825
I drive the sensory apparatus in some sense and all this multi-model I believe

01:24:29.825 --> 01:24:33.665
that's possible yeah we actually do have a,

01:24:35.345 --> 01:24:38.705
set of experiments going in that direction where you.

01:24:41.045 --> 01:24:49.425
So this is more really early but we are funded by a virtual reality company

01:24:49.425 --> 01:24:54.105
to look at that which is actually to um,

01:24:57.045 --> 01:24:57.525
uh, uh,

01:25:00.237 --> 01:25:06.517
illicit sensations that are owned without any mechanical connection.

01:25:08.477 --> 01:25:12.097
The connection there is gaze.

01:25:14.557 --> 01:25:20.037
We've shown that also at the task level. Yeah, and it seems to be pretty good, actually.

01:25:22.957 --> 01:25:26.377
It's really dramatic. Yes, it is. more

01:25:26.377 --> 01:25:31.777
dramatic than you can imagine because we built a whole new house of cards of

01:25:31.777 --> 01:25:36.017
embodied cognition yeah where we say oh we're going to ground knowledge we're

01:25:36.017 --> 01:25:39.657
going to solve the symbol grounding problem and all that stuff because there's

01:25:39.657 --> 01:25:44.157
a body yeah but the body is embodied but now thanks to you,

01:25:44.857 --> 01:25:48.997
it's the conclusion but the body is a construct it's not given,

01:25:50.237 --> 01:25:55.657
that's that's yeah yeah that's what you could conclude actually Actually,

01:25:55.857 --> 01:25:57.817
I have another contour example that's pretty good.

01:25:57.917 --> 01:26:00.697
I have to finish the paper with Jess about this one.

01:26:00.997 --> 01:26:03.777
You have an object and you ask how big it is.

01:26:04.737 --> 01:26:09.817
So you have two bigger, smaller, bigger. And then you put it here on the left

01:26:09.817 --> 01:26:11.797
side with the other hand.

01:26:11.897 --> 01:26:14.377
And then it feels a little smaller.

01:26:16.177 --> 01:26:21.717
So that's nothing surprising. You know, the one hand has a different calibration, if you want. Yeah.

01:26:22.734 --> 01:26:26.994
And then we realized, actually, that's not true. It's not the hand that matters.

01:26:27.054 --> 01:26:29.954
It's the hemispace. Ah, cool.

01:26:32.194 --> 01:26:37.174
So you have exactly the same bias. If you put the object here and you use the

01:26:37.174 --> 01:26:39.794
right hand or the left hand, it doesn't matter. It's because it's there.

01:26:39.974 --> 01:26:44.674
Then you have the haptic equivalent of the Snark effect. Yes.

01:26:45.594 --> 01:26:49.314
Right? But you also know there's a bias in the magnitude.

01:26:50.614 --> 01:26:53.934
Right, exactly. Dependent on which hands you're using to make the choice.

01:26:54.334 --> 01:26:56.474
Yeah. So, would you buy that?

01:26:57.374 --> 01:27:03.994
Generalization of the Stark effect? You could, yeah. It definitely goes into the discussion of that.

01:27:04.894 --> 01:27:10.614
But what I was pointing at is that actually, yeah, the tactile experience of

01:27:10.614 --> 01:27:13.874
the object is really related to where it is. And constructed. Yeah.

01:27:15.034 --> 01:27:22.054
That's fantastic. So, now you have demolished embodied cognition as well. Good.

01:27:25.934 --> 01:27:30.854
So the other thing that you do, you also really worry about applying the science,

01:27:30.994 --> 01:27:32.534
right? To build applications with it.

01:27:34.194 --> 01:27:37.474
Yes. I've been doing this for a long time.

01:27:37.574 --> 01:27:42.514
And in some sense also it's building the devices and the applications that is

01:27:42.514 --> 01:27:44.994
also, I think, informing and feeding back into the science.

01:27:45.654 --> 01:27:49.474
Yes. Yeah, for sure. That's a way of life.

01:27:50.174 --> 01:27:56.374
Actually, this button here, like you have on iPhone.

01:27:57.474 --> 01:28:02.794
Yeah, the Home button. Yeah. You know, it's actually a haptic effect. Sure.

01:28:04.194 --> 01:28:10.474
I did publish that many, many years ago. And actually, Apple was very nice to me.

