WEBVTT 00:00:00.017 --> 00:00:07.177 Okay, so I'm here with John Doyle. We're still both at the KITP for the workshop 00:00:07.177 --> 00:00:08.357 on network architecture. 00:00:08.717 --> 00:00:11.797 This is Paul Fouchure, the Convergence Science Network. 00:00:11.797 --> 00:00:18.337 Work and john you you presented your work on on architectures as a control engineer, 00:00:19.017 --> 00:00:22.077 you looked at at architectures in 00:00:22.077 --> 00:00:26.017 particular focusing on single cellular organisms right 00:00:26.017 --> 00:00:29.097 so so and that seems a bit of counterintuitive move 00:00:29.097 --> 00:00:31.917 for a control engineer but why why did 00:00:31.917 --> 00:00:35.257 you move to the to biology well i 00:00:35.257 --> 00:00:38.437 got interested in biology you know around 10 years ago or 00:00:38.437 --> 00:00:41.337 10 or 15 years ago um i'd been 00:00:41.337 --> 00:00:44.117 working initially motivated i'm pretty much 00:00:44.117 --> 00:00:47.277 a very much at the theoretical end my phd is in math but i 00:00:47.277 --> 00:00:52.097 was early mostly motivated by aerospace control problems because they were doing 00:00:52.097 --> 00:00:56.917 in you know the 80s the most cutting edge things but then in the 90s networks 00:00:56.917 --> 00:01:01.717 started coming you know we we the internet really grew to prominence um didn't 00:01:01.717 --> 00:01:07.037 really have a theory so i got interested more in networks and uh just kind of gravitated towards, 00:01:07.997 --> 00:01:15.017 um bacteria because i had uh colleagues at caltech who worked on it it was kind 00:01:15.017 --> 00:01:23.737 of a convenient um but also um i've found that they have in nature one of the most remarkable. 00:01:24.817 --> 00:01:27.897 And evolvable architectures the architecture of 00:01:27.897 --> 00:01:34.297 the bacterial biosphere um for one thing it evolved into us but it also continues 00:01:34.297 --> 00:01:39.697 to have tremendous evolvability and robustness and so i want to understand what 00:01:39.697 --> 00:01:45.917 gave that biosphere that robustness and i think we've now found. 00:01:47.177 --> 00:01:53.037 Deep principles that are very consistent with what we're learning about how 00:01:53.037 --> 00:01:59.117 to build technology networks so it was examples in addition to the internet 00:01:59.117 --> 00:02:00.117 of really sophisticated 00:02:00.497 --> 00:02:01.857 evolvable networks that we 00:02:01.857 --> 00:02:05.457 could compare but now before you go to internet and talk about robustness, 00:02:06.117 --> 00:02:09.417 What's what are these principles of robustness that? 00:02:09.897 --> 00:02:16.297 That you see these biological systems. Well so one thing is they are. 00:02:17.809 --> 00:02:25.269 They get robustness by having a plug-and-play, sort of what we in technology 00:02:25.269 --> 00:02:26.689 call a plug-and-play architecture. 00:02:27.049 --> 00:02:35.549 That is, on almost every timescale you look, they can quickly adjust to, 00:02:35.689 --> 00:02:37.729 say, changing environments. 00:02:39.149 --> 00:02:44.749 And so on very short timescales, they can very quickly rearrange their protein 00:02:44.749 --> 00:02:48.189 networks to respond as they needed. But on longer time scales, 00:02:48.309 --> 00:02:51.629 they can change their genomes very rapidly. 00:02:51.709 --> 00:02:53.749 They can swap genes. 00:02:54.329 --> 00:03:00.169 So the bacterial biosphere as a whole almost acts as one large gene pool. 00:03:00.689 --> 00:03:05.109 So on long time scales, they can very quickly evolve their genome. 00:03:05.169 --> 00:03:08.329 And on short time scales, they're very quick in responding to the environment. 00:03:08.329 --> 00:03:12.989 And what's interesting is I think their architecture makes both of those better. 00:03:13.169 --> 00:03:16.669 And that's one of the things that I think has surprised the biologists a bit 00:03:16.669 --> 00:03:22.889 is that the proper architecture can facilitate both rapid robustness on short 00:03:22.889 --> 00:03:26.969 time scales and rapid evolvability on long time scales. 