 I guess we're there sweet Well Welcome anyone who's watching My name is hey Andy and I also go by Nate With us today. We've got Dave Alan and Andrew The purpose of today just gonna kind of shoot the breeze about the project sort of in response to today's Christmas update kind of think about you know What it what are what are we here for you know kind of just thinking Well, we'll see how serious or casual it ends up being but I was thinking if if you guys were Open to it. We could just go around do quick introductions Like like I said, my name is Nate I'm a programmer. I Think I came across Dave's videos Searching something about artificial life or I didn't know I think you invented the term indefinite scalability So that wasn't what I searched. I must it might have been something about like genetic algorithm or something But I kind of hung around because it sort of piques my interest the same way something like Bitcoin does where You know, there's no Central authority So yeah, I'll pass it over to do you want to go Andrew? Sure So my name is Andrew. I am developer web developer also designer kind of dabble in everything I Saw Dave talk actually a couple of years ago at UCSD I'm an adjunct adjunct professor at the extension school and kind of on a whim saw his talk the advertised and and Walked down and and checked it out and and kind of got really interested in in what he had to say And so I've been following ever since and eating up all the the Propaganda and and wanting to create my own so loving this chance to Sweet now Hi, my name is Alan. I'm an engineer by trade. I really came to this Through YouTube actually, I think I was looking at artificial life as well At that time I might have stumbled on a video and from there I think Yeah, something about the the topic really resonated with me There seem to be finally another an answer to a question I had about you know, how These systems that we're creating how how are we going to escape this trap of of this system is fundamentally Unsecurable the system is Very brittle, you know, how are we gonna create the next computer? So, you know from there about six years ago, I continued thinking about this Mainly out on in my free time and just you know trying to imagine another way and potentially the future of computing so it's a very Exciting and interesting project so Yeah, yeah, you you might be here on the ground floor or who knows right? That's that's kind of the nature of research. It seems to me Yeah floor or only floor. Yeah, right exactly. It remains to be seen Yeah, anything any other Biographical stuff you want to add Dave or Well, I mean I feel like I talk about it so much. Sure. Yeah, you're the star of the show That's me. I've been into computers my whole life is sort of as soon as I encountered them. I fell in love I did Programming languages wrote a C compiler as an undergraduate With a professor that I was working with it was actually sold it compiled for the deck 10 for a lot of years back in the day and Went off to grad school and learned about this thing called neural networks and did some early work on this thing called the Boltzmann machine Which is one particular Kind of neural network And then went off to bell communications research the lab run by the local phone companies in the 90s And where I had this great job Where my job? Obligations were to work on problems of my own choosing So I guess I actually have to be clicking to switch this main video here Deliberately, I kind of thought it was go automatically, but I'll try to remember to click it deliberately anyway, I worked at Bell core for seven years and then I went to the University of New Mexico and I worked there for 24 years and I Did the genetic algorithms that Alan was talking about I did all kinds of artificial life stuff because it was all in pursuit of trying to figure out how living systems and computational systems all connect together and Did a bunch of computer security work biological inspirations for computer security and That's when I kind of got really depressed Because it it just became so obvious that there was really going to be no way To seriously make any of the systems that we had secure and it wasn't because people weren't trying it was because One bug was all it takes and to claim that you're gonna solve the last bug is just silly and But we're still we're still Dancing around as if that's possible and So I got depressed I Didn't publish a paper for five years. I got in trouble with the department And then around 2008 whatever it was. I got to the point of saying Well, there is a way that we can do better, but we have to give up on this determinism business We have to have everything the random everything's moving everything is unique and one-off and it only figures out How to get a good solid answer when it needs A good solid answer instead of figuring out that we're just going to clear the decks and make everything deterministic for start And that's what led to the movable feast machine which led to the idea of indefinite scalability as a way of driving The architectural decisions behind the movable feast machine Which led to the T2 trial project and here we are beautiful beautiful cool well, um I have And just to reiterate we are seeing the chat seems like the latency is decent decently low that is so We're watching see Tyler Coleman down there one or two other people pretty sweet So basically a full house is what I'm