01:28:12.570 --> 01:28:18.450
Invited me as a in a vip situation and it was great you got a t-shirt i got

01:28:18.450 --> 01:28:21.450
a t-shirt no i got a nice one.

01:28:25.390 --> 01:28:30.590
Well what can you do they're a trillion dollar company and they got a t-shirt

01:28:30.590 --> 01:28:35.850
they were they were quite you know uh they called me ahead of time like two

01:28:35.850 --> 01:28:39.090
years before the product came Oh, that's nice. It was nice, yeah. That's good.

01:28:39.510 --> 01:28:41.890
No, I think it's a fair company.

01:28:42.410 --> 01:28:44.150
Okay. Yeah, I'm trying. I agree.

01:28:44.750 --> 01:28:47.430
But I think from an ethics perspective, it's not a discussion.

01:28:47.770 --> 01:28:50.810
It's very difficult to speak of a fair company. Yeah.

01:28:51.910 --> 01:28:54.790
Okay, so you do build these applications.

01:28:55.330 --> 01:28:58.970
Where do you see this go? Where do you see the impact that these haptic systems

01:28:58.970 --> 01:29:01.190
are going to have in the future?

01:29:01.650 --> 01:29:06.810
How will they change our lives? Well, so the… The home button is gone now.

01:29:06.850 --> 01:29:08.510
I have an iPhone 10, no more home button.

01:29:08.810 --> 01:29:12.370
It's still vibrating once in a while, but the home button is not there.

01:29:13.830 --> 01:29:22.770
So that's a functional, yeah, yeah. But the home button now will show up on gas stoves.

01:29:24.330 --> 01:29:28.170
So that's when... So what's the real revolution?

01:29:28.450 --> 01:29:32.450
How is it going to change the way we deal with the world and each other?

01:29:32.450 --> 01:29:42.790
Well the movie theater took half a century to to really I mean it was a curiosity

01:29:42.790 --> 01:29:47.330
for a long time and it it's only after the war that it became an important,

01:29:48.050 --> 01:29:57.970
I mean no it did but it really exploded and so you say it will take a long time

01:29:57.970 --> 01:30:01.930
yeah before you have yeah so you have not a specific development base,

01:30:03.150 --> 01:30:05.930
to create that revolution, the haptics revolution?

01:30:08.850 --> 01:30:10.830
Well, revolutions are never predicted.

01:30:12.830 --> 01:30:18.050
I know you're a little embarrassed. But still, you can try.

01:30:18.390 --> 01:30:24.070
So we're trying. But mundane applications, for sure, is very important.

01:30:24.210 --> 01:30:28.030
They will come in massive.

01:30:28.850 --> 01:30:31.470
So it's not going to be the holodeck you

01:30:31.470 --> 01:30:34.870
know but mundane okay okay i

01:30:34.870 --> 01:30:38.290
understand that's cool that that's also a roadmap yeah

01:30:38.290 --> 01:30:42.670
simple simple yeah get into the real world yeah you'll also see how people will

01:30:42.670 --> 01:30:48.670
use it yeah so we made quite a tour here sort of haptics world which is extremely

01:30:48.670 --> 01:30:54.310
uh mysterious interesting and exciting and you've been hammering away at this

01:30:54.310 --> 01:30:56.230
for quite a while now very systematically,

01:30:57.110 --> 01:31:00.470
Are you doing this? So, my...

01:31:04.415 --> 01:31:08.235
Original career goal was robotics, as you know.

01:31:08.835 --> 01:31:18.315
And I was involved in almost the first force control robot way back at Purdue University.

01:31:19.915 --> 01:31:23.275
What was the load? It was really force.

01:31:23.515 --> 01:31:28.355
So essentially, you had strain gauges in the harmonic drives,

01:31:28.515 --> 01:31:30.935
actually. It was a really nice robot.

01:31:31.675 --> 01:31:34.155
All the ones I see now are actually worse.

01:31:35.715 --> 01:31:38.815
I could show you a movie. It's a beautiful way. I see that robot.

01:31:38.855 --> 01:31:44.155
It actually, you know, it tracks with, uh, uh, hard surfaces,

01:31:44.355 --> 01:31:45.615
no problem, no stability.