00:03:26.969 --> 00:03:30.209 Else but now in your so 00:03:30.209 --> 00:03:32.909 this is a bit the phenomena right these are 00:03:32.909 --> 00:03:37.349 the phenomena we want to understand but not in in terms of principles you you 00:03:37.349 --> 00:03:44.049 talked about layering you talked about constraints um and you talk about let's 00:03:44.049 --> 00:03:48.689 say feedback so so how does this now relate to these phenomena that we're trying 00:03:48.689 --> 00:03:52.669 to look at right well the way we the way in engineering Engineering, 00:03:52.689 --> 00:03:54.949 we think about, and mathematics, 00:03:55.009 --> 00:03:57.989 the way we think about describing architecture is in terms of constraints. 00:03:58.989 --> 00:04:09.209 And the IEEE Transactions paper on architecture takes that point of view, 00:04:09.229 --> 00:04:12.829 as does the recent PNAS paper aimed at neuroscientists, and it tries to explain 00:04:12.829 --> 00:04:13.809 how we think about constraints. 00:04:14.848 --> 00:04:18.008 And one of the place constraints come from is the behavior of the system as 00:04:18.008 --> 00:04:25.168 whole is constrained by the environment it must act in and if it's going to survive and persist. 00:04:26.168 --> 00:04:31.308 And then there's usually constraints on the components which come up from the bottom. 00:04:31.388 --> 00:04:34.608 What can you build? What are the resources? What can you build it out of? 00:04:35.668 --> 00:04:41.468 And then additionally, there are constraints that come out of theory in some sense. 00:04:41.668 --> 00:04:47.128 I mean, theory reflects reality. But the idea is that much of engineering theory 00:04:47.128 --> 00:04:54.288 is about the additional hard limits that come either out of limitations on computing, 00:04:54.528 --> 00:04:57.068 which is Turing, limitations on communication, which is Shannon. 00:04:57.188 --> 00:05:05.008 And what I talked about in this KTP lectures is the Bode limits on robustness and feedback systems. 00:05:05.968 --> 00:05:08.248 These aren't well known in the scientific community. 00:05:09.388 --> 00:05:11.868 And they're a little bit fragmented even within engineering. 00:05:11.868 --> 00:05:14.128 But we're trying to build unified theories now. 00:05:14.428 --> 00:05:18.108 But the final constraint then is the design choices that are made, 00:05:18.248 --> 00:05:20.928 either through evolution or through deliberate design. 00:05:21.648 --> 00:05:27.148 And Gerhardt and Kirshner, who are biologists, have a very nice phrase for what 00:05:27.148 --> 00:05:29.628 a good architecture is, is constraints that deconstrain. 00:05:29.628 --> 00:05:34.568 And you choose a few constraints, and it might be, for example, 00:05:34.708 --> 00:05:41.228 the use of ATP as an energy carrier, the use of NADH as a redox carrier, the codons. 00:05:41.648 --> 00:05:45.508 In the internet, it would be the packet formats and the TCPIP protocol. 00:05:47.308 --> 00:05:52.368 Those are chosen wisely. Then they free you up to use that as a platform. 00:05:53.125 --> 00:05:56.205 For both very robust networks and very evolvable networks and that's 00:05:56.205 --> 00:05:59.105 where layering comes in is because layering is seems to 00:05:59.105 --> 00:06:04.545 be the strategy that both engineers and biology have adopted to make flexible 00:06:04.545 --> 00:06:12.405 uh robust evolvable systems and what layering is is a particular way of structuring 00:06:12.405 --> 00:06:17.825 the chosen constraints and so that's what we talked a lot about it's not a simple topic 00:06:17.905 --> 00:06:20.885 but um it's familiar enough 00:06:20.885 --> 00:06:24.045 to people from neuroscience and from cell biology and 00:06:24.045 --> 00:06:27.645 from engineering that there is becoming 00:06:27.645 --> 00:06:30.585 i'm hoping a common language which is what i'm trying to promote about 00:06:30.585 --> 00:06:33.925 what is layering but then how do 00:06:33.925 --> 00:06:36.785 layers map to modules right because also in 00:06:36.785 --> 00:06:39.525 your analysis you you stress the notion of 00:06:39.525 --> 00:06:42.205 modules quite a bit yeah so how should i see this 00:06:42.205 --> 00:06:46.025 is related to layers and modules so layering is 00:06:46.025 --> 00:06:53.825 i think the the highest level view of what modularity is and i think it's the 00:06:53.825 --> 00:06:59.525 most uh important aspect of modularity um i think there's been a lot written 00:06:59.525 --> 00:07:02.785 in the scientific community lately about modularity that is i think 00:07:02.905 --> 00:07:11.265 naive and not wrong so much as not the most important aspects of Larry. 