saying. It's it's standing room only here on the T2 tile um But Alan or or Andrew did you want to chime in with a question or a conversation starter? If not, I'm happy to To toss one out there Okay, so like I'm really interested in and really like how how did you? come up with really you know The the the basic idea of the MFM like where did this MFM come from did it did it come from the ether, you know like You explain like what motivated you and you know to realize that but like I Don't know if that's a good starting topic, but Take a crack at it If well, you know Cellular automata are a basic approach to artificial life Since von Neumann and Ulaan as well Right the idea of having a whole grid that's all updating rather than a single CPU that's controlling everything and a lot of the artificial life models that I had done research with were kind of like Cellular automata. They were sort of more agent-based, but They were inherently spatialized. They were taking advantage of their local environment and it seemed to me that if we're gonna go beyond Serial determinism, which you know, that was the that was what I came out of Trying to do computer security and giving up thinking that computer security was a lost cause Where we are now That well, what's the alternative if the alternative is not to have one thing controlling everything the alternative must be everything going somehow At the same time and coordinating And From there the problem was okay. Well, we have Conway's game of life And these other kind of cellular automata, but they seem to me obviously equally fragile Because every single transition had to be perfect all at the same time So if we give that up if we say cellular automata, but not Absolutely deterministic. That's really the key step And everything else kind of followed from that and then realizing that what you could get from that was indefinite scalability I now have turned it around and pitch indefinite scalability first, but it came from artificial life cellular automata and Asynchronous updates having things that are going off whenever they go off Yeah, very good good stuff and yeah, just to clarify we're you know, feel free to you know follow up with other questions your your thoughts Alan we're kind of kind of just Kind of just making it up as we go along So don't don't feel like you have to be locked into Q&A, but if you want to be locked into Q&A then That's up to you Not locked into anything. No, we're I'm just I'm just messing with you man. I'm just messing with you Yeah, for sure. I'd Andrew you want to chime in there or sure Yeah, you know Dave since you spent so much time thinking thinking about Sort of like the MFM and how that plays into this, you know different way of computing Have you ever thought about like well, were there other ways to approach it or or would you you know? I just imagine you've had so many more years to just sort of think about it and and You know, would you go down a path that maybe was different but get sucked back into where you were in the first place or are there any sort of I don't know just options or or other ways to think about it that you've explored, right? I Don't think it's the move will feast machine or the highway And so I I do believe there has to be some Bigger family of which the move will feast machine is an example That you could move around within this family without losing the key Phenomena the key abilities and and the example I want to mention is not actually my own Example but one done by a colleague Lance Williams Who's also doing quite similar stuff? But he's actually coming from it a bit more of a Well from my point of view conventional Computing focus he would disagree, but he's the one that was responsible for the idea of compiling a functional programming into Distributed series of actors running on one of these grids and the stuff He's doing now some of which has been published and some of which is still new is Very much like the move will feast, but it's a slightly different set of rules and my Suspicion is is that the model that he's using is is part of that same family or almost Adjacent to the same family like that And to me the number one ticket is a definite scalability If you can show me how whatever you want to propose is actually going to scale not just to wafer level Not just to date or send a level But can as far as it's going to go and then we would measure its properties in the limit of Imagining you had plenty of copies of it so that you had to pay for the communication costs off of each minimal unit you used on a scale and so forth and Exactly how to frame it beyond that. I'm not sure. I Still believe that in definite scalability if you take it strictly Is really a pretty good razor for saying or is this Architecture going to be admissible or not to this family that we're talking about Good stuff Got got a question from the chat It might be a hundred percent sure if it's similar Hector Daniel says what about the concept of cellular wave computing? Why not use this kind of computing? Cellular wave computing is does anybody know We would have to Google it or Hector you'd have to unpack it a little for us Yeah, or and just following from from the previous response. I guess question one. Is it indefinitely scalable rights? That might be the answer right there depending on right Other question from my con I feel like the first state for the t2 tile space will be some kind of