 T Tuesday update What the heck is this? This is the first status report of the T2 tile project, which is what I've been working on a Lot for the last couple of years anyway, and in a more general sense for a decade or more That it's gotten fairly far down the road But there's still a lot left to do and it really you know I'm human. I watch YouTube videos. I like to watch, you know maker videos I think the first guy I ran into was Peter Schreephole, you know, he's great and that led me to Colin Ferd You know, I can build anything, you know, you sit back. Oh, yeah, sure. I could do that No, it's crazy the stuff that these people can do. I'm not a hardware person I'm a computer scientist more of a software person, but nonetheless I too am Building this thing. I'm working on this project. I'm trying to build this new kind of computer architecture That's very different than what the way we're computing now. I think it's important and so both as a forcing function to get me to make regular progress as well as for Advocacy as well as to help spread the word about what the project is and what its goals are and how It's making it seems like it's really a good idea to have regular reports And you know for me the things like you know Martin from Winter Gauton with the Marvel Machine X project You know, I watch that every week and I see his successes and his setbacks and his like Oh, that's really cool or I learned something or I don't think that's gonna work. Oh, it did work whatever But it really helps for people to get involved with people get engaged and to sort of gradually learn what the issues are About the particular engineering tasks that you're trying to accomplish and that you know the Martin's always saying, you know It's not about the machine. It's about the music. I'm not a musician But what I'm going for is a new way to compute that'll be better both more powerful and safer better for society So this is my first weekly report that is truly terrifying the whoops the plan the shape of it is Every week every Tuesday t Tuesday At noon mountain time. That's my goal to get the next status report out a little intro of some sort Recap what happened what actually happened in the past week Put myself on the record to make some predictions some wishes some goals for the coming week And then if in fact anybody actually sort of joins in and wants to follow along And has any Questions or you know sort of you know in the mail bag or whatever to just try to fill in stuff that was missing from before I mean From one point of view it would have been better to start this You know weekly updates two years ago, but on the other hand, you know every movie made in the last thirty years, right? They always come in right before the end right before where it gets exciting and then they do flashbacks for all the rest So that's kind of where we are a ton of decisions have already been made I've built prototype hardware prototype tiles. The goal is can we actually get the hardware finalized Get it manufactured in quantity dozens hundred or so and start using it and go for you know incredible Kick-ass demonstrations of a different way to compute So we'll see so the Q&A will perhaps be flashbacks to why the decisions are made the way they were what the goal is We'll see who knows All right, uh-uh so Let me take let's see Let's try to take like three or four minutes tops to try to just scope out the picture once over lightly So the T2 tile project is about a different way to compute still using digital hardware using all of the Manufacturing and sort of low-level and gate chips all of that stuff that we're already Extremely good at being able to design a manufacturer But to put a completely different architecture on it and the important change is giving up on this idea of hardware Determinism like that. So the software hardware Contract that was made way back when in the beginnings in the 1940s was Hardware is going to take the nasty noisy Analog physical world and clean it all up and provide pure bits pure mathematical Reliable bits that everything step-by-step every input Guaranteed the same output same program same input guaranteed to get the same output It doesn't matter how long it takes the program to run or how much memory the program takes up As long as it fits in the machine the answer has to be the same at the end that is hardware Determinism and you know, it's worth understanding that Almost nothing in reality works that way you ask, you know someone to do something They say oh, yeah, sure. I'll do it and they don't do it or they do it Wrong the world is full of that stuff in fact the idea of Software being able to just assume that it can write something and on step one of a program and it can read it back again on Step 100 billion of the program and it'll still be exactly the same thing is Remarkable and that's part and parcel of the power of traditional digital computing. The problem is it's too amazing It doesn't actually scale up and the we're all in this trap everybody who's a computer person anyway Almost everybody's in the computer person is in this trap of you know making excuses and saying oh, yes It is yes it is but in fact it really isn't I've had lots of arguments with people about lots of discussions because I'm coming out of that same tradition I'm coming out of hardware Reliability is a hardware problem Software's focus is on correctness and efficiency, but we have to get beyond that if we're going to get to the better more powerful And safer world of computing so hardware determinism doesn't scale and it isn't safe I'm not going to spend a lot of time going through the reasons for that right now But the alternative is is you know you say okay? Well, I'll make a computer that allows a little bit of errors and I'll build some sort of software redundancy I don't like that well you can do that and there's people who've done it and there's a fields full of papers research papers That are doing that kind of thing sort of little piecemeal step away from hardware determinism and tolerate a little bit of failure or even Deliberately give up on a certain amount of accuracy to try to do something faster or use less energy that sort of thing What I say we need to do with the heart and soul of the t2 tile project is we need a different assumption To take is our core assumption instead of hardware determinism rather than staying right where we've been and taking little teeny steps away We need to say what should we be heading for and that will allow us to reinvent all of the stuff that we've just sort of left the way it is on behalf of hardware determinism and What I came up with was this idea of indefinite scalability. We should build a computer. We should build a computer architecture That will scale indefinitely you can add more and add more you use it before you build it There's no notion of a sort of global boot time to turn on the whole machine because it's too big The thing is running your ad you're using adding more stuff to it. Some it's always failing If you can begin with an architecture that is indefinitely scalable Then that will force a whole bunch of issues that we sweep under the table about software Taking care of robustness as well as efficiency and correctness and so on and what I say is and I'm pretty well convinced of this at this point is That when you