 Computers keep changing the world, but their power and safety is limited by their rigid design. The T2 tile project works for bigger and safer computing using living systems principles. Follow our progress here on T Tuesday updates. This is the 50th T Tuesday update. Let's get into it. Last week we were working on putting together a power zone for grid of 4x4 tiles. It's going to have a single power supply and be isolated by power to the neighboring sets of tiles, but just connect with data. I'll have an update on that. Since then, this past week has been the Santa Fe Institute Workshop in Biological Computation that's going to spend most of the time on. What I would like to say, I see next week, is an update to the cache update protocol. Even if we don't get all the way to intertile events, which is our main goal, what I would like to see is actually having torn into the protocol to start adding additional messages to exchange information that they didn't need to do in MFMS in the simulator. We'll have more about that next week. In education and outreach, the stickers have shipped. It's a tongue twister. The domestic ones should arrive this week. They shipped yesterday morning, Monday morning. The international ones, I don't know. If you get them and you stick them on something, stick a picture up on InstaFace or wherever one sends these pictures, and that would be fun. In the hardware division, yes. Last week I had originally was thinking aluminum, but then started to think about maybe could do it with plastic after all with vertical struts that were sort of thicker to carry the weight and horizontal struts that were kind of just weaker to stabilize things. But that was not proving strong enough as it kind of tore apart the back live in the T Tuesday update last week. So I did some additional designs in particular strengthening up the horizontal direction by actually connecting adjacent guys together in pairs, somewhat based on a suggestion by Andrew Walpole. Eventually I said, well, Jesus, it would be good if I could actually connect the pairs together into sort of an arbitrary length shape. That had the unfortunate effect that I couldn't. So here is an early one of these things. This goes up, this goes down, this holds the weight, but then one of these things meets up with one of these things so that they get a smooth tiling across. And when I do that, though, only one of them fits on the build plate at a time so they're kind of a pain to print. Here's an example. They also sort of look like, you know, Simpsons aliens or native petroglyphs or something like that. But in any event, you put them together and you get it like this. And the problem is, so that's three rows of four. We need the fourth row. The fourth row goes along the bottom there, but if you just try to stick it in without having another row of the verticals, which goes way too long, they don't hang evenly because the ones that are mounted up on the actual feet are up higher. The ones that don't have it are down low. So I went back and I started making fractional pieces. So here is just the top part of what had been the pair. And that was that one I can print four up and then just connect them along the bottom to essentially make kind of like a finishing edge that would not be used if there was going to be another power zone connected up underneath this. You would get rid of these little n-things and just connect them together. But these are just sort of to cap it off. And there it is with the bottom capped off. There are similar problems needed on the sides. So these are little half verticals to provide support from one side. On the left-hand side though, since they tile like shingles from left to right, trying to stick these things in afterwards was kind of a pain. In the end, I had to actually unmount the two of the tiles in order to snap these guys in and then mount it up again. But the long and the short of it is it actually came together pretty well and it's actually pretty strong. So here is the power zone version two. And this guy is actually booted up. It's hard to know what all is going on. They're running completely different versions of the software and they're just sending information back and forth. None of which is the newest, so it's kind of silly. This was really just to test out this new mechanism for hanging and it's really quite a bit stronger. And I think this is going to do well. So that's the hard way to do it. The main event this week though is in the Science Division, the SFI Biological Computation Workshop. It was really a lot of fun. I tweeted, I live tweeted some pictures and comments so I'm not going to say very much about the actual talks today. Some of it is on Twitter. Certainly my favorite talk was from Mike Eleven about how the body remembers shapes in the simple animals like with a planaria and with like lizards. And cut off the tail and graft it on where a leg used to be and the tail will gradually turn into a leg because the shape is encoded in the memory of the body. It's super cool stuff. Crazy stuff to do with planaria and so forth. In addition, there were these 10 minute talk slides, talk slots for Friday. It looks like number one there is empty. That's because I signed up first with a ballpoint pen that doesn't really show. So let's take a look. It was 10 minutes plus two minutes for questions. Here's my talk and they'll come back and have a few moments afterwards. I'll start my timer here. All right. What I want to do is reiterate what I started talking about Wednesday morning when I was introducing myself. I want to propose a completely simple, completely