WEBVTT 00:00:03.597 --> 00:00:10.517 This is the Convergent Science Network podcast. Leading researchers in the domain 00:00:10.517 --> 00:00:16.797 of neuroscience, brain theory and technology are interviewed by Paul Verschure and Tony Prescott. 00:00:20.037 --> 00:00:23.377 This is Paul Verschure with the Convergent Science Network podcast. 00:00:24.357 --> 00:00:29.257 And in this episode, I'm speaking with Giovanni Pizzulo, who is a speaker at 00:00:29.257 --> 00:00:33.697 our summer school, Barcelona Cognition Brain and Technology summer school. 00:00:34.197 --> 00:00:42.217 And Giovanni, you started out with investigating this whole issue of the predictive brain. 00:00:42.697 --> 00:00:45.437 Yes. So what does it mean for your predictive brain exactly? 00:00:46.537 --> 00:00:51.157 Well, in a sense, the predictive brain is just a big idea. 00:00:51.317 --> 00:00:54.917 Well, the idea is that the brain is not a passive organ. 00:00:54.917 --> 00:00:58.277 So rather than just sitting down and waiting for the next stimulus, 00:00:58.457 --> 00:01:02.717 in a sense, it really always tries to anticipate what comes next, 00:01:02.837 --> 00:01:07.477 to set up some internal goals, and so to have some big internal processing. 00:01:07.637 --> 00:01:12.397 So rather than being stimulus-based, it's really trying to anticipate the necessities 00:01:12.397 --> 00:01:14.997 or maybe the opportunities for action. 00:01:14.997 --> 00:01:19.357 So that's really the big principle that drives all the brain processing because 00:01:19.357 --> 00:01:23.897 now, because the brain has some predictions, some goals, it can really organize 00:01:23.897 --> 00:01:26.317 its sensory processing, its attention processes, 00:01:26.477 --> 00:01:33.417 and its motor preparation processes for quickly responding and doing some adaptive actions. 00:01:33.617 --> 00:01:36.837 So to do that, it has to be predictive and proactive. Mm-hmm. 00:01:37.902 --> 00:01:42.202 But then where does that really start? What are the good examples of this kind of prediction? 00:01:43.322 --> 00:01:49.762 Well, in the literature, there are many, many partially disconnected literature on prediction. 00:01:50.002 --> 00:01:53.702 So predictions in sensory processing, predictions in the motor domain, 00:01:54.002 --> 00:01:56.842 predictions also in higher cognition, in that, for instance, 00:01:56.942 --> 00:02:00.142 I can try to anticipate what your next question will be. 00:02:00.242 --> 00:02:05.362 So all these literatures can now proceed a little bit disconnected. 00:02:05.362 --> 00:02:11.382 But for instance, in the sensory domain, you see a lot of predictive stuff going 00:02:11.382 --> 00:02:14.442 on, such as the anticipation of the next stimulus. 00:02:15.122 --> 00:02:21.002 In the motor control domain, you really have to get rid of what are the consequences 00:02:21.002 --> 00:02:23.862 of your action for many reasons. 00:02:23.862 --> 00:02:28.202 One big reason in the motor domain is that sensory perceptions are very ambiguous, 00:02:28.402 --> 00:02:32.802 so by also predicting what comes next, you better estimate the state of the world. 00:02:33.402 --> 00:02:38.222 Another big reason is for decision-making. So if you can really anticipate what 00:02:38.222 --> 00:02:39.522 the effects of your action are, 00:02:39.742 --> 00:02:43.162 then you can also select them on different courses of actions beforehand. 00:02:43.162 --> 00:02:47.362 Beforehand, such as, for instance, if I anticipate that by going left, 00:02:47.602 --> 00:02:53.582 I will get some big reward, and by going right, I will get some big punishment, 00:02:53.682 --> 00:02:55.162 then that's the decision. 00:02:55.922 --> 00:03:00.862 But now, in some sense, let's say also the physiological study of learning that 00:03:00.862 --> 00:03:03.722 started with Pavlov, one of 00:03:03.722 --> 00:03:07.442 the first observations of Pavlov was that the brain is predicting, right? 00:03:07.502 --> 00:03:11.862 So that's also a key feature of classical conditioning. So in some sense, 00:03:11.982 --> 00:03:15.122 the concept of prediction is with us for a long time. 00:03:15.242 --> 00:03:19.782 And as some speakers have said, like sometimes certain things get so old that 00:03:19.782 --> 00:03:20.842 they sound new again, right? 00:03:20.922 --> 00:03:27.862 So what is actually really new in this current movement of the brain as a predictor? 00:03:28.722 --> 00:03:32.902 Well, actually, I completely agree that also in classical conditioning, 00:03:33.002 --> 00:03:35.542 you have this predicting dynamics going on. 00:03:35.702 --> 00:03:41.582 But I think that we should try to really distinguish at least two kinds of predictions. 00:03:43.097 --> 00:03:45.737 It's not a sharp distinction, but it's useful to do probably. 00:03:45.917 --> 00:03:50.097 So one is an implicit kind of prediction in that, for instance, 00:03:50.257 --> 00:03:51.937 in the conditioning studies. 00:03:52.217 --> 00:03:57.377 So a stimulus, which is typically highly predictive of another stimulus, 00:03:57.597 --> 00:03:59.717 which is in turn highly predictive of reward. 00:04:00.157 --> 00:04:03.057 So the first stimulus, which is predictive of another stimulus, 00:04:03.257 --> 00:04:06.037 it becomes itself good to achieve. 00:04:06.037 --> 00:04:10.157 So in that case the brain becomes implicitly 00:04:10.157 --> 00:04:13.477 predictive in that it stores the relationships between 00:04:13.477 --> 00:04:16.737 the first and the second stimulus but to do 00:04:16.737 --> 00:04:19.737 that you don't really need to to maintain into 00:04:19.737 --> 00:04:22.737 your memory or to an internal representation of the 00:04:22.737 --> 00:04:29.637 predictive relation between the two you just attach some good label to the to 00:04:29.637 --> 00:04:34.837 the first stimulus whereas there is a second kind of prediction which is a more 00:04:34.837 --> 00:04:40.437 explicit prediction in which you keep into your brain some model of the environment. 00:04:40.877 --> 00:04:46.197 So the point is that in the first case you really have some implicit mechanisms 00:04:46.197 --> 00:04:51.177 that link good or bad to sequences of things, whereas in the second mechanism 00:04:51.177 --> 00:04:54.797 you also maintain a model of these sequences of things. 00:04:54.937 --> 00:05:02.457 For instance in the motor domain you can, for instance, let's say you have to grasp a moving ball. 00:05:02.917 --> 00:05:07.337 There are two different ways to do that. So one way is to look at the moving 00:05:07.337 --> 00:05:10.457 ball and then jump to some future state. 00:05:11.544 --> 00:05:15.844 To some future state that the ball will reach at some point. 00:05:16.404 --> 00:05:21.304 This can be done in two ways. The first way is just learning the contingencies automatically. 00:05:21.804 --> 00:05:29.104 So I know that if I see the ball moving, then I have to go to some position, 00:05:29.424 --> 00:05:31.864 which is not the current position, but the future position. 00:05:32.604 --> 00:05:36.504 Without even anticipating that position explicitly, you can do that automatically. 00:05:36.504 --> 00:05:43.864 The second way to do that is really doing what is called a forward model, 00:05:44.044 --> 00:05:47.704 so a predictive model that really tells you where the ball will be in the future. 00:05:48.024 --> 00:05:52.464 Use this forward model online, so this predictive mechanism online, 00:05:52.744 --> 00:05:56.704 and then move the hand to the predicted position. 00:05:58.364 --> 00:06:03.004 So to recapitulate this idea, there are implicit and explicit mechanisms, 00:06:03.124 --> 00:06:06.984 and the novelty of or the predictive brain hypothesis, if you want, 00:06:07.104 --> 00:06:10.544 is the hypothesis that the brain systematically incorporates, 00:06:12.024 --> 00:06:17.824 regularities of the external environment in a structured way so as to form strong 00:06:17.824 --> 00:06:22.124 models, statistical models, for instance, of the regularities of the external environment. 00:06:22.544 --> 00:06:28.324 And then it systematically uses these models to plan what to do next and also 00:06:28.324 --> 00:06:31.384 to drive perception, to drive attention, and so on. 00:06:31.384 --> 00:06:34.724 So, whereas in the first formulation of prediction, 00:06:35.024 --> 00:06:39.364 it was more an implicit prediction, in the predictive brain hypothesis, 00:06:39.584 --> 00:06:44.124 at least in some of its formulation, it's much more about building models for 00:06:44.124 --> 00:06:48.564 the world, on top of which you can run explicit predictions. Right, so then…. 00:06:49.507 --> 00:06:53.127 If you would look at, let's say, the theoretical literature on the predictive 00:06:53.127 --> 00:06:58.947 brain, what in your mind are right now the outstanding examples of this approach? 00:06:59.567 --> 00:07:03.307 Well, there are many of them. 00:07:03.567 --> 00:07:11.987 So probably in a very famous paper by Rao and Ballard, it was already in the 00:07:11.987 --> 00:07:16.267 99, if I remember well, it was about this predictive coding idea, 00:07:16.507 --> 00:07:18.287 predictive coding idea in perception. 00:07:18.287 --> 00:07:22.687 So the idea is that the brain continuously generates prediction about the stimulus, 00:07:22.907 --> 00:07:29.967 and it uses these predictions in a top-down manner, whereas it uses prediction errors, 00:07:30.247 --> 00:07:33.607 so the difference between the prediction and the sensory evidence, 00:07:33.867 --> 00:07:36.227 as a revision mechanism. 