primordial soup of atoms What is your vision on what kind of system it will? evolve into Given where I'm coming from given my research history the sort of Really natural question is is what will it evolve into? How far will it evolve? and On the one hand, I'm dying to find that out But on the other hand, I'm trying to do the engineering Without evolution at the moment the low-level stuff the transition rules the alarm behave functions the splat rules In my mind, they're really at the level of physics They're really at the level of protons and neutrons interacting and sticking together if you get them close enough whatever it is And you really don't want the behavior of a proton to evolve There's some stuff that you're kind of going to count on to build on up above it And so a lot of what I feel like I'm working on what a lot of us are working on Is really at that level of physics where it's like can we engineer stuff? Like a cell membrane or like a ability to create a state machine that can make Decisions in some distributed way. That's robust and so on And if we can then Evolution is going to pop up at the next level above once we actually build protons and neutrons and molecules and Cell wall and reproduction and so forth Evolution just happens. You don't actually have to put it in In all of this process, I want to say the goal is Computation that's useful for society The goal is a new kind of super robust scale machine that can do system control Like that can do lane keeping that that can keep your house temperature curtains, whatever In some reasonable range of values For the as long as your house lives not for until Nest gets blocked by Google and shuts down all the servers and you know all the thermostats die Yeah Until the Chinese hack into your your internet devices. Yeah, or Americans it could be Americans I don't know. Yeah, could be anybody and in fact the world today It is everybody Like that. So I don't know. I lost the thread. No, that's good. Yeah, I think you I think it was yeah, what? Well, and I feel so I feel like and my con also asked also how large Contracts will actually be for like usable programs. I had this question to like, you know if this is if these little atoms are sort of at the physics level and We kind of have to get zoomed out a bit to start to see You know these things build on top of each other like and I and I I think I was trying to say this in chat Like I know it's pretty unknown to kind of really say like well at this level This is where it starts to be useful But do you have an idea of what youth where you think that level is like like how far do you have to sort of? You know zoom out on the microscope to actually see something useful my feeling is Cells That once we have a cell membrane an engineered cell membrane that has Reasonably good properties. It's movable. It's robust. We can put lots of different things inside it and it'll keep working That then we can go to the level we can we When people think about object oriented programming we would now start making the analogy that a cell is an object not an atom not an event window but the entire cell represents one object and it can communicate internally it can maintain quite a bit of state It's going to take a while. It's going to take a whole lot of transitions For a cell to respond to a message, but that's fine I mean consider how many assembly instructions a typical method runs when you do a to serial deterministic method So I think once we have Reasonably robust cells then we're going to have message protocols going between cells and from there It's going to be relatively flat You know Living systems take cells and put them into tissues and put tissues into organs and so on and yeah We'll need that sort of thing eventually, but I think we can get to useful computation with just so imagine taking a big like 133 trials or something like and filling them up With hundreds or thousands of cells that exchange messages with their neighbors that develop a routing system sort of like you know AJ stuff except at the cell level and then Continually has signals running through it like here's the position of my right left my right wheel Here's the position of my left wheel. Here's the bump sensor stuff all coming into this collection of atoms I'm sorry collection of cells and Then decisions being made in a distributed fashion and saying well then turn a little bit more to the right Turn a little bit more to the left whatever it is like that So I feel like once we get to collections of cells. We can start to recover Programmability at the level that we just assumed it by force When we did serial determinism with assembly languages like that very interesting Does anyone know how many cells there are in the human body probably they say trillions trillion Of which like half of them are human DNA and half of them are bacteria, but we forget about yeah, all right Yeah, a lot of hosts. Yeah, right Doesn't yeah more than many cells to do something useful. I think I think that's a good way of thinking spilling up the abstractions to nests abstractions and make sure that If you have some level of function and cells you can scale that function up and Achieve some some higher level use out of that Wow, so artificial life seems to be like very much important Part of MFM right. Yeah, absolutely And then the idea is is that you can blow holes in this thing You can destroy cells and that's okay because just