Transcript: Advancing Robot Hardware Innovation through ARIA's New UK Research Initiative

[00:00:00] Introduction to Jenny Read and ARIA

[00:00:00] Audrow Nash: Hi everyone. Audrow here. In this episode, I speak with Jenny Read, who's a program director at the Advanced Research and Innovation Agency, or ARIA for short. ARIA is a new initiative by the UK to fund work and research that's not getting enough funding, but could be really impactful in the long term. They're loosely modeled on the US's Defense Advanced Research Projects Agency, or DARPA, but without the defense angle. As a program director, Jenny has carved out an area that she thinks is ripe for investment. And you guessed it, it's in robotics, mostly around manipulation. To me, this kind of initiative is very exciting because it might bust some plateaus we're seeing in robotics and let us get to some new applications and markets. I think you'll like this interview. If you're into far sighted technologies that could make a big difference, or want to understand some of the challenges and opportunities in the current state of robotics. Also, if you're interested in doing some of this far sighted work. ARIA is in the UK, but is not just for people in the UK. So you might consider applying. All right. With that, let's get to the interview. Hi, Jenny. Would you introduce yourself?

[00:01:17] Jenny Read: Absolutely. I'm Jenny Read, I'm a program director at ARIA, the UK's advanced research and invention agency.

[00:01:25] Audrow Nash: And tell me about ARIA.

[00:01:27] Jenny Read: Yeah. ARIA is quite a new funding organization in the UK. We were founded by an Act of Parliament in January 2023, and our mission is to produce transformative societal benefit to science and technology. So we were kind of new an experiment basically in a new way of funding for the UK. We're loosely modeled on the APRA DARPA model in the US that is widely perceived as having been very successful, and we haven't really had that kind of program driven, mission focused approach to funding in the UK before. And that's what ARIA is trying to do. So they kick off.

[00:02:05] Audrow Nash: Now you say loosely focused. What what aspects are you taking and what aspects are you leaving from the APRA DARPA program.

[00:02:12] Jenny Read: Yeah. Well one important thing that we're leaving from DARPA is the defense aspect. So yeah.

[00:02:18] Audrow Nash: Just ARPA no D

[00:02:20] Jenny Read: Exactly Defense, the military is actually the one area that ARIA won't fund. So that's an important change. And that has various consequences as well. So, you know, DARPA has a sort of built in customer in the form of the Department of Defense or it doesn't. So we have to think hard about how are we going to take the science and tech that we deliver or develop and translate that into societal impact.

[00:02:45] Audrow Nash: And so yeah, it's about tech transfer basically. So you're funding research so that it can you can build companies out of it and it can improve society in general.

[00:02:56] Jenny Read: That's right. So it's an interesting mix because we are not translational in the sense of being super high TRL or close to commercialization. We can fund very low. So technology hazardous level because I.

[00:03:10] Audrow Nash: Would never know that acronym okay. Yeah.

[00:03:13] Jenny Read: Yeah. All right. We we fund work that's very basic and speculative. But importantly, we always have a sense of how we're hoping it's going to deliver benefit for society. So we wouldn't fund or we haven't typically funded work that sort of blue sky in the sense we have no idea how this might benefit. We're like, we have a purpose in mind, and then we're trying to get the knowledge and the technology together that we need to solve the problem we're focusing on. And we have at that moment seven programs, and each program is focusing on a different problem area or challenge that we're aiming to solve.

[00:03:49] Jenny’s Role in ARIA

[00:03:49] Audrow Nash: Okay. Very interesting. Now, would you tell me a bit of your role in all of this? And then I guess specifically about the robotics focus?

[00:04:00] Jenny Read: Absolutely. So I'm a program director, which is ARIA's version of the RPA program managers. And that's an interesting part of our model actually. So one of our principles is we always say we fund people, then projects rather than focusing on the project first. And I applied that approach by starting out not thinking like, oh, okay, here's a particular area of science or a particular challenge that we want to solve, but rather let's recruit the program directors. So they started out and they recruited eight program directors like myself. And then each of us was given the challenge of try and find what ARIA calls an opportunity space. So an area which feels really important, if true, would be hugely important for society, but it's also underserved, so it hasn't received the attention and resources commensurate with its potential importance. And it should also be in some sense, right. So there's a reason why now is the time to dive into that space.

[00:05:01] Audrow Nash: Tell me more about that underserved part of it. Because that's an interesting thing I remember from our previous conversation.

[00:05:07] Jenny Read: Yeah that's right. I think we're trying we're trying to shift the dial. We're trying to find new spaces. We don't want to jump on the bandwagon where there's already lots of attention because, you know, that's great. It's being dealt with. So we wouldn't have much counterfactual impact there. So we're trying to find things that are currently neglected because nobody else will do them for whatever reason, whether it's because it's too risky or it's not the sort of thing where it's easy to capture value. So it's not a attractive investment proposition. And so for those reasons that particular spaces underserved and yet potentially hugely valuable to society.

[00:05:46] Audrow Nash: So you're, you're not going into areas that are receiving a ton of investment from maybe it's like American venture capital or something like this. And you're focusing on areas that they're missing that are also probably high impact. And for one reason or another are not getting funded.

[00:06:04] Jenny Read: Exactly.

[00:06:04] Audrow Nash: And there's probably also, you loo you leave a little bit of like the hype space because certain areas really blow up and the investors pay unbelievable amounts of money for it. And then it's like you won't make much of a contribution there. So you're focusing on really high impact, potentially or preferably ignored areas.

[00:06:24] Jenny Read: That's right. Where can we change the conversation.