00:07:36.227 --> 00:07:39.747 So that was a milestone in a sense. 00:07:39.887 --> 00:07:44.207 But, well, there are many other papers now. There is a very big framework put 00:07:44.207 --> 00:07:48.787 forward by Carl Freestone, which is called the Free Energy Framework or the 00:07:48.787 --> 00:07:50.047 Active Inference Framework. 00:07:50.127 --> 00:07:53.347 In that sense, Freestone really tries to look at the whole brain, 00:07:53.447 --> 00:07:57.187 not just the perceptual system, as a big prediction machine. 00:07:59.087 --> 00:08:03.927 There are many other frameworks. One is put forward by Moshibar, 00:08:04.047 --> 00:08:05.387 again, in the perceptual domain, 00:08:05.627 --> 00:08:09.987 whereas in the motor domain, also the leading view, one of the most authoritative 00:08:09.987 --> 00:08:15.847 view put forward by people such as Danny Wolpert or by Shadmer or many other 00:08:15.847 --> 00:08:19.227 people, is that in reality, 00:08:19.407 --> 00:08:23.347 what you really predict is the sensory consequences of your actions. 00:08:24.447 --> 00:08:27.907 So that framework is more tied to motor predictions. 00:08:27.907 --> 00:08:34.027 Other people, such as Mark Sherrow or others, they have tried to expand this 00:08:34.027 --> 00:08:39.167 framework from the prediction of the sensory consequences of actions to more 00:08:39.167 --> 00:08:40.707 complicated forms of cognition, 00:08:40.947 --> 00:08:45.127 such as reusing this prediction for understanding of the action of others, 00:08:45.367 --> 00:08:48.187 such as reusing this prediction for imagery, 00:08:48.467 --> 00:08:50.967 and so thinking at more abstract situations. 00:08:50.967 --> 00:08:56.787 So, I would say that there are many fields in which prediction have been studied, 00:08:56.947 --> 00:09:00.687 and also some converging frameworks. 00:09:01.027 --> 00:09:08.327 Right. So now, if we talk about your own experiments in the role of prediction in problem solving, 00:09:09.987 --> 00:09:13.587 you showed some examples of what you called embodied problem solving. Yes. 00:09:13.887 --> 00:09:20.447 Right? So how does that, which aspects of the predictive brain are these experiments exercising? 00:09:23.667 --> 00:09:28.367 Well, it's not so simple to explain the videos that I've shown, 00:09:28.547 --> 00:09:32.467 but just to recap, it's just a video of a climber. 00:09:32.587 --> 00:09:38.867 And maybe you know, but climbers prior to a competition, they have some time 00:09:38.867 --> 00:09:42.367 to look at the climbing wall that they will then climb during the competition. 00:09:43.167 --> 00:09:47.347 The nice point is that they see this climbing wall for the first time and there 00:09:47.347 --> 00:09:48.587 are many in these climbing walls, 00:09:48.707 --> 00:09:52.967 there are many climbing holes arranged all over or through the wall and the 00:09:52.967 --> 00:09:58.047 climber has one minute or a few minutes for figuring out how to better climb 00:09:58.047 --> 00:10:04.747 this wall and if you look at the video or if you look at the climbing competition you see people, 00:10:04.747 --> 00:10:11.007 People, climbers really moving the arms such as to really anticipate the next moves. 00:10:11.167 --> 00:10:13.807 They see moving the arms in one direction, then the other direction. 00:10:13.947 --> 00:10:18.907 That's really a form of problem solving because they see the climbing wall for the first time. 00:10:18.947 --> 00:10:21.407 They have to figure out a good plan for action. 00:10:21.607 --> 00:10:25.007 I call it problem solving, not only planning, because it's really complex. 00:10:25.107 --> 00:10:27.147 There are many constraints. It depends on where you go. 00:10:27.767 --> 00:10:32.327 Then you can reach or cannot reach the other hold. So you can also find in your 00:10:32.327 --> 00:10:34.667 mind many, many solutions, compare them. 00:10:35.007 --> 00:10:43.887 And in this specific kind of setups, the hypothesis is that it is really the 00:10:43.887 --> 00:10:47.527 motor system that is governing this process. 00:10:48.327 --> 00:10:55.767 So expert climbers are really able to anticipate the force they have to put into this motor act. 00:10:55.907 --> 00:10:58.767 They can really anticipate a lot of proprioceptive information. 00:10:58.767 --> 00:11:04.027 They can really figure out if a hold is in their reach or out of reach, 00:11:04.167 --> 00:11:08.587 if this hold is too little to be really grasped, if it is too far away for them, 00:11:08.647 --> 00:11:09.927 then they have to take another route. 00:11:10.167 --> 00:11:13.947 So that's, I call it embodied problem solving for many reasons. 00:11:14.027 --> 00:11:18.227 So one reason is that you really use knowledge of your body and of your motor 00:11:18.227 --> 00:11:21.227 system to figure out how to better solve this problem. 00:11:21.227 --> 00:11:24.487 Problem and another reason is that it's also 00:11:24.487 --> 00:11:27.187 overtly embodied in that you really use your 00:11:27.187 --> 00:11:31.767 body your body movements as a scaffold as a help a helping tool for solving 00:11:31.767 --> 00:11:35.907 the problem that's a bit tricky to explain without the video but in a sense 00:11:35.907 --> 00:11:40.427 by simply moving your yourself you don't need to to keep everything into your 00:11:40.427 --> 00:11:44.787 memory you use your body as part of the problem solving process. 00:11:45.747 --> 00:11:50.467 Okay, but now you could also see it as a form of, let's say, motor programming. 00:11:50.587 --> 00:11:55.707 That you say, look, I have to execute a movement sequence very rapidly. Okay. 00:11:56.115 --> 00:12:00.035 Um, and I'm just going to now rehearse this movement sequence that I can sort 00:12:00.035 --> 00:12:06.515 of automatically trigger one set of behavioral motion motor patterns after the 00:12:06.515 --> 00:12:08.675 other, and it can climb up the wall more rapidly. 00:12:09.975 --> 00:12:13.915 Yeah, that's, that's part of the, that's part of the story. So the point is 00:12:13.915 --> 00:12:15.235 forming this motor program. 00:12:15.395 --> 00:12:21.335 Uh, the, the reason why I call it an embodied problem solving is that the solution is not so trivial. 00:12:21.335 --> 00:12:26.875 So the point is that you have to try to form this motor program by assembling 00:12:26.875 --> 00:12:33.955 partial skills that you used in the past in novel ways. 00:12:34.075 --> 00:12:39.935 So there is a lot of flexibility in this new assemblage of motor or small motor programs. 00:12:40.535 --> 00:12:43.955 Yeah, but the problem solving would suggest that there is also, 00:12:44.035 --> 00:12:47.895 let's say, a goal state in the world that you want to achieve in order that 00:12:47.895 --> 00:12:49.975 you also have to manipulate aspects of that world. 00:12:50.435 --> 00:12:55.575 Yeah. Well, in this case, it's essentially making sure your body goes through 00:12:55.575 --> 00:12:56.875 a certain sequence of motions. 00:12:57.815 --> 00:13:03.935 So how would your view on embodied problem solving then generalize to problem solving in general? 00:13:04.095 --> 00:13:07.835 Because I would assume that you want to identify some more, let's say, 00:13:07.855 --> 00:13:12.715 generic aspects of problem solving as opposed to specialized aspects of problem solving. 00:13:13.155 --> 00:13:17.775 Yeah, okay. So there are two parts of this story. So one part is that even in 00:13:17.775 --> 00:13:23.675 that case, even in this climbing example, you have a goal. So the goal is reaching the top. 00:13:24.175 --> 00:13:30.675 And then it is not simply running through a trajectory, but also finding out 00:13:30.675 --> 00:13:31.855 which is the good trajectory. 00:13:32.195 --> 00:13:38.735 So it's not so different from the Tower of London or other typical problem-solving setup. 00:13:38.735 --> 00:13:43.635 So you have a goal state, and you have many, many sequences of moves that you 00:13:43.635 --> 00:13:47.495 can do, and you explore this space of possibilities. 00:13:47.835 --> 00:13:53.235 But the trick is that you explore it in intelligent ways, so you don't simply try out all of them. 00:13:53.455 --> 00:13:57.655 You really use your expertise to find out how to better explore this. 00:13:58.015 --> 00:14:02.595 The second point is that you really use knowledge incorporated into your body, 00:14:02.615 --> 00:14:06.255 your motor programs, to figure out what are the constraints of the problem. 00:14:06.981 --> 00:14:10.381 The climbing problem has many constraints because, as I said before, 00:14:10.541 --> 00:14:16.141 if you are too much to the left of the climbing wall, you cannot reach any more the holds on the right. 00:14:16.381 --> 00:14:24.581 Or vice versa, if you jump too high, you cannot probably make your body stiff 00:14:24.581 --> 00:14:27.361 enough to really hold the climbing hold. 00:14:27.461 --> 00:14:33.181 There are many constraints that you really discover while solving the problem 00:14:33.181 --> 00:14:36.021 by reenacting this motor knowledge. That's my point. 00:14:36.981 --> 00:14:41.261 And the second part of the answer was about how much this generalizes to more 00:14:41.261 --> 00:14:43.001 complex and abstract problem solving. 00:14:43.321 --> 00:14:47.821 So, well, first of all, this is just a nice video that I showed just for illustrating 00:14:47.821 --> 00:14:53.641 the possibility of solving problems by reusing motor knowledge or sensory motor knowledge. 