00:07:12.145 --> 00:07:18.505 And so the thing that people are probably most familiar with modularity is the laptop that they have. 00:07:19.385 --> 00:07:24.625 And they can, or they download this podcast and it just runs immediately. 00:07:24.985 --> 00:07:28.865 And so the idea is that the kind of modularity it has is that you can download 00:07:28.865 --> 00:07:35.345 new software, you can download new podcasts and they immediately work largely 00:07:35.345 --> 00:07:38.045 independent of which computer you happen to have. 00:07:38.848 --> 00:07:44.168 Um, and you can, you can go to the store, uh, order online, uh, 00:07:44.268 --> 00:07:45.728 hardware that you just plug in. 00:07:46.208 --> 00:07:50.268 And what makes that possible is what's hidden, 00:07:50.368 --> 00:07:54.708 what you don't see, which is the operating systems and that, 00:07:54.768 --> 00:07:59.428 that they sit between all these applications and the hardware and, 00:07:59.508 --> 00:08:01.508 uh, and allow you the modularity. 00:08:01.508 --> 00:08:05.948 And the idea is that the modularity looks so simple and so easy when it works. 00:08:05.988 --> 00:08:08.548 It usually either works perfectly or not at all. 00:08:10.208 --> 00:08:16.188 But it's the layered structure that allows that to happen. 00:08:16.348 --> 00:08:21.008 And we believe that's the same thing in biology. It's what allows for horizontal 00:08:21.008 --> 00:08:27.908 gene transfer as a mechanism for bacteria to evolve rapidly, for example. 00:08:27.908 --> 00:08:30.508 And that so they have the same sort of 00:08:30.508 --> 00:08:33.688 plug and play modularity that our technologies 00:08:33.688 --> 00:08:37.168 have when they're good but now does modularity always 00:08:37.168 --> 00:08:41.568 imply let's say a static organization of 00:08:41.568 --> 00:08:44.868 of a structure or can you 00:08:44.868 --> 00:08:47.968 have let's say virtual modules that change if 00:08:47.968 --> 00:08:51.228 you want their boundaries and their interfaces to other modules so 00:08:51.228 --> 00:08:54.848 how could i think about this well i think i think 00:08:54.848 --> 00:08:57.748 the constraints a deconstraint way of looking at 00:08:57.748 --> 00:09:00.888 this is that there's always different time 00:09:00.888 --> 00:09:03.868 scales for different aspects of the architecture there's there's usually 00:09:03.868 --> 00:09:07.848 things that are persistent on very long time scale so an example of something 00:09:07.848 --> 00:09:11.548 that is persistent on very long time scales in biology would be the use of atp 00:09:11.548 --> 00:09:16.108 as an energy carrier it's both universal across all of biology and probably 00:09:16.108 --> 00:09:23.428 has been there for billions of years um the the uh the codon usage is nearly universal and has 00:09:23.468 --> 00:09:26.368 probably been there for also similar long times. 00:09:26.508 --> 00:09:30.488 So those are constraints that really persist for a long, long time. 00:09:31.168 --> 00:09:38.168 What they facilitate, however, is tremendously dynamic responses. 00:09:38.728 --> 00:09:42.068 And so by having a common energy carrier throughout the cell. 00:09:42.888 --> 00:09:46.488 Individual modules don't have to worry about an energy source. 00:09:47.508 --> 00:09:50.508 Core metabolism make sure that that 00:09:50.508 --> 00:09:53.448 the atp charge is maintained as constant as 00:09:53.448 --> 00:09:56.488 possible and an enormous amount 00:09:56.488 --> 00:10:01.288 of sophisticated control goes on inside the cell or inside our bodies to maintain 00:10:01.288 --> 00:10:07.088 that energy charge um and then what that allows you to do is on again on every 00:10:07.088 --> 00:10:13.628 time scale adapt very dynamically to the challenges so what you see is the combination of a 00:10:13.988 --> 00:10:16.228 permanent constraint that's fixed forever. 