[00:06:26] Jenny’s Robotics Focus

[00:06:26] Audrow Nash: Okay. And so where in robotics are you focusing.

[00:06:31] Jenny Read: Yeah. So I when I came into ARIA and start looking around in robotics very much, I was an outsider. My background's actually originally in physics. And then for many years of neuroscience, it's quickly dawned on me that obviously robotics is having a bit of a moment. There's so much potential because AI and so in lots of robotics, I feel it's exactly like you say, there's not a lot of space for AI to make a difference, but it felt to me like there was one area that was being relatively neglected and that's hardware. I was very struck as a as I say, as an outsider to the field by how little robot hardware had advanced really over the past 50 years or even longer. People are saying to me, well, that's because we know the bottleneck is the control and the autonomy, right? You can tell operate existing robots to carry out complex tasks, but we can't yet have them do those same tasks, autonomy autonomously. So that proves that the problem is the control and the autonomy and not the hardware. And I think that has been true, but I doubt it's going to be true. Yeah, it's arguable, but I think it's been yeah, certainly a compelling case can be made. Up till now I think. But my sense is over the next few years, precisely because of this explosion in novel AI techniques, that's going to cease to be true. And there's going to be a new bottleneck, which is the capability of the robot bodies. And again, coming from neuroscience and spending my time studying biological brains and bodies. It was very striking just how crude current robots are compared to our own bodies.

[00:08:12] Areas of Focus Within the Hardware Space

[00:08:12] Audrow Nash: Yeah, simple heuristics get us a long way in robotics. So it's like, yeah, pretty crude. A lot of things. They just kind of work and pretty robustly. Okay. So hardware. Where within hardware is it hardware in general. Is it manufacturing. Like what are you thinking within the hardware space.

[00:08:33] Jenny Read: There's a few areas that I'm concentrating on. So one is actuation. So I feel like at the moment robotics is largely using a DC electric motors as a very mature technology. But it's a very old technology at this point. It's nearly 200 years old. And yeah, and it obviously wasn't designed for robotics. And in many ways, do you see electric motors are not suitable for robotics relativity just today. Yeah. No.

[00:09:00] Audrow Nash: I don't know too much.

[00:09:01] Jenny Read: Yeah. Low to high frequency motion. Whereas what you would typically want in robotics it's high torque and much lower frequency motion. So then oh you're.

[00:09:10] Audrow Nash: Saying that drone motors aren't appropriate for like, I don't know, robot arms or something like that where they spin that at very high speed, low torque. Yeah. These kind of things totally agree.

[00:09:21] Jenny Read: Right. And so then you need gears right to translate into a low to high torque low speed.

[00:09:27] Audrow Nash: And then they reflected inertia and all sorts of problems.

[00:09:30] Jenny Read: Exactly. You've got all this inertia that you didn't want and you're just basically fighting all the way down against the problem caused by the fact that this technology fundamentally isn't suited for the needs.

[00:09:40] Thoughts on High Versus Low Gear Ratios

[00:09:40] Audrow Nash: Well, the the high gear ratio ones, I mean, there are low gear ratio ones that I think are better. Do you do you think so or like the ones there's some companies where they're doing like 1 to 1 or 1 to 4 gear ratios. And so it's very direct. It's high torque. I think I'm not too knowledgeable in this area, but what do you think?

[00:09:59] Jenny Read: Yeah, I think there are a number of different solutions, but none of them are ideal. It won't have problems like efficiency and weight. Back drive ability. And, you know, when you compare it to biological muscle.

[00:10:13] Audrow Nash: That's very different.

[00:10:14] Jenny Read: Yeah. Very different. I'd say one of the things I'm interested in, in trying to develop in the program is fund novel forms of actuation.

[00:10:23] Audrow Nash: Okay. All right. Actuation sounds really cool. I think we'll, we'll I'd like to go through all of the different areas in hardware that you'd like to focus, and then maybe circle back to them, in more detail because I have more questions, definitely for actuation, but I don't want to get bogged down there right now. I want to see the whole thing.

[00:10:41] Jenny Read: Yeah. So another one is sensing, again, I was struck by the relative crudity and poverty of robot senses apart from vision. So typically robots, you cameras. I'm a vision scientist, so you don't need to convince me about the benefits of vision. And we are highly visual creatures, right. And what the consequence of that is that, you know, we've invented painting and photography, and we love looking at pictures. So cameras are ubiquitous to this wall and cheap and available to use in robots. But if you were designing a robot a priori, I don't think you necessarily start with cameras as top priority. And the reason I say that is if you look at biology, there are animals that cope without vision at all. I, cave salamanders with no eyes had so hunt. But I don't think there's a single animal that has that lacks touch sense. The kind of sensation is clearly super important in biology, and I just. It's not like I'm obsessed with robots having to copy biology. But evolution has been exploring the design space for billions of years, and apparently it hasn't found any workable part of that design space that doesn't have really rich tactile sensing.

[00:11:56] Touch Sensing and Robots Being Fragile Due to a Lack of Internal State Awareness

[00:11:56] Audrow Nash: So sensing and specifically touch sensing if you're interested in I think that's really cool. I mean, if we had robots with skin, all of the sensor and sensor information you could get would be amazing. And, in having the robot navigate the world and operate without like, pushing things over and whatever else.

[00:12:15] Jenny Read: Right, right. The other thing, of course, is sensing about its internal state as well, because that's the other thing that surprised me when I entered this is a robotics field, was to learn about how fragile robots are. This is remember visiting a lab and they were showing me a robot that they were trying to train to do household cleaning tasks. But the the lab chief was like, but I mustn't touch it because I probably break it. So I was like, how I going, could this be a problem? And I think presumably one of the reasons robots are fragile is that they lack sensing about their internal state or not perception. So they don't sense when they're doing themselves damage. So they're a little bit like these humans who have some, you know, genetic disorder, which means they don't feel pain. They have to live really careful lives because otherwise they die incredibly young because of all the damage they've done to their body without realizing.