like city streets if there's a pothole you can drive around it It'll take a little bit longer But then the local cells can grow and split and fill the whole fill the hole And now we have two things that are tailored to whatever function the one guy was doing and that's what I would just love to start seeing where we get You know Sell sell that differentiate into types like I'm connected to sensors I'm connected to the wheels and I'm doing this distributed state machine stuff in the middle and Then you know if you start if you start stressing the system Then it'll kind of automatically make more sensor cells to get the data faster or whatever because there'll be a payoff to it because they're needed And and then you watch the system in a way the system is evolving But it's evolving like a dynamical system evolving not like Darwinian evolution Like that, but you know again That's down the road Talking to hey Andy before we started the live stream You know it's like one of my problems is I really don't like to start painting these incredible This could happen that could happen in a hundred years pictures And I always want to stay as close to something that I'm sure I could implement if I actually pushed it through But I recognize that the world is full of people who are claiming this in principle stuff that could be implemented in a hundred years So the stuff that I'm really willing to say I see how to get to seems kind of modest Oh, we got a couple a couple comments in the chat here Good good stuff by the way, I think just my personal response to Alan Yeah, that the a life stuff seems seems very related Tyler Coleman the MFM seems rather primitive Programmatically, how do we expect to get to that? I'm sorry. We expect to get from that To evolving systems. Do we have precedent as far as evolving system and software and just in case these are related? From Joseph collard welcome, I recognize that name from somewhere I like the analogy of a chemical reaction happening in a beaker. Oh, this isn't a question This is an observation if you pour half into another beaker the reaction doesn't stop it continues in each beaker JP following on From the mention of object oriented programming. How much do you guys see this stuff as a kind of evolution of? Object oriented programming or object oriented design actor model, etc. Is a different thing fundamentally? Yeah, I think you know The fact that our object oriented programming hasn't gone away Even though it evolves over time is exactly, you know, is inheritance important or is that inheritance evil? It's still called objects. You know that kind of thing Speaks to the fact that there is something fundamentally right about having objects agents actors, whatever it is So I think and again the idea that the idea of a cell Being like an object and again tissues and organs being like large collections of objects that will get to down the road is the right or a good a useful analogy and The distinctions between agent actor I see as less important as long as We have something that that's as real an object that we can That the inside is not private just because the source shows PR IV a T E colon It's private because it's in a dedicated area of space that nothing else can get to Like that and you'd actually have to inject atoms and poke through the cell membrane to even see what was inside there That's what private should mean Yeah, we have objects Ram Really the instant the program starts running objects are complete illusion It's all just addresses in memory. You can just walk right out the end of object and be in the next object like that So I think in a way This approach the spatially distributed approach is going to be a more genuine Instance of objects a more a true instantiation of the object idea than what we currently call object-oriented programming in cpu and ram Yeah, uh, I was going to add maybe that So in in partial Agreement, I think, you know, it's it's useful to think about objects but to me these are really ways of thinking about it like object-oriented design is a is a design paradigm object-oriented programming is a programming paradigm I think the actor model is some paradigm of like actors to each other That's a useful way of thinking about the machine. I think if you spend any amount of time writing Uh, ulam programs today are programs for the mfm You spend time thinking about how actors can talk to each other. So that's it is that is a useful way of thinking about Uh, and in the message size is is quite small. So that's a that's a challenge, too Uh, how how to how to gossip and and send messages around I think that's useful uh, this the comment on spatial, uh ideas is actually Something I never thought about before is encapsulation in space is very useful It's literal. It's it's really encapsulated You know, unlike unlike object-oriented programming. I think it's just not so much uh It's just a really it's a metaphor metaphor for for What the machine is actually doing which is not like that at all Yeah, as long as the the chain of dominoes so to speak works and your your super user bit stays At user and the group permissions are assumed then you are okay assuming that They're separate Once again, but then you're back to the first bug and you're dead Oh Yeah, that was that was one um I don't know if it was illustration or analogy or example that resonated with me was Do you really want your passwords to pass through the