00:14:53.801 --> 00:14:57.641 It could also be affective knowledge. So knowledge incorporated into the same 00:14:57.641 --> 00:15:00.501 models that you use for acting in the external world. 00:15:00.841 --> 00:15:06.201 So I think that as a general strategy, that's the way we should look at higher cognition. 00:15:06.981 --> 00:15:10.881 The challenge is looking at how you solve higher cognitive skills, 00:15:10.901 --> 00:15:16.441 such as problem solving in abstract domains, by reusing these strategies that 00:15:16.441 --> 00:15:21.581 you acquired so as to efficiently deal with the external environment. 00:15:21.581 --> 00:15:28.461 So when you have these sensory-motor strategies, then probably our earlier ancient 00:15:28.461 --> 00:15:32.161 evolutionary ancestors, they 00:15:32.161 --> 00:15:36.321 had only these simple strategies to deal with their current situations. 00:15:37.101 --> 00:15:42.701 Whereas higher cognitive skills is mostly about very complex abstract situations, 00:15:43.121 --> 00:15:45.641 non-perceptually available events, distal goals. 00:15:45.641 --> 00:15:52.681 But the challenge is seeing how simple strategies that we used in the past to 00:15:52.681 --> 00:15:58.481 survive in the environment can be sophisticated and reused in these more complex cognitive domains. 00:15:58.781 --> 00:16:03.401 So that's a challenge. My example was just to illustrate the possibility that 00:16:03.401 --> 00:16:07.481 you can really solve problems by reusing systematically knowledge incorporated 00:16:07.481 --> 00:16:10.341 into your internal models in an intelligent way. 00:16:10.581 --> 00:16:14.561 And that allows also for a lot of flexibility. possibility is not for planning 00:16:14.561 --> 00:16:17.661 the next action, but also for solving very complex problems. 00:16:18.501 --> 00:16:22.001 That's a bit the wish, right? That's the wish. Yes. Then the question is, 00:16:22.101 --> 00:16:25.861 how far are you in realizing that wish? 00:16:26.201 --> 00:16:32.541 So first in the climber case, do you see a big difference between expert and novice climbers? 00:16:33.241 --> 00:16:38.801 Yeah, absolutely. Yes. So part of the story, which is also one of the reasons 00:16:38.801 --> 00:16:44.441 why I'm interested in it, is that But in a sense, it is the expertise that not 00:16:44.441 --> 00:16:46.061 only modulates the climbing ability, 00:16:46.401 --> 00:16:48.701 but also the ability to think about problems. 00:16:49.940 --> 00:16:54.140 That's supporting for the embodied cognition view, for the view that it's really 00:16:54.140 --> 00:16:59.180 knowledge incorporated into your motor system that helps you not only in executing the actions, 00:16:59.380 --> 00:17:06.100 but also for instance in imagining the actions, in reusing the repertoire for solving problems. 00:17:06.460 --> 00:17:11.600 At the same time, we have evidence that knowledge incorporated in your motor 00:17:11.600 --> 00:17:16.060 system also helps you understand what other people are doing. 00:17:16.640 --> 00:17:19.420 Right. So that's the idea in a sense. 00:17:19.820 --> 00:17:22.920 Yeah, that's in the step into the social cognition component. 00:17:23.060 --> 00:17:26.740 But in some sense what you're saying, look, the standard view would be, 00:17:26.820 --> 00:17:32.040 let's say we perceive the world, we have some sort of high fidelity in the end 00:17:32.040 --> 00:17:35.400 interpretation of the world on the basis we should make decisions and perform action. 00:17:35.620 --> 00:17:38.680 But now the consequence of what you're saying is say, look, I'm 00:17:38.680 --> 00:17:41.300 actually acting in this world and the way I act in the 00:17:41.300 --> 00:17:45.180 world is now modulating or directly filtering the 00:17:45.180 --> 00:17:48.340 way I'm going to perceive this world absolutely so the processing 00:17:48.340 --> 00:17:51.280 goes backwards in that sense so the expert 00:17:51.280 --> 00:17:56.140 climbing is looking at this world very differently as a novice climber yes and 00:17:56.140 --> 00:18:02.020 you have evidence for that uh well we yes we perform them uh well of course 00:18:02.020 --> 00:18:06.720 we have a tiny evidence up to now but i think the this evidence points in the 00:18:06.720 --> 00:18:09.800 right direction so one memory study that we performed, 00:18:10.080 --> 00:18:12.900 comparing novice and expert climbers. 00:18:13.780 --> 00:18:20.740 So we simply asked these novice and expert climbers to look at three climbing routes, 00:18:21.020 --> 00:18:25.080 one simple, that both of them were able to execute, one difficult, 00:18:25.320 --> 00:18:28.920 that only the expert group was able to do, and one impossible, 00:18:29.020 --> 00:18:31.120 which was not climbable, actually. 00:18:31.280 --> 00:18:37.040 It was not really a climbing route. It was just a random displacement of climbing holds. 00:18:37.967 --> 00:18:43.707 And so the task was simply remembering the climbing holds in the right sequence. 00:18:44.067 --> 00:18:49.967 What we found is that in the easy condition that both novice and expert were 00:18:49.967 --> 00:18:55.407 able to climb, both were also able to remember quite well. 00:18:55.567 --> 00:19:01.047 So there was no evidence for a better remembering for experts. 00:19:01.207 --> 00:19:07.387 While for the difficult route, only the experts were really able to remember it well. 00:19:07.387 --> 00:19:12.507 And this, in our opinion, this points to the fact that really the way experts 00:19:12.507 --> 00:19:18.867 structure the perceptually also the situation can also help in this memory task. 00:19:19.807 --> 00:19:24.967 And so we had this control condition in which for the impossible route, 00:19:25.127 --> 00:19:30.107 in that route, we did not find any advantage for the experts because this ability, 00:19:30.307 --> 00:19:34.627 this increased memory ability is really tied to the climbability of the wall. 00:19:34.627 --> 00:19:38.867 It's not a generic ability to remember climbing holds without context. 00:19:40.127 --> 00:19:47.767 But to respond completely to your question, I would say that this sensory motor approach, 00:19:48.027 --> 00:19:52.687 so this sensory motor influence of thinking on cognition, to me, 00:19:52.687 --> 00:19:56.387 at least, it acts on two timescales, at least two timescales. 00:19:56.527 --> 00:19:58.927 One timescale is learning and development. 00:19:59.789 --> 00:20:03.889 So the point is that because I have some sensory motor skills, 00:20:04.189 --> 00:20:06.769 then I change the way I perceive the world. 00:20:06.849 --> 00:20:12.129 So I structure my memory and my perceptual abilities in such a way that then 00:20:12.129 --> 00:20:14.129 supports better cognitive abilities. 00:20:14.429 --> 00:20:17.929 So that acts on a longer time scale of learning and development. 00:20:18.389 --> 00:20:23.669 The second way the sensory motor system supports cognition is more in the online cognition. 00:20:23.669 --> 00:20:27.509 Cognition, because I can reenact my motor problems right now, 00:20:27.609 --> 00:20:32.249 my motor programs right now, so I can use them so as to support, 00:20:32.409 --> 00:20:35.509 for instance, imagery or action perception. 00:20:35.729 --> 00:20:40.529 So there are two sides of the same coin. So in a sense, this framework predicts 00:20:40.529 --> 00:20:43.909 that if you increase your sensory motor abilities in one domain, 00:20:44.109 --> 00:20:48.889 then you really shape your perception, your memory, and all your cognitive processing. 00:20:48.889 --> 00:20:56.549 And then you are also better able to run imaginary experiments based on this 00:20:56.549 --> 00:20:58.989 motor expertise that you can now reenact. 00:20:59.249 --> 00:21:05.069 Right. But then would the expert climbers recognize the unclimbable wall more 00:21:05.069 --> 00:21:07.629 rapidly than the novices, novice climbers? 00:21:10.069 --> 00:21:15.549 Well, probably, well, if the task is exactly recognized whether or not it is 00:21:15.549 --> 00:21:18.329 climbable or unclimbable, that's the task. 00:21:18.329 --> 00:21:22.989 If this is the task, well, of course, this is a bit of a tricky task because 00:21:22.989 --> 00:21:27.569 unclimbable could be, is not such a clear concept. 00:21:27.649 --> 00:21:32.989 But anyway, yes, I would say that in that case, what will happen is just that 00:21:32.989 --> 00:21:35.469 the climbers try to climb it mentally. 00:21:35.469 --> 00:21:40.709 So they run this imaginary climbing simulation, and then because the expert 00:21:40.709 --> 00:21:45.209 will anticipate some failure in climbing, because they have high confidence 00:21:45.209 --> 00:21:48.949 in their internal models, they will say, no, no, no, this cannot be climbed. 00:21:49.269 --> 00:21:54.229 Whereas for the novices, maybe they will try out to anticipate. 00:21:54.369 --> 00:22:01.169 They will also fail, but without noticing the, well, without trusting too much their simulations. 00:22:01.269 --> 00:22:04.369 So that will be my... Right. 00:22:04.609 --> 00:22:10.329 But now, still a problem for this point of view is that you could say, 00:22:10.369 --> 00:22:13.929 look, the climber is a highly specialized human being. 00:22:14.929 --> 00:22:19.349 So it's not very surprising that they have certain cognitive capabilities because 00:22:19.349 --> 00:22:24.329 they have been overtrained in dealing with certain kinds of environments like climbing walls. 