00:10:17.792 --> 00:10:23.452 But that choice, if done wisely, enables very, very much dynamics. 00:10:23.632 --> 00:10:30.832 Now what happens is on fast time scales, even what's a module is changing very rapidly. 00:10:31.112 --> 00:10:39.052 So for example, just in say in metabolism, you'll the depending on what, 00:10:39.052 --> 00:10:45.272 what biosynthetic pathways are needed, we may come and go on demand. 00:10:45.272 --> 00:10:48.232 Um and and what 00:10:48.232 --> 00:10:51.212 what and depending again on what's in the environment what 00:10:51.212 --> 00:10:54.052 was a food source may go away 00:10:54.052 --> 00:10:57.012 and now all of a sudden there that's a say an amino acid that 00:10:57.012 --> 00:11:00.272 you have to start manufacturing again this is bacteria are the 00:11:00.272 --> 00:11:03.592 most plastic in this regard we're not very flexible right but 00:11:03.592 --> 00:11:06.412 at the same time we can eat all sorts of things um and so we're 00:11:06.412 --> 00:11:09.232 omnivores and our ability to do that is 00:11:09.232 --> 00:11:12.052 again because we have a 00:11:12.052 --> 00:11:15.192 sort of front-end process that turns 00:11:15.192 --> 00:11:18.432 whatever we eat into a few basic building blocks from 00:11:18.432 --> 00:11:21.452 which we synthesize all of our tissues okay but now if you 00:11:21.452 --> 00:11:24.832 if you describe it like this coming from engineering you 00:11:24.832 --> 00:11:28.092 seem to be implying that actually engineering also 00:11:28.092 --> 00:11:31.292 works according to these principles yeah is it actually true 00:11:31.292 --> 00:11:34.452 for for sort of the larger systems 00:11:34.452 --> 00:11:37.852 that we have put together as humans well one. 00:11:37.852 --> 00:11:40.672 Of the problems is the larger systems we put together as humans 00:11:40.672 --> 00:11:44.412 are not sustainable and so they 00:11:44.412 --> 00:11:47.792 have design flaws and one of the ironies is uh 00:11:47.792 --> 00:11:50.712 when i first started getting interested in biology the biologists 00:11:50.712 --> 00:11:53.752 would tell me oh you know you're going to see biology's kludgy it's 00:11:53.752 --> 00:11:57.232 an accident it's a tinker evolution's the tinkerer um that's 00:11:57.232 --> 00:12:00.332 very true but in fact you don't 00:12:00.332 --> 00:12:03.872 you don't tend to see uh the gross 00:12:03.872 --> 00:12:10.092 design flaws in biology that are all over in engineering and the reason for 00:12:10.092 --> 00:12:14.892 that is that that much of the core uh these core protocols in biology have had 00:12:14.892 --> 00:12:21.432 millions to billions of years to evolve um and And sure, 00:12:21.512 --> 00:12:25.012 we have some funny aspects of because we were built out of fish parts. 00:12:25.552 --> 00:12:29.752 And so building a human out of fish parts is going to have some problems. 00:12:30.832 --> 00:12:34.232 But the the biochemical layer. 00:12:35.311 --> 00:12:39.571 It seems to be spectacularly uh well designed from an engineering point of view, 00:12:40.271 --> 00:12:44.471 so i think and so what the what you what's hard to do with any of these systems 00:12:44.471 --> 00:12:51.451 is to sort out what is a hard necessary constraint and trade-off and and what is a, 00:12:52.091 --> 00:12:56.571 an accidental design flaw either of evolution or of engineering and then you 00:12:56.571 --> 00:13:02.131 see both um so the problem with engineering though is in our at the big scale 00:13:02.131 --> 00:13:04.711 we're profoundly unsustainable, 00:13:05.311 --> 00:13:09.531 In our energy, transportation, waste, water, food, almost everything. 00:13:10.451 --> 00:13:13.591 And the more you learn about it, the more disturbing it is. 00:13:13.711 --> 00:13:17.491 The more I understand ecosystems and their interactions with our technologies, 00:13:17.651 --> 00:13:19.611 the more frightened I am. 00:13:20.311 --> 00:13:24.951 I think I'm, funny, I'm kind of a global warming skeptic. I think it's much 00:13:24.951 --> 00:13:27.011 worse than the scientists are telling us. 00:13:27.051 --> 00:13:31.711 I don't think they have the tools yet to do the sort of worst case scenario 00:13:31.711 --> 00:13:33.551 analysis that we do in engineering. 