same circuit as Everything you look at on the internet or you know Password as a stand-in for anything you care about, you know health stuff financial stuff And it's like wouldn't that be cool if there were somehow away those Could be separate And and that that was a really interesting idea Yeah, and so I I I tried to say that at the beginning of the path to best effort paper where it uh, it's like In the in the past people thought this made sense Sending all of your information through the exact same place in space Um, and and how we're just going to go. Oh my god. What were they thinking? Yeah, uh, once we get an alternative Once we get the engineering stack built up high enough That people can come into it and start programming at the level of cells Exchanging messages and these messages are going to be Big heavyweight slow things compared to the messages that I think alan was talking about Uh, uh, you know inter from one atom to a nearby atom in the event window or close to the event window That's going to be sort of microscopic signaling that would be part of this giant message floating by on its way To the cell membrane looking for a related guy like that But then people are going to be able to think about it from the level of cells and And intercellular messages and and neurotransmitters and so on and so forth and just just as we do today Ignore all of the stuff that we're doing below it and just complain. Oh geez, you know message passing takes three milliseconds Oh, I can't stand it Like that, you know, and and and we go, you know, you're welcome Uh So yes, uh, absolutely and and you know, I think at that point Once it's that obvious People in the traditional computing community Will it'll be easier for them to get it. They'll be easier to make the mapping I mean, we keep trying to make the mapping From what we already know to this crazy new thing and we try to apply it to atoms We try to apply it to bits inside an atom. We try to apply it to that and we'll try to apply it to a tile Whatever it is, but I think where we really want to apply it is to a sort of floppy cellular bag And and then we're going to go for primitives from there. So so getting to a decent kind of A good cell is is really the signal Software achievement that we're going to try to get to once we have enough hardware That we can say, you know, it's actually okay if the cell takes 10,000 sites Of space because 10,000 is a small amount of space in the way that we should be thinking Maybe someone Is is thinking this um, but I think I would like to bring it up is You know, we have this technology stack. Uh, that is very successful today Which is traditional, you know computer computing so You know, I would say why why can't we build this? At different levels of the stack what you know, why can't we build security at this level of the stack? Why can't we have physical security and separate things from each other to Kind of you know, why does basically why does the mfm need to exist? I know someone's thinking it Before before we hit that can I can I pause for like yes a quick advertisement? Can we uh, would you mind? Bragging about your shirt, Alan Oh, yeah We got the uh, Alan is a an early patron of the andrew wapol Uh Maybe andrew wants to talk about it I think Great shirts great quality t-shirts Um from the andrew thank you very much. They're awesome. They feed you I think people uh It starts a lot of conversations. So um Yeah, and that's great great Beautiful. So so yeah, right now they're being sold at cost. There's it's not funneling back into the project or anything We've been vaguely talking about crowdfunding or something for tiles because there's there's a little bit of cash involved but no no hard plans um And uh got a quick question from the chat. Who are these people? probably good to Refresh that if you guys are comfortable with that before we go to the why this project exists um My name's hey andy Aka nate. Um, I'm just a fan. I'm a programmer um andrew I'm andrew wapol. I'm a designer and web developer Hi, uh, I'm alan uh software engineer by uh by trade And this is Dave very cool So yeah, I just wanted to get that quick advertisement in um before we dug into Uh The big question If that's cool Oh, why does the mfm need to exist? Is that the big question? I That was if we want to go there. We don't have to go there, but I think yeah I guess I'm happy. I'm happy to go there. I mean part of the problem is This idea of design lock in that when you make decisions in a computer architecture You you don't just make one decision at a time of cpu and ram is not one or two Architectural decisions. It's dozens. Uh, uh, it's it's representing numbers in binary. It's assuming the bits are really good It's assuming the memory is 100 stable and so on and so on and so forth and So the challenge is when you say, well, we need to change this We need to come up with a new approach and you say, well, what are you going to do? You're going to say you're going to let ram bits fail well, uh And then what the if statement might just go the wrong way for no reason Uh, uh, what are you going to do? And so the problem is If you take just one little step like say, okay, the the cpu is going to fail every so often Our memory is going to fail every so often. You're in a position where it's absolutely clear The right thing to do is to fix that And and go back to where you were make it so that memory won't fail At