[00:13:05] Audrow Nash: Oh, for sure. Or like, if you lose nerves in your feet and then you have a pebble in your shoe and that really the infection and wounds that you can get there.

[00:13:13] Jenny Read: Exactly.

[00:13:14] Audrow Nash: Okay. So I like those points quite a bit. So sensing so actuation specifically around motors. We're not novel sensing not novel forms of actuation and then sensing around touch and proprioception. Is that what it's called?

[00:13:30] Jenny Read: Yeah. Well, there's proprioception and nociceptors and so forth. Perception is sensing your current state of your body and nociceptors sensing what is being damaged, essentially.

[00:13:40] Focus on Integrating Body Design and Neural Control in Robotics

[00:13:40] Audrow Nash: Oh, cool. I don't know that word. Sweet.

[00:13:42] Jenny Read: Yeah.

[00:13:43] Audrow Nash: Any other focus areas?

[00:13:45] Jenny Read: Yeah. So the other area that we're getting quite a lot of attention to, and again, coming back to this idea of evolution, one of the things that again struck me about robotics is how little attention has been given to things like morphological computing. The advantages you can get by making control more efficient by designing the body appropriately. And that's again, something that I think we see very commonly in biology. And it's because in evolution, brains and bodies obviously evolve together in lockstep. Where is my sense to that in robotics? Yeah. Often you get a mechanical engineer designing a robot. Yeah. Choosing the actuators, choosing the limb lengths and so on. And then they sort of hand it over to a computer scientist or a software engineer or go, go figure out how to control this thing. And it's always a bit like it might be of like the Frankenstein movie, you know, you've got this body that is like, bring it to life. Yeah. And then looking back, even I realize I actually do this. I can trace this back in sort of Western tradition is basically the book of Genesis, right? God forms man from clay and then breathes life in. And obviously that's not at all how we think of evolution having happened. And I think that the tight integration between the control and bodies in evolution is why we see this incredible efficiency. So that's how you can get something like a B with like a million neurons doing all these amazing.

[00:15:17] Audrow Nash: Incredible.

[00:15:17] Jenny Read: Things. Right? Yeah. Flight obviously of all these sort of decisions and memory and communication like in which direction is the best honey and so on. And I think part of the reason they're able to do that is it just needs to spend so many fewer computational resources compensating for the inadequacies of their bodies.

[00:15:37] Audrow Nash: Yeah. Yeah. They're just driven by impulses in a sense that are highly efficient and very appropriate generally to their environment.

[00:15:44] Jenny Read: Right. Exactly. So when I see people saying, oh, look, the hardware is fine, because, you know, I can tell I operate a robot to do this. It's kind of like, well, yes, you can, but you're throwing an entire human brain at that problem, right? And the human brain is a lot of computational resources, power. And what you're achieving could be done by a mouse brain. If that robot brain is deeply integrated into a mouse body, so can we find ways of achieving that kind of efficiency gain in robotics? A few people have looked at this, right? Because obviously you can't actually evolve robots like one that it reproduce. And two, we haven't got billions of years to play around with. So you have to do it in simulation. And people have obviously tried to explore similar ideas, but I think up till now you just haven't been able to simulate things of sufficient complexity to really make any progress. And that's why up till now, actually the best deal was to have a smart mechanical engineer design your robot body. But maybe we're reaching the point now where that's no longer true, and maybe we can have we can experiment with generative AI exploring in simulated worlds. You know, suppose this control strategy gets integrated with this particular body design. Can we discover unexpected efficiencies?

[00:16:59] Audrow Nash: Yeah, that kind of thing is really cool. Yeah. I remember several years ago looking into work like this and it's super cool. And in theory, it's amazing. I just haven't seen that compelling of examples recently, but I haven't been in the academic space for quite a while. So maybe that kind of thing has happened, or the results that you achieve might be amazing in simulation, but they're very hard to generalize to reality because they overfit the simulator. And the simulator has all sorts of shorthands for complex physics phenomenon.

[00:17:29] Jenny Read: Right, exactly. So clearly, a big part of this effort is going to be improving our physics based simulation and developing techniques for closing this into real gap, for sure.

[00:17:39] Audrow Nash: Yeah. Very cool. Okay. Do you have any more focus? Because that already seems like a lot to me.

[00:17:43] Jenny Read: Yeah, that pretty much covers it. But yeah, you're right. It's a lot.

[00:17:47] Audrow Nash: Okay, so I hear novel actuation, sensing that's around like touch sensing and then also like state sensing where they're like, am I in pain or something like that so they can realize their state and not the word for it. But then, and then morphological computing, which you said, and that's like a tighter integration between hardware design and control strategy. Probably with a lot of leaning on simulation. Is that right?

[00:18:16] Jenny Read: Correct. Yeah. I should say we even though nociceptors and proprioception I think are important, we haven't actually we're not funding any work in that area just based on.

[00:18:25] Audrow Nash: Yeah, that seems very hard to me. Yeah. But I don't know the technology state in that area too.

[00:18:31] Jenny Read: Like, okay. Well, so I just I think proprioception does come in with some of the novel artificial muscles, which can, report on their own state effectively. Yeah.

[00:18:43] Audrow Nash: And if you have encoders and, motors, you can tell where the arm is.

[00:18:46] Jenny Read: Right.