00:22:24.329 --> 00:22:27.389 So how would you see this sort of. 00:22:28.669 --> 00:22:31.829 But you want to look at this as an example where 00:22:31.829 --> 00:22:37.249 you say well it's actually very basic sensory motor patterns or it's really 00:22:37.249 --> 00:22:42.789 action itself and the way we generate action from which the rest of cognition 00:22:42.789 --> 00:22:49.509 is sort of bootstrapped or is depends upon it's constrained by so how is it 00:22:49.509 --> 00:22:51.229 going to work exactly how do you see that work, 00:22:52.128 --> 00:22:57.168 Well, I would say that now we don't have any perfect story yet, 00:22:57.308 --> 00:23:02.028 but the point is really that action and the goals implied in the action and 00:23:02.028 --> 00:23:06.068 the predictions implied in the action, they are really key to cognition and 00:23:06.068 --> 00:23:08.668 to develop increasingly more complex cognitive skills. 00:23:09.448 --> 00:23:13.108 So at the beginning, so that could be a simple story. 00:23:13.188 --> 00:23:17.168 So at the beginning, the brain is simply organized around these goal selection 00:23:17.168 --> 00:23:19.248 and specification and selection tasks. 00:23:19.248 --> 00:23:22.428 So they have to integrate information from the external world, 00:23:22.508 --> 00:23:28.728 from the memory, from affective states, to very, very quickly jump into a good action selection. 00:23:29.248 --> 00:23:35.348 And to do that, as I told before, you're mostly relying on your internally generated 00:23:35.348 --> 00:23:40.168 goal state, affective processes, attentional processes. 00:23:40.348 --> 00:23:44.108 The stimulus helps you, but then it's the brain, which is autonomous, is doing the job. 00:23:45.068 --> 00:23:50.908 But then as the sensory motor abilities of this primitive architecture develop 00:23:50.908 --> 00:23:54.788 over time and increase your abilities, they increase their ability to control 00:23:54.788 --> 00:23:56.388 the external world and to predict it. 00:23:57.388 --> 00:24:04.108 So they start incorporating more and more knowledge, more and more also structure 00:24:04.108 --> 00:24:09.208 from the external environment into their predictions, into their motor control actions. 00:24:09.208 --> 00:24:12.848 So how are you going to validate this prediction? 00:24:13.068 --> 00:24:16.568 Is it a pure experimental exercise or are there other ways to get this validated? 00:24:17.641 --> 00:24:24.281 The framework, you mean? Well, the framework, that's much more a lifelong research 00:24:24.281 --> 00:24:29.561 program, if you want, because we are trying to do many, many validations. 00:24:29.621 --> 00:24:32.661 A few empirical data, of course, already exist. 00:24:32.761 --> 00:24:37.741 If you think of the literature, how similar the brain networks for imagery and 00:24:37.741 --> 00:24:41.241 for execution or for motor preparation, they are. 00:24:41.241 --> 00:24:46.721 So you already see that some mental operations are really supported by the sensory 00:24:46.721 --> 00:24:49.321 motor system in the brain. So that's some evidence. 00:24:50.221 --> 00:24:55.541 There is evidence, well, there are also patient studies that say that in some 00:24:55.541 --> 00:25:02.441 conditions, the imaginary action cannot be really stopped from being overtly executed. 00:25:02.441 --> 00:25:05.321 So that's another part of the story because 00:25:05.321 --> 00:25:08.761 in this story it is it is the internalization of 00:25:08.761 --> 00:25:11.981 the predictive mechanism that really supports cognition so you 00:25:11.981 --> 00:25:15.941 first predict predict things in the external world then you internalize this 00:25:15.941 --> 00:25:20.861 capability of predicting and this eventually lets you to rehearsing entire sequences 00:25:20.861 --> 00:25:25.601 of action without executing them but then a prediction of this framework is 00:25:25.601 --> 00:25:31.341 that if you really do this covert mental imagery if you want or discover covered predictive magnets, 00:25:31.581 --> 00:25:37.781 and you cannot really separate these internal processes from the overt execution, 00:25:38.181 --> 00:25:41.161 then this implies that what you imagine, you immediately do. 00:25:42.161 --> 00:25:48.781 And there is evidence that this unfortunately happens in patients with bilateral parietal lesions. 00:25:49.561 --> 00:25:54.481 And other evidence of this close connection between what you think and what you do is, 00:25:55.018 --> 00:25:58.298 also exist for utilization behavior for instance utilization behavior 00:25:58.298 --> 00:26:01.518 patients they are not really able to inhibit their 00:26:01.518 --> 00:26:04.618 motor programs for grasping objects and uh just 00:26:04.618 --> 00:26:09.738 reacting in a very uh quick way to what they say so this uh this tells a story 00:26:09.738 --> 00:26:16.538 on on how uh this more um abstract thinking part is tied to the sensory motor 00:26:16.538 --> 00:26:21.378 programs okay but now this these this this in how do you what's What's the role 00:26:21.378 --> 00:26:23.798 of the internal simulation in that case? 00:26:23.938 --> 00:26:27.018 Because on the one hand, you're saying, well, we run these internal simulations. 00:26:27.138 --> 00:26:28.698 Apparently, they would run in parallel. 00:26:29.738 --> 00:26:33.858 But on the other hand, if you look at the deficits that you mentioned in these 00:26:33.858 --> 00:26:40.998 patients, it's not that they are executing a whole sequence of actions in parallel, right? 00:26:41.078 --> 00:26:45.638 In utilization behavior, you would grasp a single object and the whole organism 00:26:45.638 --> 00:26:46.578 would be focused on that. 00:26:46.578 --> 00:26:53.118 So therefore, is this kind of evidence coming more from the clinic, 00:26:53.178 --> 00:26:57.418 from clinical studies, really supporting this hypothesis of internal assimilation 00:26:57.418 --> 00:26:59.198 and the role of action in the structure of cognition? 00:26:59.658 --> 00:27:05.138 Well, for the first evidence I mentioned of the patient who was unable to inhibit 00:27:05.138 --> 00:27:11.878 the imagined action, that's quite clear, because then what you think is what you do. 00:27:11.878 --> 00:27:16.338 So if you imagine, let's say, pointing to one direction, then you point in that 00:27:16.338 --> 00:27:18.298 direction, you cannot really stop from doing that. 00:27:18.518 --> 00:27:23.238 So in this framework that I'm supporting, the idea is that you run this imaginary 00:27:23.238 --> 00:27:24.638 simulation of pointing. 00:27:25.998 --> 00:27:30.018 This is typically done by also inhibiting the overt motor execution. 00:27:30.158 --> 00:27:33.038 But if you have a deficit, you cannot inhibit that output. 00:27:33.198 --> 00:27:36.558 It rapidly turns out into an overt action. 00:27:36.558 --> 00:27:41.518 In the utilization behavior, the link is tiny in a sense. 00:27:41.518 --> 00:27:48.398 The idea is that one idea in the literature of affordances, also the canonical 00:27:48.398 --> 00:27:52.118 neurons literature, is that when you look at an object, 00:27:52.338 --> 00:27:58.458 then you automatically, let's say, pre-potenciate some motor programs that are 00:27:58.458 --> 00:28:00.558 good for interacting with that object. 00:28:00.558 --> 00:28:07.058 Jet okay but typically this does not result into an overt execution so in a 00:28:07.058 --> 00:28:12.398 sense this is just a pre potentiation which can be also interpreted in terms 00:28:12.398 --> 00:28:15.358 of simulating grasping this cap in front of me, 00:28:16.030 --> 00:28:18.710 So in that case, I'm not reasoning about simulating the action. 00:28:18.790 --> 00:28:24.570 This is just a more automatic process of mentally rehearsing of the good action that they can do. 00:28:25.190 --> 00:28:31.530 Again, if you don't have this, let's say, this mechanism working well, 00:28:31.650 --> 00:28:33.990 you cannot really inhibit also the overt execution. 00:28:35.010 --> 00:28:38.830 I would say that while in my first example of the imaginary situation, 00:28:39.190 --> 00:28:42.590 the link is clearer, in this second example, it's less clear. 00:28:42.590 --> 00:28:47.070 But to me, it's another supporting example that your internal mental process 00:28:47.070 --> 00:28:50.770 is always oriented towards anticipating possibilities for action. 00:28:51.210 --> 00:28:56.850 In some cases, this is just a simple automatic process that tells you what is possible to do. 00:28:56.910 --> 00:29:01.230 In other cases, this is more intentional imagery. 00:29:01.510 --> 00:29:05.370 So you really run long-term simulation imaging sequences of actions. 00:29:06.270 --> 00:29:09.110 But I see a continuity in this. 00:29:09.110 --> 00:29:13.930 So the good thing about this framework is that you do not have to postulate 00:29:13.930 --> 00:29:17.090 any different set of cognitive representation for different tasks, 00:29:17.210 --> 00:29:22.870 any module for thinking that is completely segregated from the action control and specification, 00:29:23.270 --> 00:29:28.090 but you have a continual reuse of the same abilities in more complex ways. 00:29:28.885 --> 00:29:33.845 Okay, but then would you see this as being, let's say, layered in some way, 00:29:33.985 --> 00:29:37.865 that you have, let's say, sensory motor capabilities of, let's say, 00:29:37.865 --> 00:29:41.845 varying levels of complexity with some sort of discrete steps between them, 00:29:41.925 --> 00:29:43.245 or is it really a continuum? 