least however long it takes to do with whatever computation you can make money by doing And so that's the challenge that I faced. That's the challenge that society I think has been facing Since von Neumann when von Neumann said in a few years We are going to come to a new architecture that's going to support the probability of failure of every Yeah, and now we're 70 years later after that and we're not doing that so The reason one one answer is the reason the mfm needs to exist is because we need to flip eight or 20 design decisions all at once To get to an alternate approach that now pulls together on its own basis Instead of being a garbagey version of serial determinism It's a great version of its own thing like that now You know, that's like telling people There's a new world across the Atlantic Ocean And if you're just willing to take on scurvy and and dying and whatever you can go there It's a big ask No one knows if that's really true So I see that and that's one one answer answer number two and then I'll shut up is Once we build cells and membrane signaling and all this kind of stuff like that we could if we wanted Implement something like cpu and ram on top of it We we could now actually say yes You you could you could make a distributed cpu That exchanged all these things to decide what message to do next and you can make a bunch of distributed stuff that would act like ram And it would you know, you'd send out a request by address Sure, and it would come back and give you the Contents of that address and so on and it would be it would be slow But it would be inherently movable and robust in all these good things So we could have what we currently are doing back But it's been lifted up to this robust movable level And I think the reality is once we get that far we'd say well, you know, yeah Actually implementing dos 3.0 is not the best use of all of those cells We'll go direct to utility Within the attractor of robust first computation instead of now using this platform to implement what we already had Good stuff. Good stuff Anything from you andrew other than visit mfm.rox um You know, I think I think you kind of just touched on it, but just It sort of plays in my head like like this idea that you know if well We come back to security as sort of the main issue and and rightfully so Um For a system like this, but then you might have somebody make the argument and say like well I've got my calculator. It works quickly and uh, you know, I pull it out of my pocket. And so like like do you I think I might be getting caught too because you know, you think about the distant future And you might think well a robust system like this that that does have all those levels stacked on top of it You don't really need to to to think about this question, which is you know, do we have both? Is that okay? because it's reliable enough even in its robustness that You know, you might as well have your calculator on it, but Until we get there until you know, you you have so many layers that that reliability exists Uh in an effective way You know, do you see these things? Sort of needing to get along and and needing to sort of say well, let's go after these applications Because security matters more Um, and you can keep your calculator in your pocket for a while. Um, you know I don't know if there's a question there, but I I put in a pitch again for the path to best effort paper That at least it's a weird thing This paper, uh, but I found it so I put it up on the web for people to look at Really good read Thank you. Thank you But at least it takes a stab at andrew your question at saying where are Tiles distributed first robot robust first stuff going to get design wins In a world where we have zillions of traditional computers And so on and and it suggests an answer to whether I think that's actually the answer that's going to happen or not Who knows? But you know spoiler alert A password manager a secure password manager Would be that that's what the story postulates In effect and and the point is that yes, you ask can this movable feast stuff work along with Conventional stuff and the answer is yes, absolutely, but it works along via protocols happening at the edges It sends messages. It receives messages and so on the idea that the stuff that's executing inside this Is the same universal MIPS arm x86 instructions and is executing in the other thing that goes right out the window But that doesn't mean that movable feast sensory robust control systems can't talk by Interfaces by message protocols back and forth with horrible disgusting not robust Traditional technology. So that's the idea And we need to get a killer app of some sort enough that we can drive the hardware costs down So that we can have tiles like water Um, and at that point that will be where you were saying might as well have your calculator Run on top of this thing because you're buying this stuff by the square. Y'all right That makes it that kind of you're kind of tying also into so to like kind of get out of the future and and all of that sort of presuppositions and stuff When when you build, you know for t2 tile the next big thing is sort of I don't know. Did we settle on lotus or power lotus is Yeah, I think a ring lotus. I'm thinking the build is going to be a ring of lotuses seven lotuses So when you build that Um, what you know not to fully jump ahead, but I will What's what do you what are you excited to do with