[00:18:47] Jenny’s Background and Journey Into the Field, Including the Funding Approach and Involvement

[00:18:47] Audrow Nash: So we have proprioception or however you say it, just to a decent level. But we don't know if like, you don't have skin so you can't tell if you're bumping things and a lot of it's more open loop and you can observe at the motors if they're bumping things because they'll, you'll generate a force on them and they can observe that, but it's still not as good. And especially with the high gear ratios, you might clobber stuff before you realize that you have indeed clobbered it. Right? Okay, so tell me about so I get the areas you're focusing. That's really awesome. Your background in neuroscience and biology. Tell me how you're here and how you are. Like, I guess I'd love to hear your background and how you got involved with all this. And you mentioned they were funding people, not programs, per se. Tell me how you started with all this.

[00:19:42] Jenny Read: Yeah. So as I mentioned, I started out in physics. I actually did a doctorate in theoretical astrophysics, but after that I moved into computational neuroscience. I started modeling synaptic transmission. And then I got into vision science. And I spent many years studying stereoscopic vision. So how do you get a sense of depth? By combining the images in your two eyes, which is it sounds kind of niche, but it's actually a really nice model system for studying perception and understanding. Yeah, right.

[00:20:13] Audrow Nash: How is it I mean, robotics, it's like we use it all the time. Stereo vision for depth, cheap cameras. You just need processing and you can get a lot of information, very rich information about the world.

[00:20:24] Jenny Read: Right? Right. Exactly. And it's very nice. Question for perception as well. Yeah. How is it that electrical activity in neurons can give you a sense of depth? So I was studying stereoscopic vision in humans, and other primates for many years. And then my career took a bit of a swerve because I discovered that insects have got stereotypes. Just to a chance conversation in the coffee room, a colleague was mentioning to me that praying mantis have got stereoscopic vision. That was news to me at the time. And it really surprised me, because within the field, there's a sense that stereoscopic vision is something really expensive. Exactly what you told us told the correspondents from.

[00:21:04] Audrow Nash: Their little head in the praying mantis has the stereoscopic vision.

[00:21:08] Jenny Read: Exactly. I'm like, wait, hang on. Insects got a stereoscopic vision. So I applied for funding to study insect stereo is to understand, like, how are they doing this? Is it similar to human stereoscopic vision or is it different? Long story short, it's kind of similar and different. But that was a really interesting, research program. But it also got me thinking about robotics for the first time. And I think that's because insects are so simple. It's kind of natural to think of them as little robots. So I thought, yeah, right. So I was trying to reproduce whatever stereo algorithm was going on in the head of the praying mantis.

[00:21:48] Audrow Nash: Then that that's really cool. Yeah, yeah. I see how that lends to robotics.

[00:21:52] Selecting Focus Areas, Funding, and Logistics of the Program

[00:21:52] Jenny Read: That's right. You start to have to think about the motor behavior as well, because, I mean, one of the things that it's clear is that they that's the story vision I think works best in the center of their vision. So, you know, they have to detect something and then move their head towards it. So now you really have to focus. Yeah. You have to put the close the loop. Right. You have to put the motor behavior in when you think about the sensory input. So yeah I was thinking about robotics from that perspective other than an idea came along and we have recruited for a program directors and made it clear that you could apply to work on something that was outside your existing area of expertise. I just thought, well, it's probably crazy, but why don't I pitch to them to come and do something on robotics?

[00:22:34] Audrow Nash: Very cool. Okay. And then so how does this all so you have your focus areas. How do you go about selecting like what were you looking for. How much funding and support are these companies or groups getting? How does all this work? Kind of from a logistics of the program perspective.

[00:22:53] Jenny Read: Yeah. So the first thing to say is it varies a lot from program to program. Because program directors have quite a lot of freedom to set things up how they think is best going to advance their particular program.

[00:23:04] Audrow Nash: Did how are you doing it? Yeah.

[00:23:07] Jenny Read: So in my case, I had a budget of around 50 million pounds. I in my original call for proposals, I said, you know, the maximum 20 individual project. 20 million. Actually, I haven't I'm not planning to make any such large awards. I've probably made more smaller awards, but, you know, fewer, larger ones, do. The thing about my program is extremely collaborative. So a lot of the creators, as we call them, the, people who are actually going to be doing the research, even if they applied separately, going to work with each other. And that's. Yeah, that's something that I certainly feel very positive about. But I was struck by what an appetite there was. From the community for that. And I think that's one of the things that attracts me about the idea of sort of synergy and, you know, the whole be more the sum of the parts because you can work together collectively on this common goal and the co-design area that I mentioned, whether you're trying to optimize the design of a body and also the control, that's a great area where I have a whole bunch of different creators, taking particular skills in areas like probabilistic programing or coziness into real gap or Bayesian optimization or the nitty gritty of actually building robot hardware. So they're all coming together with these different lenses and perspectives. And I think that's going to be super helpful.

[00:24:41] Audrow Nash: Yeah, I think so too. Well, that's really cool when you're saying that they work together, are they working together inside of your program or across the whole area program? Is it like oh go ahead.

[00:24:54] Jenny Read: It's within a program. So they're working together.

[00:24:58] Audrow Nash: So maybe like sharing code, sharing resources, sharing know how all towards the advancement of all the goals that are okay. I like that a lot.

[00:25:08] Jenny Read: I like that I should say it's not mandatory. So there are some creators within the program who, you know, maybe they're a company. They don't want to share anything, because of their, you know, commercialization goals. That's fine too. But I'm happy that there are many creators who are very open about sharing code, sharing ideas, sharing approaches.