embrace indefinite scalability, so we build a computer architecture I can go from here to Pluto if we want and we'll be using it when you know, it's barely gotten past Jupiter or whatever The way software is going to work the way computer the way computations are going to work in this indefinite scalable hardware Is going to look much more like living systems. They're going to move to get out of the way. They're going to protect themselves They're going to heal. They're going to reproduce For the same reasons that living systems do to preserve themselves to get their jobs done to occupy available resources so I'm suggesting that we should put aside Virtually all of our intuitions all of the stuff that we lean on when we do traditional Programming software engineering and so forth and do a whole new start Why would anybody do that as Carl Sagan said extraordinary claims demand extraordinary evidence. So The project is well heck let's come up with the extraordinary evidence Let's just build it The way you can build a indefinitely scalable machine is by making an individual little tile of Computing that can plug together with more and more of its own kind and there's no absolute address There's no serial number. There's nothing unique about them They're like grains of sand that you can pack together and make a bigger and bigger and bigger sheet of these things And you know in the future we can hope that it'll all be optimized and these individual tiles be Microscopic and you know you might be able to repair your computer by just sort of scooping stuff out of a different machine Packing it into the thing that you need who knows but for now for a demo for research to show the idea Say we'll just take some some cheap Linux board something like a raspberry pi now in the end We've picked a board and it's not the raspberry pi, but we'll get to it Build a custom circuit board to mount the thing on that sort of remaps the IO pins in a way so that the Tile this individual little computing thing can connect to copies of itself just to the nearest neighbors You know east and west and so forth. Maybe stick on some kind of little display so we can see what's going on Do that build it make a hundred of them something like that and Actually just show how it works. Let's make it work. That's the project and you know, so this Stays report the t2 tile project a lot of it is going to be about hardware because that's where we're currently at in the project But the hardware itself is just a step in a larger program the point is to build you know a tabletop full of a hundred of these tiles or more and Actually start being able to use that as a computational resource To build a new kind of software that follows this living computation thing. That's the plan All right taking too much time So where are we? What's the status report? You know, I've been working on this on and off for years in the last week What I spent most of the time doing was actually working on a written report for a guy that's helping me out I don't know. Maybe we'll make it public who knows going forward But just trying to sort of summarize where we all were I also Hacked a bunch on the software that I'm using for these slides It's got a lot of other capabilities that I'm working on as well Why do we need to do that for the t2 tile project? Well, it's not exactly clear how much we do but It's fun and it's also true that I have discovered that in all the years Working on this kind of living computation best effort all this stuff is that it actually affects my software engineering design It affects the way I implement normal stuff And it may turn out to be the case that having The viz system built up may end up being a little useful examples It's all very much close to the the traditional approach It's one of those little bitty steps away, but it might be evocative and in the last week I've been convincing myself. I really have to make these status reports or I really have to make the first one and Put it up there coming up In for next week what I would like to have is an actual little Intro video, you know, I talked for what four or five minutes about the meaning of the project I'd like to boil that down to a one frickin sentence And have some kind of little video bumper that we can put on at the beginning in general As I'm rotating away from taking care of other stuff and focusing on the visualization so forth to get back to Focusing on the t2 itself. I just got to get you know, I've got to get kikad going again hardware stuff Just to refresh my t2 brain Also, we have a new version of the programming language oolong coming out that has Modifications to the simulator to make it work better with these tiles It's time to get oolong four out of there the original intent was to get oolong four out in August that didn't happen and also Make myself be here again so Q&A well, we're taking we've already taken up 15 minutes So we'll leave Q&A for another time if if anybody does want to come along wants to be getting on the ground floor Certainly one thing that would help me is What ideas you know all of this area, you know, there's living there's living computation indefinite scalability best effort robust first All of these different ideas that for me they all lead to the same place their views of you know The blind men and the elephant But for different people with different backgrounds different ones can be helpful and you know I can't tell people say oh summarize it. I said well, I can't summarize it unless I know what you know So I'm really much more of a counterpuncher if people are willing to ask questions. I would love to answer them How do I feel about doing this? That's the question. I'm asking myself to get us started Well, you know, I feel terrified And because you know right fundamentally if you keep the secret if you're doing it just by yourself Then it's kind of like CPU and RAM. It's kind of like serial computing. You can you know You can have your desktop covered with half constructed pieces of project as long as you want as long as nobody else Needs that space and you can work on it whenever you work on it Just like in serial determinism the computer goes however fast it goes and you have to wait until it gets to the end But the instant you become event-driven the instant you start having I owe to the environment where there are other active entities Now it's no longer Serial determinism now there are other agents out there other active components that have their own time constants And you know, you need to do it or you're gonna lose it and so forth So, you know that that's the good and the bad about committing to a weekly or not weekly You know at my age a week feels like it's another week I really wanted to do sort of a monthly report, but I'm saying weekly. We'll see so I'm terrified because really You know a Week is gonna be here before I know it really but on the other hand I need to do it Why do I need to do it because of course? It's emblematic of the process. I'm saying that we need dynamic sort of democracy Interactive computing rather than I am in charge of everything Everybody wait until I'm ready and then we'll run. So here we are The next update will be out Tuesday August Tuesday, October 16th at noon Mountain Time. I Welcome you to join me if you want and we'll see how it goes Thanks for watching