00:13:34.651 --> 00:13:40.091 And we take for granted that our planes rarely crash, and that's because engineers 00:13:40.091 --> 00:13:41.591 do that kind of worst-case analysis. 00:13:42.011 --> 00:13:45.351 So we do certain things very, very well in engineering. We've gotten to the 00:13:45.351 --> 00:13:50.431 point where we can build platforms like airplanes and vehicles that are very 00:13:50.431 --> 00:13:55.851 sophisticated and very robust, but we're not so good at the networks yet. 00:13:56.251 --> 00:14:02.311 And the internet was a spectacular innovation, but it now has a lot of recognizable design flaws, 00:14:02.311 --> 00:14:05.091 but we've built so much on top of it 00:14:05.091 --> 00:14:08.191 that it's hard to now change 00:14:08.191 --> 00:14:11.031 those if you make a if you make a mistake in 00:14:11.031 --> 00:14:13.831 a core protocol and then you build a 00:14:13.831 --> 00:14:18.951 lot on top of it it can be very hard to now change that right so if atp wasn't 00:14:18.951 --> 00:14:24.191 a good energy carrier biology would have very little choice but to just stick 00:14:24.191 --> 00:14:29.651 with it yeah exactly so but now so so what but this seems to be also a conundrum 00:14:29.651 --> 00:14:32.151 a bit in your, well, these ladies are getting a bit noisy, right? 00:14:32.231 --> 00:14:34.771 So maybe we should change location. 00:14:39.051 --> 00:14:42.671 Because I have to filter all this stuff out later again. Keep on going. 00:14:44.831 --> 00:14:49.991 Because what I want to bring up now is that in your, so the one that now we 00:14:49.991 --> 00:14:55.591 see, if you want, a bit of a contrast here, right, between your analysis of 00:14:55.591 --> 00:14:59.131 life, of biology, biological systems, Thank you. 00:15:00.139 --> 00:15:06.599 And the success of this framework you see in nature, in actual development of 00:15:06.599 --> 00:15:11.019 these large integrated systems like internet, right? 00:15:11.059 --> 00:15:15.619 Because like you said earlier, if you look at how internet has been put together, 00:15:15.759 --> 00:15:18.099 it's one big kludge, right? It's not really. 00:15:18.239 --> 00:15:27.319 But it was based on a few very profound and visionary insights. 00:15:28.759 --> 00:15:32.039 And it's not necessarily from from control engineering 00:15:32.039 --> 00:15:35.299 in any way no no in fact it was operating systems 00:15:35.299 --> 00:15:39.099 uh and engineers right and um and 00:15:39.099 --> 00:15:42.619 they now they knew elementary 00:15:42.619 --> 00:15:45.279 control theory and made quite a bit of use of 00:15:45.279 --> 00:15:47.899 it at a time and they did it at a 00:15:47.899 --> 00:15:50.939 time when uh communication theory was almost 00:15:50.939 --> 00:15:54.619 entirely dominated by information theory and 00:15:54.619 --> 00:15:59.439 so decades ago there was a major battle between sort of the information theorists 00:15:59.439 --> 00:16:09.039 and the operating systems people over how to build these networks and and for 00:16:09.039 --> 00:16:11.379 almost accidental reasons in the u.s. 00:16:11.379 --> 00:16:16.879 The OS system people won and they built the internet and now what's happened 00:16:16.879 --> 00:16:21.439 in the last 10 to 15 years as we really now have a much more integrated theory 00:16:21.439 --> 00:16:26.059 that includes information and control theory and And greatly explains both the 00:16:26.059 --> 00:16:29.239 successes and the flaws in the current Internet architecture. 00:16:29.899 --> 00:16:34.679 And one of the big challenges now is to use those insights to build a next generation. 00:16:35.099 --> 00:16:40.239 But the Internet's a good example of a situation where the really brilliant 00:16:40.239 --> 00:16:46.059 engineers are always way out ahead of any of the theoretical understanding. 