00:29:44.905 --> 00:29:49.005 Well, that's a hard question. one important 00:29:49.005 --> 00:29:51.905 thing in this framework that probably answers partially to 00:29:51.905 --> 00:29:54.805 your question is that we have 00:29:54.805 --> 00:29:57.805 to probably focus we have to talk about 00:29:57.805 --> 00:30:00.965 actions as the unity and also we 00:30:00.965 --> 00:30:03.885 know that action can be specified at different levels of detail 00:30:03.885 --> 00:30:06.825 so there are actions that are specified 00:30:06.825 --> 00:30:09.765 at the level of movement of a single finger whereas there 00:30:09.765 --> 00:30:12.625 are other actions in which the action but also 00:30:12.625 --> 00:30:15.905 the goal of the action is specified at a more abstract 00:30:15.905 --> 00:30:18.985 level that is grasping this uh this object 00:30:18.985 --> 00:30:21.825 this cup whereas there is still another level of 00:30:21.825 --> 00:30:27.545 action and intention associated which is uh let's say drinking from this cup 00:30:27.545 --> 00:30:35.585 so the actions uh all these all these uh levels are are in a sense they they 00:30:35.585 --> 00:30:39.325 are they support one another in a sense so of Of course, 00:30:39.325 --> 00:30:46.145 the more abstract actions have to be finally specified in more fine-grained 00:30:46.145 --> 00:30:48.025 terms of the movements of the fingers. 00:30:48.165 --> 00:30:53.765 But this is a structure, a cognitive structure that has different layers probably. 00:30:53.945 --> 00:30:59.405 Or we don't know exactly, but at least you have a big structure in which you 00:30:59.405 --> 00:31:03.925 can really specify actions at different levels of complexity, 00:31:04.025 --> 00:31:06.625 a different level of abstraction with more complex intentions. 00:31:06.625 --> 00:31:12.525 And the arguments that we do is not that we always use the lowest level, 00:31:12.645 --> 00:31:13.905 the lowest possible levels. 00:31:13.945 --> 00:31:19.225 At some point, you can really think and do some imagery by using actions and 00:31:19.225 --> 00:31:23.245 their associated intentions at intermediate or even at abstract levels. 00:31:24.185 --> 00:31:28.965 So that's probably answers. Right. So now one way to test some of these ideas, 00:31:29.305 --> 00:31:33.005 you were performing and presenting experiments on joint action. 00:31:33.145 --> 00:31:43.125 Yes. So how does joint action now help you to understand this model of cognition and action in the end? 00:31:43.385 --> 00:31:50.565 Well, the reason why I went into joint action in my presentation is that, as I told before, 00:31:51.025 --> 00:31:55.725 I would like to come out with a convincing story or more complex cognitive abilities 00:31:55.725 --> 00:31:59.345 developed based on simpler cognitive abilities. 00:31:59.345 --> 00:32:04.245 So here the steps that I presented are, okay, I start doing my action, 00:32:04.365 --> 00:32:12.185 but then I am in a social domain, and so maybe I need to do some action together with you. 00:32:12.942 --> 00:32:17.242 So in this case, I can reuse a lot of what I know about my action system to 00:32:17.242 --> 00:32:21.122 anticipate you and to get into your intentions. 00:32:21.162 --> 00:32:26.222 So to get rid of your motor system or your motor action of your intentions. 00:32:26.542 --> 00:32:32.782 In this case, this is just the beginning of a story in which I reuse my sensory 00:32:32.782 --> 00:32:38.942 motor knowledge or my action execution and prediction abilities to get into 00:32:38.942 --> 00:32:40.362 more and more complex situations. 00:32:40.362 --> 00:32:44.682 In this case, it is a social situation, but it could be even a non-social situation. 00:32:45.502 --> 00:32:52.622 Still another step in this passage is that, okay, now I'm able to use my predictive 00:32:52.622 --> 00:32:54.082 abilities to predict you. 00:32:54.262 --> 00:32:59.142 I use my body and my action system as a model to understand your body and your action system. 00:32:59.982 --> 00:33:05.422 And then I maybe use these abilities to plan long-term joint action. 00:33:05.422 --> 00:33:08.862 To do that then i have planned some more coordinating actions 00:33:08.862 --> 00:33:11.942 so i have to plan how to really achieve them in practice i 00:33:11.942 --> 00:33:14.842 have maybe to to invent some way to better 00:33:14.842 --> 00:33:17.602 coordinate with you and maybe at some 00:33:17.602 --> 00:33:20.542 point you invent more and more complex things you extend the 00:33:20.542 --> 00:33:23.742 boundaries of your control to the control of my body or 00:33:23.742 --> 00:33:27.202 to the control of our body or our combined actions 00:33:27.202 --> 00:33:30.162 and maybe also to the control of your internal 00:33:30.162 --> 00:33:33.202 states so your beliefs your intentions so 00:33:33.202 --> 00:33:36.102 i do some action to change your intention so not only 00:33:36.102 --> 00:33:38.962 i control my body i now control me and you and i 00:33:38.962 --> 00:33:41.942 control you and i know that uh even after 00:33:41.942 --> 00:33:46.462 this interaction you will have some new beliefs i extend my control abilities 00:33:46.462 --> 00:33:52.582 control capabilities from myself to you so that that's uh this the very beginning 00:33:52.582 --> 00:33:58.042 of a story in which we begin to to go into more and more complex cognitive abilities 00:33:58.042 --> 00:34:00.242 is starting from simpler ones. Right. 00:34:02.742 --> 00:34:09.322 So for the joint action experiment, you were looking at hand movements essentially. Right. 00:34:09.846 --> 00:34:12.966 You're tracking hand mostly yes and then you 00:34:12.966 --> 00:34:16.166 also build a model of that yes okay so so 00:34:16.166 --> 00:34:20.726 which aspect now of the joint action control of 00:34:20.726 --> 00:34:28.226 hands do you capture in this in this model well if we think the model specifically 00:34:28.226 --> 00:34:37.006 then uh well in a sense the key idea of the model is that it's not completely is not new. 00:34:37.066 --> 00:34:41.406 There are many similar models, but the idea in a sense is that I really use 00:34:41.406 --> 00:34:46.506 my own action abilities as a model to understand your action right now. 00:34:47.126 --> 00:34:54.086 So the model can predict for instance the hand trajectory, but at the same time 00:34:54.086 --> 00:34:59.086 it can also give some hint into your intentions and goals. 00:34:59.306 --> 00:35:03.446 Because I know what are the goals that are linked to my model actions. 00:35:03.786 --> 00:35:10.926 How does the model know that? Well, let's imagine I see you moving your arm towards a cup, okay? 00:35:11.126 --> 00:35:15.326 So now the hypothesis is that I run an internal simulation of the possible actions 00:35:15.326 --> 00:35:19.826 that I could be that are quite compatible with the actions that I'm seeing. 00:35:20.026 --> 00:35:24.446 I start predicting better and better your trajectory, and then let's see that 00:35:24.446 --> 00:35:27.586 this model fits quite well the data, so your hand trajectory. 00:35:28.126 --> 00:35:29.606 Because I know what is the goal 00:35:29.606 --> 00:35:34.086 of my hand action, I also can have some hypothesis of what is the goal. 00:35:34.086 --> 00:35:38.886 So because I know that my trajectory eventually reaches the cap and then grasps 00:35:38.886 --> 00:35:43.086 it, then I can also infer or at least hypothesize that you will do the same. 00:35:43.206 --> 00:35:44.666 So I also know what is your goal. 00:35:45.546 --> 00:35:53.186 That's the key idea. But then the intentional labeling of the sequence is then 00:35:53.186 --> 00:35:59.906 with reference to your own action or that is in reference in some way to observing the other? 00:36:00.786 --> 00:36:06.606 So the idea of these kind of models is that you use what you know about the 00:36:06.606 --> 00:36:11.206 link between the trajectory of your arms and your goal. You know them because you execute them. 00:36:11.366 --> 00:36:17.226 You use exactly the same link to get some understanding of what the other is doing. 00:36:17.386 --> 00:36:21.906 So that's the idea. So you have first some sort of self-monitoring system that 00:36:21.906 --> 00:36:24.826 develops a self-model that you then generalize to the other. 00:36:24.906 --> 00:36:29.586 This would be the step. Exactly. So I will not necessarily call it a self-monitoring 00:36:29.586 --> 00:36:33.806 in that you simply need a forward model and an inverse model. 00:36:33.946 --> 00:36:36.306 So you need some control capabilities for yourself. 00:36:37.086 --> 00:36:42.586 And then you transfer, you use this exactly as a model to understand the other. 00:36:42.766 --> 00:36:46.446 So that's more or less the key idea. And this idea is pursued by many people 00:36:46.446 --> 00:36:49.166 actually also. Right. Yeah. Okay. 00:36:51.292 --> 00:36:55.892 So what's, do you feel confident then that this original idea you had about 00:36:55.892 --> 00:36:59.532 this more, let's say, embodied problem solving, action as grounding, cognition, 00:36:59.872 --> 00:37:04.912 what tells you in this joint action task where you now are generalizing this 00:37:04.912 --> 00:37:07.692 model to social interaction, that this is working? 00:37:07.812 --> 00:37:10.992 Because in some sense, I could also argue, well, it works because you have been 00:37:10.992 --> 00:37:13.832 imposing, you have really made the task pretty abstract, right? 00:37:13.892 --> 00:37:18.472 Because I'm just observing these hands moving in one plane. plane, 00:37:18.472 --> 00:37:24.292 I don't have to do anything about, let's say, invariant recognition of posture, of limbs and so on. 00:37:24.532 --> 00:37:30.332 So to really now generalize this to, let's say, a realistic scenario where you 00:37:30.332 --> 00:37:35.632 have to observe another human moving about in space, is that just a small matter 00:37:35.632 --> 00:37:39.032 of programming or are there some fundamental steps still missing? 