that once you have that? sort of there Yeah, you know just to stand it up and get it working is Super petrifying enough Uh, uh, but the next thing after that is I want a benchmark drag and res running on a ring lotus and say we have now demonstrated Two air three air 1.5 events per site per second across an indefinitely scalable grid And if we decided to go back and make seven more ring lotuses and make a I don't even know what the next name would be We would still get 1.5 air whatever it is because that's the nature of an indefinitely scalable metric So that's step one to draw the line in the sand 1.5 air has been achieved zero point zero one air has been achieved. I don't know what it's going to be We'll find out zero Yeah, exactly And epsilon air has been achieved The rest is up to the reader Like that and um So that's task number one and then task number two is what I really want to do is what I'm better at which is Software and get up working on getting to the cellular level So that we could take a ring lotus and use it to the control the Stupidest robot you ever saw a giant platform big enough with car batteries to drive the whole thing And then we could start virtually shooting holes. Yeah, uh in the grid. I mean they're a little expensive To show how it behaves how it works when you start stressing the system and instead of falling over like every single Traditional implementation will do it's going to do surprising and sensible stuff Even if you knock out vast majorities of it because the cells will still have enough to find a connected path through the hardware That's still there and they'll send messages and they'll do what they can do But even that, you know, if that happens in 2019, we're doing great right cool So that that kind of leads me to like the next piece. I I'm the kind of person I always want to get to like You know, if you have that vision for that future, how do we get there? And and and draw that so when you talk about these, you know, these first cells What what's known and what's what's kind of unknown? Like is there a place for us to start to look and poke at and start to find Our way through that path, you know, do we do we do we say we want to build a cell that can do XYZ and then work backwards from that or do we just start playing at the elemental level and building a little army of things and then Finding the bridge somewhere in between there And can would it be okay to pause and we got we got a little bit of stuff in the chat. Is that cool? Sure. So we have Where to what to do next to head toward that vision with a pin in it Cool. Um, some we got David new house in here who apparently knows more about biology than I do, which is great Excellent question from my con Have you ever played with the idea that this kind of system could be developed biologically instead? Since humanity is starting to fiddle with direct DNA manipulation, etc Tyler Coleman brings up the term wet wear Right. Yeah The a life distinction is hardware software wet wear hardware robots software Game of life and wet wear actually using chemicals water solvents and so on I think people doing wet wear are extremely cool But if you think Dealing with oolong is primitive then try dealing with wet wear Um, I mean, they're they're building cell membranes by using detergent and oil You know that kind of thing. So it just snaps out and makes a non-polar membrane inside Polar solvent inside water Like taking grease out of your way Um, and which again is incredibly wicked cool but I'm a software nerd and I want to get to the level where I'm programming And so I think, you know, yes, I want people to be working on the biological stuff and also yes In the long run if we can figure out sort of general principles of distributed robust stuff Then those general principles will somehow find application In the detergent and oil wet wear future like that, but I can't really contribute much to that to the wet wear side. It's just not my skills and and my my Ambition is based on We've now spent 70 years figuring out how to put transistors and wires on on silicon And we can do it extremely cost-effectively Extremely large numbers extremely small and the question is how should we connect them up? Like that so to take the existing technology we have for making silicon chips and say all you need to do Is use a different architecture and connect them this way instead of connecting them that way Then we can have a whole new thing to me. That seems like an opportunity Because now the world is full of extremely cheap 2d silicon gates and what can we do with that? What was the other stuff? And then I think yeah, I just wanted to it had been a minute since we had Looked at the chat just wanted to get that one. And yeah, sorry to cut you off there. Andrew You were asking something more practical about actually doing things, which I don't know why you would want to talk about that So what to do next? You know the the c2 11 cell membrane Is out there somewhere the c2 12 membrane the one that appears in the retirement talk and the one that appears in the The a-life talk from tokyo this past summer Um, and that's where I'm going to go To try to flesh out this cell membrane vision and and really Once you start to understand how the c2 11 membrane is working It really is a lot like oil and soapy water Making little elastic things that just react completely dumbly if the inside stuff pushes the membrane