00:16:46.499 --> 00:16:49.699 And so we're always playing catch up with the best. 00:16:49.819 --> 00:16:54.259 It's the same way, I think, in medicine. And I think the best doctors are always 00:16:54.259 --> 00:16:56.879 making leaps way out ahead of any scientific basis. 00:16:58.079 --> 00:17:03.619 And then, and I think the scientific community in some ways quit doing this, 00:17:03.699 --> 00:17:07.259 is trying to come back and help the doctors flesh out the details. 00:17:07.459 --> 00:17:12.859 I think the doctors have, for a few decades, really been on their own as the 00:17:12.859 --> 00:17:17.239 scientific community pursued genomics as the answer to everything. 00:17:17.239 --> 00:17:20.279 Thing right but the only problem is of course that with respect 00:17:20.279 --> 00:17:23.239 to internet we might be already so much committed into 00:17:23.239 --> 00:17:28.199 one certain way of structuring the system that we have no chance ever to to 00:17:28.199 --> 00:17:32.379 re-engineer or redeploy anything that's a big worry yeah that's a big worry 00:17:32.379 --> 00:17:39.019 and so um i'm a little less concerned about that than i was from a technical 00:17:39.019 --> 00:17:40.899 point of view in the sense that i think. 00:17:42.036 --> 00:17:48.176 The architectures that I think are the most appealing that are being proposed 00:17:48.176 --> 00:17:53.416 actually have the internet as kind of a lousy special case. 00:17:53.756 --> 00:17:58.916 So I think it could be organically grown around a new architecture and then 00:17:58.916 --> 00:18:00.296 it's a little bit like a city. 00:18:00.376 --> 00:18:06.556 You could sort of leave the old parts as a kind of scruffy old neighborhood. 00:18:07.376 --> 00:18:13.016 I think that's possible. I think the challenge is going to be probably in these 00:18:13.016 --> 00:18:20.556 big systems, not technological, but the interplay between technology, markets, and policy. 00:18:20.556 --> 00:18:27.196 And I think if there's a system that we understand most poorly, 00:18:27.416 --> 00:18:30.676 it's the interplay between technology and policy and markets. 00:18:30.976 --> 00:18:39.576 And right now, extremely ideologically driven, not the least bit rigorous. 00:18:40.096 --> 00:18:45.796 And I would say of the fields that I've looked at, I think the area of economics, 00:18:46.636 --> 00:18:53.056 and finance and markets in that world is among the most ideological driven of any. 00:18:53.156 --> 00:18:56.516 And so the problems ultimately there may not be just purely technological. Right. 00:18:56.856 --> 00:19:01.036 But now to get back to the first architecture, right? 00:19:01.096 --> 00:19:06.476 So in some of those Also in the discussions here in this workshop. 00:19:07.456 --> 00:19:10.916 We want to get to some notion of, let's say, brain architecture. 00:19:11.456 --> 00:19:15.536 But then let's first try to understand what we mean with architecture. 00:19:15.856 --> 00:19:22.276 And you have used this analogy with, let's say, a garment to illustrate how 00:19:22.276 --> 00:19:23.256 you think about architecture. 00:19:23.736 --> 00:19:30.056 Could you maybe explain? Yeah, I've been trying to look for an example of an 00:19:30.056 --> 00:19:37.696 architecture that would be so simple you could explain it to a school child, right? 00:19:38.876 --> 00:19:42.816 Because I think we need these concrete metaphors. And of course, 00:19:42.836 --> 00:19:47.596 anything that simple is dangerous because it can't have many of the important 00:19:47.596 --> 00:19:51.956 features that we care about. But the reason I picked on clothing is everybody knows about clothing. 00:19:52.396 --> 00:19:57.096 And the thing about clothing is that it shows two very different kind of layers. 00:19:57.996 --> 00:20:04.136 And the PNAS paper that's now online and soon to be in print discusses that. 00:20:04.136 --> 00:20:08.356 And the idea is that the most familiar one is sort of inner to outer wear. 00:20:08.496 --> 00:20:09.876 So you have an inner layer that's 00:20:09.876 --> 00:20:13.836 close to the skin. And I'm imagining clothing in a cold environment. 00:20:14.016 --> 00:20:20.016 We're naturally adapted to live in a very hot environment and don't really even 00:20:20.016 --> 00:20:21.356 need clothing in a hot environment. 