[00:25:27] Audrow Nash: Oh, yeah. I mean, especially I think when you go deal with problems that are as hard as these, there ends up being a lot of boilerplate, like there ends up being a lot of things that are undifferentiated, and don't, don't allow you to test your idea or like, they're just necessary steps on the way there and like, might as well share all those. And so then, so you have companies and research labs or like, who are the people that are applying for some chunk of your 50 million pounds?

[00:25:59] Jenny Read: Yeah, all of the above. So we're very open as to who we fund. Big companies, small companies, academic research labs, even individuals, everyone's eligible, we can also fund outside the UK everything within a set of programs. So. Yeah. And in my case as well, I don't think I've actually, ended up selecting any individuals who have had it as part of the program. But apart from that, pretty much every other type of entity.

[00:26:26] Audrow Nash: That's cool. So you mentioned you can fund people outside of the UK. Oh, actually, before we getting to that, where are you guys now in the process? Like do we have I think you were saying before that you have a few people or you have groups, but you're finishing up negotiating or where are you now?

[00:26:46] Jenny Read: That's right. We haven't yet signed any contracts for the main program funding. We're in the process of negotiating the details and hope to sign our first contracts before Christmas.

[00:26:58] Audrow Nash: That's awesome. How many? Roughly how many entities? Like companies, research labs. Are you? Possibly funding.

[00:27:06] Jenny Read: About 26 for the program.

[00:27:09] Audrow Nash: That's awesome. And they run the gamut of the areas that you said actuation, sensing and morphological computing.

[00:27:18] Jenny Read: That's right. And some of them are also what I called challenge specification creators. So I was keen to get people involved who are not in robotics themselves, but have some problem that they would like robots to solve so that they could work with the creators. Yeah, and I should say, I think I haven't mentioned yet that within robot hardware, the focus of the program is specifically on dexterity. So we're interested in people who have dexterous tasks, like, for example, you know, assembly in manufacturing, the sort of tasks that still have to be done by human beings. What would it take to have a robot be able to do those tasks?

[00:27:56] the Dexterity Challenge in Robotics, Its Importance, and the Focus of Investment

[00:27:56] Audrow Nash: I like that a lot. You want to tell me a bit more about the dexterity challenge and like what skills I get the importance of it. And I also think that it's an area in robotics where there needs to be a lot of work. And you see a lot of companies working on this. This actually from my perspective, this seems like a much hotter area or more more money and more investment and more attention all over, is going into this area, but I suppose maybe not as much on the hardware. Like you still see two finger grippers. But tell me about this area a little bit more like.

[00:28:30] Jenny Read: Well, I think, you know, just as you say, dexterity is just so important because basically anything we would want robots to do for us is going to involve the precise application of forces to objects in the world, which is essentially dexterous manipulation. You know, we're seeing robots start to be used, you know, to inspect factories and so on. But I think that their use is going to be so limited until they can really manipulate objects in a way comparable to human beings. I mean, not a new point, obviously, as you say, that's why there's so much tension going into it. But my sense is this is an area where the hardware really is going to prove limiting. And the tactile sensing, for example, is going to be really key here, but also the actuation. Right. And it's just hard to have dexterous fingers if each one is being driven by an electric motor.

[00:29:23] Audrow Nash: Interesting. Yeah. So this is and so what it sounds like to me is this is kind of a everything, combines around this area of the population because I can see how hardware might be a limit and your sensors might be a limit, and then your ability to design hardware in tandem with how it's being controlled would be a limit. And simulation would be a limit, right? This is kind of like all of those points. They all focus under manipulation in some sense.

[00:29:55] Jenny Read: Yeah. Yeah, exactly.

[00:29:58] Audrow Nash: Okay. Yeah, that seems really cool. One thing that's interesting to me is, I've talked to a lot of companies that use, like the two finger grippers or the solution grippers and they're like, yeah, we tried a bunch of things, but these are just the best working ones kind of thing. Where do you imagine there's a lot of room for improvement? I'm sure there is. But, what are kind of what are the things that you're excited about trying or I don't know, where do you see there being like, low hanging fruit perhaps in this area?

[00:30:27] Jenny Read: Yeah. I hesitate to describe any fruit as low hanging in this area, but I think I already mentioned middle hanging. Yeah, I already mentioned.

[00:30:35] Audrow Nash: Lowest possible hanging.

[00:30:36] Jenny Read: Yeah. Assembly tasks. So, the sort of things where you're trying to clip something on or maybe stretch and pull over. So. Right. You may have to handle, deformable materials. You may want to, you know, it's plugging in cables in the cable harness. I think all of those things, it's really going to be challenging to do with a pincer gripper.

[00:31:01] Audrow Nash: Yes.

[00:31:02] Jenny Read: Okay. Disassembly using.

[00:31:04] Audrow Nash: Mostly on deformable materials within manipulation space because that that is something I don't see any good. I haven't I mean, I'm not terribly in the space, but, I haven't seen any compelling, like, solutions for deformable objects. Like, that's a very hard area for sure.

[00:31:25] Jenny Read: What's interesting to hear? I wouldn't say we're focusing on it specifically, but it's definitely something we're aiming to address as part of this. For example, taking care that the simulation techniques we develop can hand deformable objects. Yeah.

[00:31:39] Audrow Nash: Yeah, that's really hard. Like it seems like people work really hard to develop very good simulation models for deformable things. And if they work well, they're very slow and then that doesn't. That's kind of the opposite of what you need if you want to train things on it. So say you want like a reinforcement learning circuit in there. Then that really limit to you.

[00:32:00] Jenny Read: Yeah.

[00:32:01] Audrow Nash: Okay. Yeah. Well that's really cool. Okay. So we have those big areas, and we have 26 companies that you're funding and they're getting.