outrageous, but I think actually as good as you can possibly have definition of life, definition of computation and show how that they're related. And then I want to spend time on my own work as well. Here it is. In the end, living systems and computational systems turn out to be the same class of systems. They define each other. They are not two independent things. They define each other in a circularity, a circular definition, which is what makes it completely unsatisfying. But in the end, all definitions are circular. When you look in a dictionary, you can't actually find a brick in the dictionary. You can only find words pointing at other words pointing at other words. So we are always in the land of circular definitions, but usually we don't know it because we have disciplinary boundaries where we take things for granted as axiomatic. Other disciplines take different things as boundaries. So what I want to do here first is shrink the circle down to its limiting case of surgery to see how they all inter-define each other. Okay, so here it goes. What is life? Life is how we describe a physical system that does work to preserve itself. It uses computation to decide what work to do. Okay, well then we need to know what just computation is. What is computation? Computation is how we describe a physical process that selects meaningful actions based on senses. Okay, have we made any progress? Well, we have to know what meaningful is. Meaningful is how we describe a physical action that somehow tends to preserve life. There it is. That's biology, computation, and philosophy all in one. They just go around and around. And normally, you don't have to worry about it because normally you can take some formal notion of meaning, some formal semantics. I mean, computer science struggles terribly when you bring up meaning and they get it off the table as quick as they possibly can. But this, I suggest to you, is really what it's all about. Life does work to preserve itself using computation to decide what work to do. The meaningfulness of a computation derives from the fact that it helps a living system preserve itself. So, look, go ahead. Hold that thought when Artemis gives his presentation. Hold what thought? I just gave it to you. Everybody's holding it. Okay. All right, so we have sampled uses of this, okay? Does this matter at all? Well, it allows us to make progress on a lot of the things that have already come up in this workshop, starting with the fluttering asking ways. Do we want to interpret the fluttering Aspen leaves as actually performing a computation or is it just mere dynamics? And the answer, according to the fundamental circularity, is it's performing a computation if it helps the Aspen in some way. And in fact, I've been educated since then by Melanie and others that, yes, of course, the Aspen leaves are exquisitely evolved to take advantage of even the lightest wind to flutter around because that has great things about modulating how much sun they get, and yada, yada, yada. I don't know. But it helps them. But it helps them in this particular case at the evolutionary level. That's where the actions are being taken to do it. When they're actually out there going, they might as well just be passive dynamics because the work to preserve it was done at the evolutionary level. Similarly, do the planets compute their orbits? That's, you know, the people bring that up all the time in trying to dismiss notions of what is a computation and what isn't. And here we're going to say, well, probably not. Can you tell a story about how the planetary orbits help to preserve a living system? Well, the orbit of the Earth creates the seasons that we decide when to plant. So maybe you could. But probably not. But you see how it works. You can choose. This is a descriptive view. Shining information, same thing. You know, at one level, people say a bit is, you know, the amount of information that allows you to make a prediction better than chance at a certain level like that. And that fits right into this view. A prediction that will help you somehow survive better, enhance your life. But on the other hand, it's channeled information is normally thought of in terms of channel capacity. And that's the case where it goes completely nuts, where static has the most information at all. How do we make that go away? Because static is not helping a living system preserve itself. Therefore, it's not computation. The definition has some bite. All right, moving on. Consequences for computer science, I won't spend too much time on this, but I want to say, computability theory, I'm sorry to say for folks who love it, is evil. Computability theory is about things that can be computed in principle with infinite amount of time, infinite amount of resources, infinite amount of tape. And it's for that exact reason that almost none of the results that come out of computability theory actually have relevance to implementable systems that might make life better. That needs to be distinguished from computational complexity, which could, in fact, say things about this computation versus that computation. But a whole lot of people have gotten mind damaged by computability theory and have led them down unfortunate roads. Okay? I'm going to skip, I'll skip the other ones of these I would love to talk about them, but in the interest of time what do I say we do about this? How can we use this understanding of the relationship between life and computing? Number one, what we need to do is say, this whole idea that hardware guarantees 100% reliability, guarantees 100% repeatable deterministic execution for whatever