00:37:40.532 --> 00:37:44.812 Well, to answer this question, probably we have to go into a bigger picture. 00:37:44.812 --> 00:37:50.292 The bigger picture to me is that now, although I have emphasized these motor 00:37:50.292 --> 00:37:52.072 predictions, that's not the whole story. 00:37:53.032 --> 00:37:56.472 In a sense, my opinion is that the brain is a smart guy. 00:37:56.772 --> 00:38:02.372 The brain always uses all the knowledge that it has, at least in principle, to do the job. 00:38:03.092 --> 00:38:08.612 The point here is that in more realistic joint action or action observation 00:38:08.612 --> 00:38:10.932 scenarios, there is a lot of information 00:38:10.932 --> 00:38:15.012 available. There is perceptual information, which is available. 00:38:15.212 --> 00:38:18.932 There is some context information, such as, for instance, I see what are the 00:38:18.932 --> 00:38:22.192 objects within your pre-personal space. So that's my prior. 00:38:23.192 --> 00:38:31.172 I can have a lot of prior information on you. For instance, I know that you like very much the beer. 00:38:31.372 --> 00:38:35.752 So if there are in front of you, I see some beer and I see some Coke, 00:38:35.912 --> 00:38:39.012 I would maybe anticipate that you will grasp the beer. 00:38:39.012 --> 00:38:46.392 And then I also use this motor simulation that I told, because it is also useful. 00:38:46.552 --> 00:38:50.432 So I'm not assuming that the brain only uses this motor simulation part. 00:38:50.652 --> 00:38:55.092 It uses the motor simulation to support this action understanding, 00:38:55.492 --> 00:38:59.292 because in this context, it is very salient and very useful. 00:39:00.192 --> 00:39:04.292 Why is it very useful? Because, well, I have a very good model of my body. 00:39:05.084 --> 00:39:08.684 So it's a very good model, very good predictive model for your movements. 00:39:08.924 --> 00:39:15.264 Of course, I have also some purely perceptual predictive mechanisms that can help me predicting you. 00:39:15.364 --> 00:39:21.764 But I would say that in the case of human understanding, the model of myself 00:39:21.764 --> 00:39:24.604 is a very, very good model of you in most cases. 00:39:24.744 --> 00:39:28.144 So this is what the brain will use the most. 00:39:28.424 --> 00:39:32.424 But if you go into the biggest picture, you can think of it as a big Bayesian 00:39:32.424 --> 00:39:36.844 process happening. So in a Bayesian process, you basically use all the information 00:39:36.844 --> 00:39:41.644 you have, information which comes first, then can be also used as a prior to 00:39:41.644 --> 00:39:42.724 run the rest of the process. 00:39:42.884 --> 00:39:48.284 And then all the new information that is collected is fused in an intelligent way. 00:39:48.864 --> 00:39:52.484 Intelligent here means that it is weighted depending on its uncertainty. 00:39:52.724 --> 00:39:59.404 So if the motor system is a very good model, the information it provides will be used very much. 00:39:59.404 --> 00:40:04.324 Much if this information is not good or it is failing then it will not be used 00:40:04.324 --> 00:40:08.524 very much yeah but you seem to now to drift away a little bit from your original. 00:40:10.564 --> 00:40:15.824 Proposal because i thought you started out by saying look cognition is really 00:40:15.824 --> 00:40:19.984 predicated on action yeah right but now you seem to say something like well 00:40:19.984 --> 00:40:24.704 motor the motor system is one of many sources of information you could consider 00:40:24.704 --> 00:40:27.824 because the bayesian system basically use all the knowledge it has, 00:40:27.904 --> 00:40:30.644 and it has knowledge about the motor system, but also about, 00:40:30.744 --> 00:40:34.764 let's say, perceptual systems, or from memory. 00:40:35.124 --> 00:40:38.484 So aren't you a bit drifting now in that argument? 00:40:38.984 --> 00:40:44.544 Well, I don't think so, because as I told before, the way you collect this information 00:40:44.544 --> 00:40:46.644 in the first place is through learning. 00:40:46.904 --> 00:40:50.284 And through learning, the motor system has really influenced you a lot. 00:40:51.064 --> 00:40:55.104 So also through learning, this perceptual information that I have, 00:40:55.244 --> 00:40:58.104 I have acquired it by also interacting with the world. 00:40:58.224 --> 00:41:02.664 So the motor system really is used very much not only online, 00:41:02.864 --> 00:41:04.984 as I told now, but also during learning. 00:41:05.832 --> 00:41:10.752 Of course, we don't have to be radical on that. So we don't think that only 00:41:10.752 --> 00:41:13.492 the motor system does something in the brain. 00:41:13.612 --> 00:41:15.852 There are many parts of the brain that are used. 00:41:16.212 --> 00:41:20.632 And probably the point is slightly different. 00:41:20.832 --> 00:41:24.912 The point is not all about the motor system. So the point is that the ability 00:41:24.912 --> 00:41:27.172 that you have to interact with the external world, 00:41:27.332 --> 00:41:30.312 which can be supported by many systems in the brain, 00:41:30.432 --> 00:41:33.412 prominently by the motor system, but well many system 00:41:33.412 --> 00:41:37.032 you reuse them consistently to achieve new 00:41:37.032 --> 00:41:40.012 cognitive abilities so to develop and and 00:41:40.012 --> 00:41:45.692 then to achieve new new goals specified at different levels okay but um so then 00:41:45.692 --> 00:41:50.292 you're saying well action let's say impregnates all other aspects yes of of 00:41:50.292 --> 00:41:54.992 cognition and perception through the cause of learning okay and this so it has 00:41:54.992 --> 00:41:58.852 a direct and otherwise an indirect impact yeah yeah Yeah, there are both. 00:41:59.032 --> 00:42:03.052 I would say that the first one is through learning. The second one is during 00:42:03.052 --> 00:42:05.052 overt interaction with the external world. 00:42:05.212 --> 00:42:10.432 And then through the internalization of this action and prediction process, 00:42:10.512 --> 00:42:11.832 you can imagine new scenarios. 00:42:12.092 --> 00:42:15.792 But this doesn't mean that when you can have some motor simulation, 00:42:16.032 --> 00:42:18.992 you close your eyes, you don't use percentile information. 00:42:19.232 --> 00:42:22.492 That would be foolish from the point of view of the brain. and and 00:42:22.492 --> 00:42:25.252 there is an important uh point of this view 00:42:25.252 --> 00:42:28.172 is that uh as you know in action understanding 00:42:28.172 --> 00:42:31.132 there are many views one view is that you 00:42:31.132 --> 00:42:34.292 i call this the mirrorless view is that 00:42:34.292 --> 00:42:38.652 the first thing that is active is the recognition of the goal then the recognition 00:42:38.652 --> 00:42:43.912 of the goal in the mirror system of course it eventually helps also predicting 00:42:43.912 --> 00:42:49.452 the action there is a second view which is related but in some sense also very 00:42:49.452 --> 00:42:53.312 different because this view is that what you do first is prediction. 00:42:53.672 --> 00:42:58.612 So it is prediction which comes first and prediction helps then recognizing the goal. 00:42:59.917 --> 00:43:04.397 But still another view is that all these motor simulation things, 00:43:04.677 --> 00:43:09.037 they are not so important because we really use much more abstract knowledge 00:43:09.037 --> 00:43:12.577 of what the other person should do, given the context. 00:43:12.657 --> 00:43:15.197 So that's more theological knowledge, if you want. 00:43:15.837 --> 00:43:24.177 Well, in the view I'm supporting, it is not so useful to say what is more used and what is less used. 00:43:24.557 --> 00:43:28.497 So what comes first can be used as a prior for the rest of the things. 00:43:28.497 --> 00:43:33.457 And what is more reliable is used more than the other things. 00:43:33.757 --> 00:43:35.757 It happens, at least to me, that 00:43:35.757 --> 00:43:39.257 this motor simulation in that specific context, they are very reliable. 00:43:39.697 --> 00:43:44.857 This is why I think they play a prominent role, not because of some categorical distinctions. 00:43:45.357 --> 00:43:49.777 And also relating to this issue of what comes first, goal recognition or prediction, 00:43:50.457 --> 00:43:52.317 that's really depending on the task. 00:43:52.477 --> 00:43:56.557 So if I have a big prior knowledge of what the most likely goals will be, 00:43:56.557 --> 00:44:01.917 then it is probable that I will force some hypothesis or some probability distribution 00:44:01.917 --> 00:44:05.977 over those goals much before starting predicting you. 00:44:06.637 --> 00:44:11.497 But if I cannot see what are the goals, let's say the objects you can grasp, 00:44:11.677 --> 00:44:17.877 I will probably start by predicting and by prediction I will get into some hypothesis of what is your goal. 00:44:18.237 --> 00:44:20.477 So everything can be very flexible. 