[00:32:10] Jenny Read: Some part of what the North.

[00:32:15] Audrow Nash: Companies and labs, just organizations in general, that you're funding. What's the timeline for this, or are you giving away all of the 50 million right at once, or is it, over several years, you start to give more of it, or how does this all work?

[00:32:29] Jenny Read: Yeah, we're basically committing it, all at once. These, each creator has a particular timeline, typically, but ranging from about two to about four years for this program. So we'll be spending the money over that period. But we're allocate doing it now.

[00:32:47] Audrow Nash: That's awesome. Okay. And so if a it with your allocation how do you like one of the things that venture capitalists do that I think is kind of clever is they fund a bunch of startups a little bit, or sometimes they'll fund a bunch of startups a little bit later rounds that make it different, especially early ones. They'll fund a bunch a little bit and then they'll reinvest potentially later as, traction is being shown with the startups. Will you be doing anything similar or is it all allocated right away or how do you think of that?

[00:33:19] ARIA’s Allocation Strategy

[00:33:19] Jenny Read: We have a few where we are. We've got a couple we were explicitly allocating. Exactly. So either we have a go no go point where we'll decide whether to run dog in most cases we have a we're making the allocation. Yeah. Upfront I should say all our funding. We remain very involved in the delivery of the projects and we'll be discussing with creators, you know, whether things are going well, whether we need to pivot, worst case, potentially terminating if it's turning out things or not working quite well. Yeah. Which is obviously to be expected. Right. It's high risk. Oh yeah. So you know you build in well let's try this and put a milestone here and see if we reached that point. So depending how things pan out then if we've decided not to pursue one particular project that potentially frees up resources to then double down on another project with. But we're somewhat limited in our ability to, you know, if we have a large amount of follow on funding within this program because each program has a particular budget, what it will say is, I think looking at offer to go back to what we can learn from that. One of the things that's impressed me about the offer is how they have a great record of pursuing ideas over a very long time scale, over decades. I would love to see how you're doing that. So even though program directors tenure is limited, so was appointed for initially three years, it will be a maximum of five years and then I'll move on. But maybe another program to will come into this space and then they can double down. Yeah.

[00:34:56] Audrow Nash: Interesting, Is that a common thing with RPA? I mean, I've talked with like I talked with Tim Chung a while ago who was, DARPA program director. He did the sub t subterranean challenge, a bunch of robots in caves and underground environments. And he is no longer there, so, I don't know, maybe their tenure is not that long.

[00:35:17] Jenny Read: For this kind funding. Yeah, that that's a key part of the model.

[00:35:21] Audrow Nash: The churn. So you keep getting fresh ideas and that's pretty cool.

[00:35:24] Jenny Read: Yeah, I think it's cool.

[00:35:26] Audrow Nash: So you have these companies they're funded for 2 to 4. I keep saying companies you have these organizations, they're funded for 2 to 4 years. They're doing these high risk things. Now, you said they're not all in the UK or it's not a requirement that you fund companies that go against organizations in the UK. How are you working with global communities or like global organizations that may want to apply for this kind of thing? And what requirements do you have? Like do they have to move to the UK or how does it work.

[00:35:57] Jenny Read: That they demonstrate UK benefit? Which could be through offering something really essential to the program? We certainly encourage them to consider moving to the UK or setting up an office in the UK.

[00:36:12] Audrow Nash: Or is it anywhere in the UK or is it like, do you guys have a hub in one location?

[00:36:17] Jenny Read: No, I mean, we have our headquarters in London, but we, in many ways I would actually say is because, you know.

[00:36:24] Audrow Nash: Space constraints and things.

[00:36:26] Jenny Read: Just like, you know, geographic diversity, we value diversity of any axes. So if you want to come to the UK but not London, then that's great. You know. So yeah, it's about UK benefit not being in any particular place

[00:36:39] Audrow Nash: Do you have any like what. Do you have any examples of company, organizations that you're considering that are not in the UK? And what kind of problem space are they in?

[00:36:49] Jenny Read: Yeah. So one company was a nonprofit is basis. It's. Yeah, it's a nonprofit applied research organization studying intelligence. It's currently, in New York, but, it's following the, are the funding, which, as I say, when you negotiate in that funding to set up an office in the UK and they'll be working with other creators on this co-design optimization challenge.

[00:37:18] Audrow Nash: Okay.

[00:37:19] Jenny Read: That really cool. Yeah, yeah. Go ahead. Because another example is, a company in Denmark, plastics, startup that's developing, electro elastic actuators for soft motion. Cool. Yeah. So it'll also be working with the UK creators and enabling UK creators to have early access to prototypes.

[00:37:45] Audrow Nash: Very cool. Let's see. So what are some other examples of organizations that you have that are in the negotiations I suppose. Like what are some promising ones. What are you. I mean I'm sure they're all promising, you know. What are some I don't know, it's it's kind of spanned the space of different organizations you're planning to work with. I'd love to hear some examples.

[00:38:10] Jenny Read: Yeah. So one is I could name a small company d z p technologies, founded by materials chemist called Slacker Stover. And they're innovating in printed electronics. So they're looking to create new components, robotics such as stretchable and flexible and tronic components.

[00:38:30] Audrow Nash: Oh, that'd be really nice.