00:20:21.456 --> 00:20:24.776 But where we really need clothing is harsh, cold environments. 00:20:25.216 --> 00:20:29.676 And so in those environments, we often have an inner layer that's close to the skin. Right. 00:20:30.522 --> 00:20:34.562 And then you have an outer layer that protects you against the weather from 00:20:34.562 --> 00:20:36.582 the outside, wet, cold, wind. 00:20:36.982 --> 00:20:41.042 And then a middle layer that provides warmth. And what's interesting about this 00:20:41.042 --> 00:20:43.882 view of layering is it's so simple anybody can understand it. 00:20:44.042 --> 00:20:49.642 What you have is three very different kinds of layers that combine synergistically 00:20:49.642 --> 00:20:54.902 to create a whole that feels good to the skin, protective of the outside, 00:20:55.062 --> 00:20:56.102 and as warm as necessary. 00:20:56.102 --> 00:20:59.262 And tremendous plug 00:20:59.262 --> 00:21:02.962 and play modularity you can swap in different parts 00:21:02.962 --> 00:21:06.002 of the garments but it also shows another 00:21:06.002 --> 00:21:09.022 point which is enormously confusing within 00:21:09.022 --> 00:21:12.202 the scientific community not in engineering but most 00:21:12.202 --> 00:21:15.402 combinations of garments do not produce functional outfits 00:21:15.402 --> 00:21:18.502 they really while there's enormous variability 00:21:18.502 --> 00:21:21.222 in what you can do there's still 00:21:21.222 --> 00:21:24.282 pretty strict strict rules in what can go where or they don't 00:21:24.282 --> 00:21:27.762 work at all so you can't you can't make underwear 00:21:27.762 --> 00:21:30.922 outerwear you can't make socks into hats 00:21:30.922 --> 00:21:34.142 you can't you know for the most part so that's one 00:21:34.142 --> 00:21:42.662 way of looking at the layering and but another is the composition of fiber which 00:21:42.662 --> 00:21:49.542 is then um spun into yarn which is then knit or woven into cloth which is then 00:21:49.542 --> 00:21:53.262 sewn into garments A whole other dimension of layering. 00:21:53.562 --> 00:22:00.782 Very different than the first, but again, illustrates properties both of biology and technology. 00:22:01.062 --> 00:22:07.362 And so, in that PNS paper, I try to flesh out that example as one that I think would help. 00:22:08.449 --> 00:22:13.189 Both engineers and neuroscientists talk to each other because it's a, 00:22:13.229 --> 00:22:14.469 it's such a simple example. 00:22:14.549 --> 00:22:18.629 Now the danger is it's too simple. It doesn't have dynamics. 00:22:19.869 --> 00:22:23.489 The, the, these compositional forms are very, very simple. 00:22:23.569 --> 00:22:27.889 The way clothing just sits on top of, you know, one layer of clothing sits on 00:22:27.889 --> 00:22:33.309 top of another doesn't really, doesn't really lock in like you would in a, 00:22:33.309 --> 00:22:35.829 in, in many other systems. There's a lot of things that are wrong with the thing, 00:22:35.909 --> 00:22:37.029 but it gets a starting point. 00:22:37.469 --> 00:22:41.169 Okay, that's good. So if we have layers, we also have different, 00:22:41.229 --> 00:22:44.969 let's say, processes that give rise to the key properties of these layers. 00:22:45.409 --> 00:22:49.109 But now, if I would like to have, let's say, a more parametric description of 00:22:49.109 --> 00:22:54.409 such an architecture, what are the key, let's say, parameters that I should worry about? 00:22:55.249 --> 00:22:59.529 Well, in this, I mean, again, using this, stretching this probably to the point where we'll break, 00:22:59.529 --> 00:23:10.429 um there is there is the rules by which the constraints by which you can assemble 00:23:10.429 --> 00:23:17.729 so like i said most most combinations of garments do not produce a functional outfit, 00:23:19.149 --> 00:23:27.469 and there's a few rules that you can write down in principle anyway and we certainly 00:23:27.469 --> 00:23:32.769 learn them And children learn them as to what make a functional garment. 00:23:32.929 --> 00:23:37.369 But the interesting thing is, once the garments are selected...