00:44:21.337 --> 00:44:26.977 Okay, but you are saying ideally you take a model-based approach towards yes 00:44:26.977 --> 00:44:31.837 in this case also social perception you want to say yes you want to get to the 00:44:31.837 --> 00:44:36.457 goal of the intentional state of the other agent as quickly as possible and 00:44:36.457 --> 00:44:40.057 from there you just reconstruct the rest of the action, 00:44:41.077 --> 00:44:44.737 yeah yes in some sense so yeah. 00:44:45.793 --> 00:44:49.793 Typically, I want to go into the goal when this is useful for my task. 00:44:49.953 --> 00:44:57.013 So if my task is understanding which cup will you grasp, that's the goal information that I need. 00:44:57.173 --> 00:45:01.373 Whereas if I don't want to hurt you, it's more the trajectory that I want to know. 00:45:01.453 --> 00:45:08.553 So for example, when we are driving, I don't really need to know where are you going. 00:45:09.053 --> 00:45:15.953 I don't care. I don't want to cross you or to hurt you. So it's not so much 00:45:15.953 --> 00:45:22.433 the goal of what you're doing that is important, but it is the trajectory that you are doing right now. 00:45:22.993 --> 00:45:27.013 Of course, also the goal is interesting because the goal tells me a lot about your trajectory. 00:45:27.333 --> 00:45:35.453 So I would say that what you ask, what you want to infer is very much task dependent. 00:45:36.613 --> 00:45:42.213 And as a side effect, you will also infer many other things that give you useful 00:45:42.213 --> 00:45:45.473 priors or useful hints. experiments. So that's my view. 00:45:46.033 --> 00:45:52.253 You don't only run experiments, right? To test these ideas, you also use robots. 00:45:52.533 --> 00:45:57.693 Yes. So how have robots helped you to make progress on these issues? 00:45:58.973 --> 00:46:04.953 Well, yes, we are running many computational and also robotic experiments. 00:46:05.573 --> 00:46:14.873 So So one thing that we have recently seen is that probably this is very interesting 00:46:14.873 --> 00:46:18.213 again for the argument that more and more complex cognitive abilities can be 00:46:18.213 --> 00:46:20.453 developed on top of this sensory-motor interaction. 00:46:20.673 --> 00:46:24.573 One thing that we are investigating with computational models and robots is 00:46:24.573 --> 00:46:27.093 the ability to signal in social context. 00:46:27.393 --> 00:46:32.073 So just to explain this very quickly, the point is that if we are supposed to 00:46:32.073 --> 00:46:39.933 do one joint action okay let's say building some tower together but of red and 00:46:39.933 --> 00:46:44.033 blue blocks okay one red one blue one red one blue okay now we are interacting 00:46:44.033 --> 00:46:45.453 we are supposed to do that, 00:46:46.033 --> 00:46:52.173 and we are able to coordinate to do that but now let's imagine only i know. 00:46:53.121 --> 00:46:55.741 What is the goal? What is the tower to be built? You don't know. 00:46:56.201 --> 00:47:00.041 So one thing that we are investigating with robots is how we come out with some 00:47:00.041 --> 00:47:05.241 good coordination with only one person knowing the task. 00:47:05.641 --> 00:47:09.981 In that case, without using overt or linguistic communication. 00:47:10.521 --> 00:47:14.181 So what we have seen also in robot, that's a very hard problem. 00:47:14.441 --> 00:47:19.081 Because when you run computational robotic models, then you put your ideas into practice practice, 00:47:19.241 --> 00:47:23.401 and you see that by simply predicting the action of the others doesn't work 00:47:23.401 --> 00:47:27.201 too much because then I start predicting you, 00:47:27.321 --> 00:47:31.961 but so the guy who doesn't know the job has to predict the actions of the other, 00:47:32.061 --> 00:47:36.261 but it is always one step later, it's too late. 00:47:36.441 --> 00:47:41.741 So it doesn't work very well. So we come out with the idea that the guy who 00:47:41.741 --> 00:47:46.361 knows the job can also in some sense support the predictive processes of the 00:47:46.361 --> 00:47:48.261 other or help the other in some sense. 00:47:49.521 --> 00:47:56.701 So how would I help my other robot? Well, one simple example is that I can make 00:47:56.701 --> 00:48:02.981 my behavior more predictable, or I can make my behavior more informative for you. 00:48:03.581 --> 00:48:10.581 Such as I can use my choice of action to letting you understand very well what are my intentions. 00:48:11.461 --> 00:48:15.441 So, you typically try to understand my intentions, but if the uncertainty is 00:48:15.441 --> 00:48:21.121 too high, I can in some sense help you with the wise choice of actions. 00:48:21.321 --> 00:48:22.761 That can be very intuitive. 00:48:23.021 --> 00:48:26.041 But it could also be a way to test hypotheses about the other. 00:48:26.581 --> 00:48:30.361 It's not just making your behavior more predictable. It must be a way to test 00:48:30.361 --> 00:48:32.441 whether the other has the right model of you. 00:48:32.681 --> 00:48:37.581 Absolutely. Yes, that's the idea. So, we design… But it's not the same thing, right? No, no, no. 00:48:37.621 --> 00:48:41.341 Well, they are related. So, in one sense, 00:48:42.337 --> 00:48:48.237 Also, in my example, let's call the guy who has the knowledge the leader and the other the follower. 00:48:48.797 --> 00:48:54.657 So in a sense, the leader can have some uncertainty on what are the models of the follower. 00:48:55.697 --> 00:49:02.517 So, of course, he can use some actions and monitor the reactions of the other 00:49:02.517 --> 00:49:08.417 person just to know what are the good models, what are the models of the other 00:49:08.417 --> 00:49:10.137 agents, what is he doing. 00:49:10.297 --> 00:49:17.737 Right. And when the leader infers that the follower does not have good models, then he will fail. 00:49:17.857 --> 00:49:22.337 Then he can, in a sense, try to help the follower. 00:49:22.617 --> 00:49:24.617 So the two processes are interconnected. 00:49:25.057 --> 00:49:32.537 So by monitoring the process and by always trying to keep track of the uncertainty 00:49:32.537 --> 00:49:36.737 of the other person, you really can plan helping actions. Right. 00:49:37.017 --> 00:49:41.037 But now the thing was interesting that, so we go from action now to interaction. 00:49:41.317 --> 00:49:48.177 Yes. And in your research plan, that's really a very crucial step, right? 00:49:48.197 --> 00:49:53.197 Because you really seem to think that or propose that it's by getting a handle 00:49:53.197 --> 00:49:56.957 on interaction that you really can get to, let's say, a more generalized understanding 00:49:56.957 --> 00:50:00.477 of social behavior and also cultures. 00:50:00.717 --> 00:50:03.497 Yeah. So how should I see that generalization exactly? 00:50:05.217 --> 00:50:11.657 The point is that I think that for many disciplines, such as the study of language 00:50:11.657 --> 00:50:15.937 and communication, the natural starting point would be naturalistic joint action. 00:50:16.708 --> 00:50:21.168 That will be a nice starting point because now the hypothesis is that the mechanism 00:50:21.168 --> 00:50:25.848 that we use for interacting with the others in sensory motor domains, 00:50:26.148 --> 00:50:27.808 then they provide a scaffold. 00:50:28.088 --> 00:50:34.688 So they provide some help for developing, let's say, for instance, 00:50:34.928 --> 00:50:35.828 linguistic communication. 00:50:35.828 --> 00:50:40.948 The hypothesis, which has been called by Levinson this more interaction engine hypothesis, 00:50:41.288 --> 00:50:47.748 is that by simply interacting in the external world, we really have some strong 00:50:47.748 --> 00:50:52.148 pragmatic abilities to infer the action of the other, to do some truth-taking, 00:50:52.388 --> 00:50:53.328 to do some joint attention, 00:50:53.728 --> 00:50:56.308 to anticipate the action and the intentions of the others. 00:50:56.488 --> 00:51:01.028 And this is a strong universal basis also for language communication. 00:51:01.028 --> 00:51:05.828 So, he hypothesizes, for instance, the children use this interaction engine 00:51:05.828 --> 00:51:12.168 to develop, so to understand and to learn their natural language. 00:51:12.348 --> 00:51:15.068 And that's exactly a point that we want to make. 00:51:16.208 --> 00:51:21.728 So we started studying this kind of joint actions. Now we come out with a formulation 00:51:21.728 --> 00:51:26.768 of the problem in which the two persons happen to solve a joint action optimization process. 00:51:27.948 --> 00:51:33.408 This joint action optimization process means that it is not only so I don't 00:51:33.408 --> 00:51:37.728 have to care only about my action, but also the joint outcome of the action. 00:51:38.508 --> 00:51:43.088 Because I have to care about the joint action and the joint outcome of our action, 00:51:43.188 --> 00:51:46.168 which is a joint goal, then I also have to care about you. 00:51:46.728 --> 00:51:51.868 This is why in some cases, as I told before, I help you, not because I'm altruistic. 00:51:52.008 --> 00:51:54.928 I'm helping you because by helping you, I help the joint goal. 00:51:54.928 --> 00:51:59.988 All. So I'm sorry but I'm not so much altruistic, that's all individualistic 00:51:59.988 --> 00:52:05.828 but at the same time I think that this helping action, 00:52:06.589 --> 00:52:09.209 is is really the first primitive form of a 00:52:09.209 --> 00:52:12.069 communicative action because as i told before i 00:52:12.069 --> 00:52:15.049 can make my action more observable by 00:52:15.049 --> 00:52:18.389 you more predictable by you more understandable or more 00:52:18.389 --> 00:52:21.349 diagnostic for you diagnostic here simple 00:52:21.349 --> 00:52:25.229 means simply means that if you have two hypotheses on what i'm doing i make 00:52:25.229 --> 00:52:29.769 my action uh let's say for instance i exaggerate the movement such as to make 00:52:29.769 --> 00:52:34.009 you understand very quickly what i what is my right intentions that's diagnostic 00:52:34.009 --> 00:52:39.109 but that That exaggeration that I do, it's really a form of communication. 