01:31:45.775 --> 01:31:49.395
It wasn't a, anyway, it worked so well. I thought it was not a problem,

01:31:49.535 --> 01:31:50.955
you know, like not interesting.

01:31:55.515 --> 01:32:00.075
And, and then I drifted. Uh, and then what happens is that I got,

01:32:00.075 --> 01:32:03.835
um, a bit frustrated by the rate of progress in robotics.

01:32:04.015 --> 01:32:07.555
Uh, I was doing like straight manipulation, a lot of software,

01:32:07.795 --> 01:32:13.635
mechanical engineering, new structures, ultralight arms and stuff like that.

01:32:14.595 --> 01:32:23.275
And then one day, a colleague of mine came to my office from the Research Institute

01:32:23.275 --> 01:32:25.955
in Quebec and said, oh, I have a great idea.

01:32:27.623 --> 01:32:30.603
I'm working on the... There's a problem.

01:32:32.843 --> 01:32:42.583
The blind community uses Braille mostly for literacy and also for work.

01:32:42.863 --> 01:32:51.743
And so far, the computer works were motivated by the fact that you had line

01:32:51.743 --> 01:32:54.103
commands like Unix and DOS.

01:32:54.103 --> 01:33:00.463
And so you had quite a lot of blind folks who would actually be system managers

01:33:00.463 --> 01:33:05.683
and programmers because, you know, essentially lines is what you can do. It was.

01:33:06.103 --> 01:33:08.423
And then came Windows and then boom.

01:33:09.243 --> 01:33:14.723
Flickering everything. Yeah. It was really not wanted but it was so I said I

01:33:14.723 --> 01:33:16.023
have an idea. Perfect, yeah.

01:33:17.303 --> 01:33:23.763
Let's turn the window screen into something you can touch. And so we went to see a small.

01:33:26.383 --> 01:33:29.283
Accessibility company in Quebec, which still exists actually.

01:33:29.563 --> 01:33:31.683
It's one of the biggest in the world now.

01:33:32.503 --> 01:33:40.443
It was proposed to make a device that would let blind users feel the screen.

01:33:43.643 --> 01:33:49.783
I remember they gave us $5,000 to do that.

01:33:51.183 --> 01:33:54.463
I milked some money from the space agency and and

01:33:54.463 --> 01:33:57.443
then uh and then we actually made the stuff

01:33:57.443 --> 01:34:01.703
it worked beautifully was a small robot that big and then you would touch it

01:34:01.703 --> 01:34:08.903
and um and then you would actually uh feel the all the icons and uh you know

01:34:08.903 --> 01:34:16.943
the desktop very nicely and uh and and then i i thought i was really interesting.

01:34:17.943 --> 01:34:24.223
So I could use my engineering acumen for making it work well.

01:34:24.883 --> 01:34:32.623
And then another guy called Bill Buxton, he was a lot of fun and he wanted one.

01:34:34.303 --> 01:34:40.723
And then after that I never went back, basically. How long ago was that? That was actually in 91.

01:34:42.203 --> 01:34:48.543
How did you get rid of the Windows paperclip and the paperclip troll that then would help you?

01:34:53.063 --> 01:34:57.003
But it was really interesting, like an anecdote, but many I think those are interesting.

01:34:57.203 --> 01:35:02.083
When you gave that device to the blind users, so in your visual desktop,

01:35:02.183 --> 01:35:06.503
you would align the icons like in columns, like in a grid.

01:35:06.503 --> 01:35:13.423
But the blind users that they would feel the icons and push them around they

01:35:13.423 --> 01:35:17.683
would have you know a tactile sensation to them and they would be radiocentric and they.

01:35:18.999 --> 01:35:22.959
So the most important would be in the middle, and the less important would be on the side.

01:35:23.159 --> 01:35:28.539
So it was very different from a visual world. It was really a haptic world.

01:35:29.719 --> 01:35:35.839
And then there's a lot actually of… Well, haptic is a more serial or linear search, right?

01:35:35.899 --> 01:35:39.139
It's not linear, it's very… Or a random access, a random access to vision.

01:35:39.859 --> 01:35:42.039
No, I think that's wrong.

01:35:43.119 --> 01:35:46.919
It's not exactly a correct... Oh, you can tell me. Yeah.