the duration of the computation is, it's got to go. It doesn't scale, because, in fact, you have to say how big your biggest computation is up front before they design the machine. That's no good. So what we need instead is a new computer architecture that's based on best effort hardware. Hardware that's trying to get the right answer, trying to be repeatable, but it reserves the right to get the answer wrong, and not only that it reserves the right to get the answer wrong in any possible way. There is no probability distribution that you can use to limit the possible errors the hardware may dispense. If you accept best effort hardware, in exchange you are offered indefinite scalability. You could build a computer from here to Pluto if you have space, heat, cooling, and room to lay out the tiles, because that's how you can build an indefinite scalable computer. You build it not by building it once and for all, but you build it by building a unit of computation that can tile space with copies of itself. And then you can just keep on going. You'll be using the computer while you're still building it. There's no global boot time. It's too big. If you can figure out how to write software to build a computational stack that will work on an indefinite scalable computer, you will find the software mechanisms you've used to be like life. They will be doing all of the robustness, the healing, reproducing themselves to consume available resources as well as improving reliability. Dot, dot, dot, dot, dot. So I said about 12 years ago now, what the hell? Let's just do that. Let's just build it. So in 2008 these are our first tile of computation. Each of these is an individual thing that plugs into copies of itself. It was marketed briefly under the name Illuminato X Machina and I wrote the operating system part that didn't pick the name. And it was great. Learned a lot. Learned problems that it had needed to build a new version. At the same time we were starting to say, well, how do you write programs on this spatially distributed computer? The answer was some kind of cellular automata, because they're spatially distributed. You need to be doing stuff all over the place. But it cannot be a synchronous deterministic cellular automata because the hardware doesn't support that. So we started developing these asynchronous failure-prone models of computing doing things like copying and so on and so forth. This is from a 2011 Topics and Operating Systems paper, if you can believe that. I'm running out of time. Also we had to do advocacy. People are so entrenched in digital determinism that focusing on efficiency that we need to do a lot of unlearning. And this is one that we did. Suppose you were just doing sorting. But for whatever reason your comparison routine had a 10% chance of failure. Well then what happens? The failure of the sort, the errors that come out when you use an efficient algorithm, quicksort, mercsort are huge. But when you use an inefficient algorithm that everybody in computer science hates, both sort. It's great! Why? Because both sort compares items redundantly and it only moves them one step at a time. So if it makes a mistake, it doesn't make a very big mistake. It's because they maximize the leverage of each observation that they make in the world. You try to make a single comparison and then move things as far as you care. Which is what makes you efficient and what makes you fail big when you fail. I'm over my time. We've built programming languages to start building up the stack. It's not enough just to say here's my time. We have to say what's the mid-level software? Multiple languages. This is a spatial programming language expressly designed to represent two-dimensional transformations because two-dimensional spatial transformations are a pain in the butt to implement in typical, potential languages. Much more natural to make the programming file itself be a grid and directly specify them. And this is where we are today. We've implemented little cellular membranes that in their own little way like cellular membranes. So this will just go around and oh by the way and so here this is the T2 tile. This is the next generation tile that tiles with copies of itself. Six of them go around it because it makes a brick wall stired and you can see back where I sit there's one running there right now running a very simple basic physics for a benchmark. Thank you so much for listening. So that's it. I didn't include the questions, there were just a couple afterwards because I hadn't really asked permission for video taping, I think. This is the first time I actually tried to use the webcam on the new laptop to capture me while I was capturing the screen. It didn't work terribly. The sound wasn't great. The questions were mostly about boundary cases for the definition of life. Like a pendulum is a pendulum doing work, why shouldn't that count as alive and I'm like well yeah, but it would wear down friction and the funny thing to me was that the theorist guy that I was talking to was kind of irritated that I wouldn't assume away friction because as if that was a small matter it's like you know yeah well okay so imagine you have a living system that doesn't have to eat, that's what it would equivalent to not needing to worry about friction. Okay, but it was a lot of fun, I learned a lot and the fundamental circularity of life, computation and meaning feels pretty good to me, I think there's more mileage in that yet to be explored. Next update will be out in a week, we're getting very close to a year of T Tuesday updates thank you so much for being here, have a good week.