[00:38:31] Jenny Read: Right. Exactly. No need to try to build skin on robots then. Those kind of components be handy. So that's a nice example. I think of. One of the things I wanted to do with this program was bring people from outside robotics to work on the problem, and that's a good example. I think. Another one we've got. So it's some academics also working on the co-design, problem. So we have a johns at Imperial College London and they the Bristol are collaborating closely together, on that challenge, looking to develop new techniques for optimizing, robot bodies and control policies and then bridging the SIM to VR gap by covertly prototyping robot manipulators, you know, seeing how well, how closely they performed what was expected. Cool. Yeah. So they'll be working with basis and with Quaker and Nova. It's a researcher at Cambridge. Again, all bringing their different skills and unique talents to bear on different aspects of this challenge. So there's the company I could mention, it's wave drives. So that's a, small company near Bristol, and they're developing a really interesting non-contact magnetic magnetic transmission linear actuator.

[00:39:51] Audrow Nash: That's going to take me a minute to unpack. Non-contact. What is that? Say it again.

[00:39:55] Jenny Read: So it's a linear actuator. Think like a piston sliding in and out, but it's not actually in contact with the cylinder around it. Oh they tell the.

[00:40:04] Audrow Nash: Magnets to do that.

[00:40:05] Jenny Read: And Yeah. That's right. So there are the davantage of that. There is there isn't friction. There isn't where there isn't a need for lubrication. So it's.

[00:40:14] Audrow Nash: Yeah, interesting sounds finicky to me, but it could be very, very useful to. Yeah. No friction and all that.

[00:40:23] Jenny Read: It's likely they'd be developing their technology in areas like aerospace and prosthetics. And obviously for this program they're developing it for robot dexterity specifically.

[00:40:34] Audrow Nash: What a cool thing I can imagine. Like, in like sci fi movies and things, when you have like a floating arm that you can control for all this thing because magnets are holding it up, like that kind of thing would be super, super cool. I wonder if we'll ever get to something like that. Okay, these are great. And any other ones you want to mention?

[00:40:51] Jenny Read: Well, you were talking about magnets and flow too. It reminded me of a project that's actually not part of the dexterity program itself, but part of our seed funding. So another thing we do, oh yeah, is that within the wider opportunity space, in my case, robot hardware, we can make small seed awards, to explore really interesting ideas. And one of the seed awards we made, is it again interesting to bring people from outside robotics? So this is a chap called Christopher Wool. He's actually a cardiologist. Yeah. And he's founded a company called Colored Cardiology Devices. And he's looking at, developing a ventricular assist device to aid patients with heart failure. But it's talk to.

[00:41:34] Audrow Nash: A catheter or what kind of things?

[00:41:37] Jenny Read: No, I don't believe so. I think just to aid the blood flow. Although. Don't kiss me too closely. Oh, yeah. Exactly. Cool. So based on that, he got interested in motors, right? Because he needs a motor for that purpose. And so he applied for seed funding for ARIA to develop, a bearing less motor. So, yeah, rather than resting or bearing C devotes is essentially suspended in a magnetic field. So again the is a rotary motor this time. But the same idea that you don't a cool. Yeah.

[00:42:11] Audrow Nash: Yeah I can't wait till that stuff works. Yeah. No friction nowhere.

[00:42:15] Jenny Read: Right.

[00:42:15] Audrow Nash: It's bearings not seizing up at some point. That would be so wonderful.

[00:42:19] Jenny Read: Exactly.

[00:42:20] Audrow Nash: Gosh, we're getting into the future quickly.

[00:42:22] Jenny Read: Is it? And so you can see the strategies that we've got a kind of portfolio of actuation techniques, some of which are different forms of motors and some of which are completely different technologies. Like the electrode, active isolators or the Hazel actuators rules of funding. A US company called Artemus that's developing, hazel hydraulically amplified soft electrostatic actuators. So we're kind of spreading out batsmen. So a lot of different actuation technologies. Cool. Very cool. And ensure roboticists in the future have more components to choose from.

[00:42:59] Bottlenecks in Robotics Development, Including Manufacturing Capacity and Frictionless Motor Innovation

[00:42:59] Audrow Nash: That would be awesome. How do you. So one of the things that's a big bottleneck because so this sounds like it's very wonderful. And future focused for this I view a massive bottleneck being like you come up with an amazing way to do motors that are frictionless. So, it's still like manufacturing capacity is a tremendous difficulty. Like a lot of the research stuff, like maybe it'll work once, but it's like way too expensive to manufacture or it's just hard to manufacture, especially with, like, aging demographics and all sorts of, like, everything getting more expensive in general. What, do you have any. Are you guys maybe it's out of focus, but what are your thoughts around potentially trying to help the manufacturing bottleneck that some of these technologies may encounter?

[00:43:53] Jenny Read: Yeah. You right. It's definitely something we need to think about. I think to be honest, that probably when you really get close to manufacturing at scale, that probably tends to require more resources that we're likely to be able to make it very true. Right. So we would be aiming to get to the point where it's attractive enough for other people to push it. Yeah, but we're definitely thinking about how can we demonstrate the portability of that. Yeah. So we're working, for example, with the Manufacturing Technology Center in the UK. Oh, which basically exists for this kind of reason, to try and help companies figure out how to manufacture what they need.

[00:44:27] Audrow Nash: Okay. I'm pretty ignorant how how significant like, does the UK do a lot of manufacturing or does it? I don't know much I suppose.

[00:44:37] Jenny Read: Yeah, no we do. We have quite a strong manufacturing sector actually. Yep.

[00:44:41] Audrow Nash: Yeah. Awesome. Okay. So those are all great examples of different companies and initiatives. It's cool to hear that you guys. I mean, you're leveraging and thinking about how it goes once it gets out of kind of the research and early stages. And then are actually are any of the technologies more mature or are they all very high risk, very early?