01:35:47.419 --> 01:35:53.539
Vision is also serial because you saccade and search, but it does it at essentially,

01:35:54.199 --> 01:35:56.939
300 faster than the hand. No, no way.

01:35:57.579 --> 01:36:03.379
Okay, but I can... So you can actually scan... I can go rapidly and also change direction.

01:36:03.559 --> 01:36:07.599
I've read Russia in my... Exactly. Yeah, yeah. So the dynamics of vision are

01:36:07.599 --> 01:36:10.359
so fast that... They compare me with the time consensors.

01:36:10.579 --> 01:36:17.499
And in fact, you can do that in the lab. If you slow down a vision like touch, you get the similar.

01:36:17.959 --> 01:36:23.359
In my defense, I could then say a vision could approximate random action. Yeah. Yeah.

01:36:24.399 --> 01:36:31.559
But it's really a matter of steel. The question is that this is 35 years in haptics. Yeah.

01:36:32.839 --> 01:36:37.979
And for some reason the field is still sort of in its beginnings. There's a lot to be done.

01:36:38.279 --> 01:36:41.559
So what is Vincent's law if we want to make some progress here?

01:36:41.659 --> 01:36:43.519
What law should we follow? What's Vincent's law?

01:36:43.799 --> 01:36:49.839
Oh, in the engineering side?

01:36:50.119 --> 01:36:56.699
In understanding engineering, you choose to have impact in this domain.

01:36:58.339 --> 01:37:06.379
Well, that's pretty tough. Yeah, do things that work.

01:37:09.679 --> 01:37:14.439
You never know. How does it? Overselling is a big problem. Okay.

01:37:15.239 --> 01:37:22.339
Like in robotics, there's a lot of that, a lot of claims and very little delivery.

01:37:22.899 --> 01:37:32.339
So get real. So the overselling is very dangerous because it makes people tired or bias negatively.

01:37:34.099 --> 01:37:38.359
So that's a piece of advice. Don't oversell. Be, you know.

01:37:39.279 --> 01:37:43.859
Be real. Stick to the facts. yeah stick to the fact yeah that's a good point

01:37:43.859 --> 01:37:48.839
and then okay if i'm gonna go up to paris and um visit you four years from now

01:37:48.839 --> 01:37:51.939
what's the one prediction you would like to see,

01:37:52.879 --> 01:37:56.579
thoroughly tested in that time frame that's essential to your program.

01:38:00.932 --> 01:38:10.792
So I have, yeah, I think it will be possible to reduce the, yeah,

01:38:10.852 --> 01:38:11.932
you were asking about principles,

01:38:12.212 --> 01:38:23.152
the, at least a certain aspect of principle to a countable amount of invariance.

01:38:25.152 --> 01:38:32.272
And they could be biomechanical they could be physiological or they could be

01:38:32.272 --> 01:38:34.492
mechanical or drawn from the laws

01:38:34.492 --> 01:38:38.872
of physics but I think there is not an infinite number of them there is,

01:38:40.332 --> 01:38:46.032
not more than the one in vision they have been well researched so that's a prediction

01:38:46.032 --> 01:38:53.212
I could make that you could actually put them down in a you know, a discrete matter.

01:38:53.652 --> 01:38:57.432
You're just detecting what the invariances are. Well, I know quite a few,

01:38:57.572 --> 01:39:02.532
like the one I showed about the curvature, so that's a good one.

01:39:03.852 --> 01:39:06.432
But there are... So you're saying

01:39:06.432 --> 01:39:10.472
there's a handful? There will be like a handful of invariants. Yeah.

01:39:12.292 --> 01:39:16.792
Very good. That's a telly word. Thank you very much for this conversation. Pleasure.

01:39:21.032 --> 01:39:26.872
The CSN podcast was produced by the Convergent Science Network of Biometrics

01:39:26.872 --> 01:39:33.612
and Biohybrid Systems, a project funded by the European Sevens Research Framework Programme.

01:39:34.832 --> 01:39:40.152
For more interviews, recorded lectures or upcoming conferences in the field

01:39:40.152 --> 01:39:46.392
of biometrics and biohybrid systems, go to csnnetwork.eu.

01:39:46.320 --> 01:39:54.320
Music.

01:39:46.692 --> 01:39:48.552
And thank you for listening.

01:39:50.552 --> 01:39:52.212
Bye.