00:52:40.009 --> 00:52:44.189 That's really the bootstrap also of all other forms of communication, 00:52:44.409 --> 00:52:45.649 including linguistic communication. 00:52:46.129 --> 00:52:51.829 So the general hypothesis here, this is why I'm interested in this passage from 00:52:51.829 --> 00:52:56.189 joint action to more complex form of cognition, is that by interacting, 00:52:56.389 --> 00:52:58.149 because there are some constraints, 00:52:58.389 --> 00:53:01.349 some joint constraints that we have to fulfill, feel 00:53:01.349 --> 00:53:04.329 then in this joint optimization 00:53:04.329 --> 00:53:07.029 framework it's quite obvious that i have 00:53:07.029 --> 00:53:10.669 to do something also for you for your sake right and 00:53:10.669 --> 00:53:15.409 in our formulation that that that means that in addition to having some motor 00:53:15.409 --> 00:53:20.289 intention we also have some communicative intention the communicative intention 00:53:20.289 --> 00:53:26.609 is has a cost because for instance for exaggerating the movement i pay a cost 00:53:26.609 --> 00:53:28.689 in terms of biomechanical a cost of something. 00:53:28.809 --> 00:53:34.009 So such as, for instance, when I am in a noisy pub, I have to over-articulate. 00:53:34.409 --> 00:53:39.549 Or if you think the mother is, for example, when the child directed speech, 00:53:39.809 --> 00:53:42.909 so you exaggerate, that's done for the sake of the other person. 00:53:43.955 --> 00:53:47.375 Ultimately, we say that's done for the sake of achieving a good joint goal, 00:53:47.535 --> 00:53:48.875 which is understanding one another. 00:53:49.115 --> 00:53:55.355 But that's really the first form of motor of communicative intention that I have. 00:53:55.535 --> 00:53:59.255 So it's for you, it's really the signaling mechanisms that would allow the bootstrapping 00:53:59.255 --> 00:54:01.815 of communication and language, right? 00:54:02.015 --> 00:54:05.875 But that's still a project that will be realized in the future. 00:54:06.135 --> 00:54:11.935 Yes, but I would say that the important component is always the missing one 00:54:11.935 --> 00:54:14.915 from the current studies, which is the more pragmatic component of language. 00:54:15.095 --> 00:54:21.175 So the pragmatics of language as the early pragmatists, which study language know very, very well. 00:54:21.355 --> 00:54:25.715 So the pragmatic component is prominent because then the ground ego symbols, 00:54:25.855 --> 00:54:28.855 the grammar, of course, they are hard problems, but the pragmatics, 00:54:28.855 --> 00:54:35.715 they are the most important part, such as for instance, knowing Knowing what requesting means, 00:54:35.955 --> 00:54:44.255 knowing what telling means or promising means or all these joint action and joint attention, 00:54:44.575 --> 00:54:49.635 tour-taking mechanism, all of them provide a big scaffolding for the emergence 00:54:49.635 --> 00:54:51.315 of linguistic communication. Right. 00:54:51.775 --> 00:54:55.655 So I got that. But now, so to finish up, two questions. 00:54:57.115 --> 00:55:01.115 So you have this really big ambitious program, and in the end, 00:55:01.155 --> 00:55:07.075 you'll generate a whole of culture from action, which is going to be still a 00:55:07.075 --> 00:55:10.155 long trajectory, but you're well on your way. 00:55:10.655 --> 00:55:13.615 But in doing this, what would be Giovanni's Law? 00:55:14.615 --> 00:55:22.335 Giovanni's law? Okay. Well, my law would be maybe starting understanding cognitions 00:55:22.335 --> 00:55:27.515 from the primitive equipment of our ancestors. 00:55:27.875 --> 00:55:32.935 So that would be more an evolutionary route towards the more complex cognitive skills. 00:55:33.275 --> 00:55:35.955 Well, I can probably give three 00:55:35.955 --> 00:55:39.535 examples that are very simple in the three domains that we target now. 00:55:39.695 --> 00:55:41.855 One is the social domain that we have discussed. 00:55:42.055 --> 00:55:47.215 So you start from controlling your body, your actions, to controlling our actions, 00:55:47.335 --> 00:55:49.055 to controlling then your mental states. 00:55:49.235 --> 00:55:53.255 That's communication in a sense. Communication is really putting something into your brain. 00:55:53.675 --> 00:55:57.215 And then you also develop culture and pedagogy, for instance, 00:55:57.315 --> 00:56:00.295 for controlling that at a very, very longer timescale. 00:56:00.355 --> 00:56:05.775 So by developing some culture, the human beings can really control the behavior 00:56:05.775 --> 00:56:09.555 of many people over millennia probably, and also by pedagogy. 00:56:09.555 --> 00:56:10.895 So that's in the social domain. 00:56:11.375 --> 00:56:16.715 You can see this also in the control of the external world, on this environmental 00:56:16.715 --> 00:56:19.575 scaffolding or the environmental shaping. 00:56:19.655 --> 00:56:24.195 So we start from the control of our body to the control maybe of the reality. 00:56:25.269 --> 00:56:28.249 Our peripersonal space with some objects and then 00:56:28.249 --> 00:56:31.169 you start controlling tools for instance 00:56:31.169 --> 00:56:34.609 for doing more complex actions and but then 00:56:34.609 --> 00:56:37.589 tools you have also to build tools maybe at some 00:56:37.589 --> 00:56:40.329 point you build the city of barcelona so then in a sense 00:56:40.329 --> 00:56:44.149 you extend the boundaries of your control from the control your body to the 00:56:44.149 --> 00:56:48.749 control of the whole environment we humans really modify our environment very 00:56:48.749 --> 00:56:53.149 much such that then they support much better our cognition so that's another 00:56:53.149 --> 00:56:57.769 story not only in the social domain also in the environmental domain the third 00:56:57.769 --> 00:57:00.069 the third and last one is in the more, 00:57:00.669 --> 00:57:04.969 thought domain or the cognitive control domain so you just start from controlling 00:57:04.969 --> 00:57:10.129 your action and then you end up in some way by controlling your thought processes 00:57:10.129 --> 00:57:15.549 and by controlling your cognition over long long long time scales for instance 00:57:15.549 --> 00:57:17.009 one example that i do well, 00:57:17.649 --> 00:57:21.149 let's imagine i want to become a prominent scientist 00:57:21.149 --> 00:57:24.029 which apparently doesn't doesn't happen right 00:57:24.029 --> 00:57:27.209 now so i have to control my behavior for 20 years 00:57:27.209 --> 00:57:33.009 maybe so the way i do that is by well by cognitive control in a sense by setting 00:57:33.009 --> 00:57:37.429 some goal states into my mind then then control my behavior on longer longer 00:57:37.429 --> 00:57:42.589 longer time scales then again you have this control of the body the or the control 00:57:42.589 --> 00:57:45.929 of my short-term actions that become the control of my entire life. 00:57:48.669 --> 00:57:53.509 I have given three examples of how this research program, I would say, 00:57:53.669 --> 00:57:57.829 would start from very simple sensory motor abilities to very, 00:57:57.909 --> 00:58:02.329 very complex abilities that cross the boundaries of many disciplines. 00:58:02.489 --> 00:58:05.789 Sociology on the one hand, and let's say architecture on the other hand, 00:58:05.869 --> 00:58:08.469 or maybe the study of consciousness on the other hand. 00:58:08.649 --> 00:58:11.689 And you have even integrated it in your own career planning. 00:58:12.089 --> 00:58:16.449 Ah, of course. so the last thing the very last question then is prediction so 00:58:16.449 --> 00:58:21.529 if we come back then five years from now as a famous scientist to the summer school again. 00:58:23.308 --> 00:58:27.088 So what's the one prediction we can test on you by then? 00:58:27.188 --> 00:58:31.408 So what's the one prediction you would like to make today that you feel very strongly about? 00:58:31.708 --> 00:58:34.068 And we can ask you about five years from now. 00:58:34.668 --> 00:58:39.928 Okay, so about the future of this field or about the… Your research program. Okay. 00:58:41.088 --> 00:58:46.468 So about my research program. Well, I think I elucidated the principle, 00:58:46.828 --> 00:58:52.328 my key goal, which is understanding higher cognition, how our cognition is based 00:58:52.328 --> 00:58:55.248 on sensory motor and predictive abilities. 00:58:55.528 --> 00:58:59.468 So the prediction is that at least we will have some nice demonstration in these 00:58:59.468 --> 00:59:04.388 three fields that I've mentioned before, much stronger demonstration than the ones that we have now. 00:59:04.648 --> 00:59:09.928 And from that, we can really try to build an understanding of how the brain 00:59:09.928 --> 00:59:11.348 implements higher cognitive skills. 00:59:11.648 --> 00:59:16.168 So at least one demonstration for each of the three fields. That's very good. 00:59:16.248 --> 00:59:19.428 So Giovanna Pizzuto, thank you very much for this conversation. Thank you, Paul. 00:59:22.148 --> 00:59:28.148 The csn podcast was produced by the convergent science network of biometrics 00:59:28.148 --> 00:59:34.628 and biohybrid systems a project funded by the european sevens research framework program. 00:59:36.240 --> 01:00:02.886 Music.