[00:45:06] Jenny Read: Oh they vary. So yeah, they're not all very high risk. If you say Artemus, the company making haze like traitors, they already have some products out. And the what they're doing is, I would say, you know, towards the close to market and, then, you know, we have Tawfik Hassan that Cambridge is developing an electronic skin and that's very early. So this is a Cambridge University and that's very lab based. So yeah, we fund the range.

[00:45:38] Measuring Success in Initiatives and Using It To Justify Future Programs

[00:45:38] Audrow Nash: Oh, yeah. Okay. So for all of this, how do you measure success? Like how do you know if the, batch of organizations has gone well or how can we see if it's, the whole initiative was a success? And also, I'm assuming then we use that idea of success to justify future, programs. And continuing in this vein, maybe. Or how do you think about success?

[00:46:05] Jenny Read: Yeah, that's a really tough one. Right? It's famously hard to assess success. I think it becomes clearer with hindsight. So I think, yeah, the thing we say about, yeah, is we should be able in a generation to look back and it just be completely obvious. How are we a funded research has, transformed society and benefited the UK. And yeah, I get back to DARPA. But you look at that, I sort of point to the internet or many vaccines and so on. In the short term, obviously, we're looking for signals that we're heading in that kind of direction. So I guess it's have demonstrating new capabilities. New products that we can say. Yeah, you could see that they are we are funding was critical in getting them there.

[00:46:55] Audrow Nash: Awesome. Yeah. It is a hard thing to measure with all these especially in the really short term. But I think it'd be really cool to see some major things you can point to in a generation in this kind of thing.

[00:47:07] Jenny Read: Absolutely. And in the shorter term I think things just like new communities for Asians, potentially no new interest in robotic hardware.

[00:47:15] Audrow Nash: Definitely. Yeah. I think one of the things that's really exciting about what you're doing to me is that you're getting people that are not involved in robotics traditionally into this. And so what it's doing, in my opinion, or from my perspective, would be getting them exposed to this world that I think is going to be very impactful in the future. And so we can have more hands helping out in robotics in general from all of this.

[00:47:39] Jenny Read: But yeah. Exactly.

[00:47:41] Audrow Nash: How does so you mentioned at the beginning that the this is already the full program or is it like you mentioned it started in 2023, this initiative, is it guaranteed to continue. Like is it a commitment from the government to continue for decades, like you're saying, measure in a generation? Or how are we, thinking about the future with all this?

[00:48:07] Jenny Read: So are we was set up for ten years. Pilot initial to parliament. And then it'll be reviewed. Obviously, we are hoping that by that ten year point, there will be enough to point to the government. Realize this is really the case.

[00:48:21] Audrow Nash: Make it bigger.

[00:48:22] Jenny Read: Make it bigger, make them carry on. But at this point, it's not a ten year commitment.

[00:48:28] Key Takeaways and Next Steps for the Audience

[00:48:28] Audrow Nash: Okay. Let's see. And then what are you so thinking about the listener. What do you hope they take away from this. Or is there any like like should they join the next batch or like what are kind of calls to action for anyone who's listening to this?

[00:48:50] Jenny Read: That's a great point. I think part of what I would like to take away is that they've now heard about ARIA, which maybe they hadn't in the past, and maybe go on our website and sign up for updates. We have, as I mentioned, seven programs running at the moment. Some closed calls are closed, some are open. We've also recruiting our second cohort of program directors, who'll be starting early in the new year. So there'll be a second round of programs coming out. So yeah, if you sign up for updates, you'll find out what those are and, you know, potentially want to apply.

[00:49:24] Audrow Nash: Yeah, I'll put a link if it's out. When we published this in the not so distant future, I'll put a link, so that people can find it more easily to sign up, but then also, maybe to the submission, because that'd be really exciting to have more people in the innovation space in general.

[00:49:40] Jenny Read: I think that'd be great. And I think for my program specifically, I would say, you know, please get back to me with any feedback about what you've heard, if you have any, like, hardware needs, if you're excited about any of the work that I'm funding that I've mentioned, please get in touch. We aim to have some workshops and dissemination activities to talk to people about what we're doing, and obviously will be super keen to get any new components or actuators or sensors or whatever out and try it out and test for free flow of interest in them. So, you know, please get in touch if that's you.

[00:50:17] Audrow Nash: Yeah, that would be awesome. Do you have like so from your perspective, three years or five years when your tenure is over for the ARIA program? What do you, what would be like success, in your opinion? Like, what do you hope you've gotten to in the time?

[00:50:39] Jenny Read: Great question. I think it's going to I'm really hoping that the creators that I'm funding are going to achieve something amazing, and I'll just be so excited if even half of what they've proposed comes off. Oh, yeah.

[00:50:55] Audrow Nash: Yeah, yeah, that would be awesome.

[00:50:58] Jenny Read: I think if we could have them. Yeah. I think, if we could demonstrate the value of the co-design approach, demonstrate that, yes, we have actually reached the point where simulation is powerful enough that this is an entirely new way of designing robots and controlling them. I think that would be amazing. If robots are going around with tactile skins that were developed, funded by ARIA, that would be awesome.

[00:51:24] Audrow Nash: Yes. Skins would be a big value add, I think, especially putting robots in the home and things like this and simulation. I feel like if you made significant contributions to the simulation space, that would just be that would open up so much. I think simulation is a real bottleneck in my opinion.

[00:51:40] Jenny Read: Interested.

[00:51:41] Audrow Nash: So anyways, Jenny it was wonderful talking with you and thank you for telling me about the ARIA program.

[00:51:46] Jenny Read: Great. Thank you. It's been an absolute pleasure.

[00:51:49] Outro

[00:51:49] Audrow Nash: Bye everyone. That's all for my conversation with Jenny Read. I hope you enjoyed and see you next time.