grows out If the inside stuff disappears the membrane collapses back in and and that's it um and what I think We need is Better stuff to put inside it That's say, okay We're gonna we're gonna assume that a membrane like this is going to be our basic membrane to work with So it's not like we say do we need a better stronger robustor membrane or whatever we say Let's have the membrane be like this and have figuring out what we need to put inside How the stuff inside it should behave so that we can make the whole cell move We can make it eventually split in a controlled fashion The last demo we got Before I switched from software back to all this hardware stuff Was uncontrolled reproduction essentially cancerous reproduction where it's splitting all over the place The next stop I would love to see is controlled Reproduction that somehow you you drop an extra command atom in there someplace saying Reproduce now and all the goo in the middle doubles up its count and segregates apart And the membrane will just automatically grow and stretch around it And if you pull it far enough the membrane will just go And now you have two Like that it's about getting control the reproduction once you get control reproduction You can totally have evolution if you want it It's going to be hard in fact to avoid it So that's what I think the problem is you know the c2 11 membrane is written in splat It's got a bunch of escapes to Oolong inside it So you really kind of need to be comfortable with oolong to even understand what's going on in splat And you know, it's only like 80 or 100 lines or something of splat But Like that so from one point of view we just we need to build up the skill set And we need to get as many people who are comfortable With oolong and then with splat and it may in fact be possible to get comfortable with splat Without getting comfortable with oolong and andrew that was something that was sticking in my mind about mfm.rox the idea of Actually making a splat parser that doesn't have an escape to oolong It just has some built-in predicates and actions like you were talking about That people could play around actually writing splat rules and it would be super cool if people could get far enough to actually make a membrane I mean a c2 11 analog in Splat minus minus Like that although I think we would then find out that you know, damn This is kind of running pretty slow Like that, but then it's no worse than mfms. It's no worse than the simulator we have now Yeah, I think I think then as far as mfm.rox goes like that's that's kind of my vision is You know, I don't I don't really i'm not super concerned about you know aligning it more with oolong though I I think some things need to it needs to from a conceptual perspective There's more alignment to do but I think there's this opportunity to You know let people play with the concepts and I think You know coming into it like it took me a long time and a lot of watching your videos to like really Get even like maybe like I'll say the concepts are kind of gelatinous in my head. They're not fully solidified But that's kind of good enough and fun enough That makes it really interesting and and so I think the idea there is to like provide a tool that Kind of helps people with that in a hands-on way Right. Well, and I think it already is to a degree and the more development they can do the more sort of programmable Or little mini, you know, you can customize this element by clicking here or something like that And then just see how the behavior changes The better, you know, right Well, so it's coming close to the top of the hour. Do we have other things that we want to try to touch on? I think I'm okay Sorry, I think we're doing great. Um, I think Some people are are asking questions about like how do I get started which is kind of what you guys are are talking about um I I put the mfm rocks link in the chat um There's also uh an ulam laying Wiki that has nothing on it right now, but stay tuned. I think I'm going to uh Try to include some splat in ulam Demo's in there as I uh learn more of these things myself too Right. Yeah to the degree that we can And after I get access I mean all all that I'm trying to do which is new and outside my comfort zone Is to do all the development in public and to try to Make the mistakes be public make the discoveries be public and so on And let that be a trail of breadcrumbs at least In as a down payment for you know proper tutorials People that know how to explain you know To sequence these head exploding concepts that you really kind of need like six of them Before you can really start getting it like that and all of these things as somebody said I can't remember that Maybe it was Andrew or AJ that you know, I've forgotten what I didn't get Like that so the idea that you know That a site and a tile are two different things for example. Yeah, yeah You know is is not obvious from the outside and the idea that the tile itself is really kind of an artifact Rather than a truly it's an engineering decision to even have a tile Is when given that that's what you see it makes perfect sense that one's model would begin with tiles I don't know. I don't know how to push that forward with 24 hours in the day I I hope that basically the the wiki can be a place to You know disambiguate and and learn basic concepts and even like once we put Have more code examples and stuff and start building our library of Adams and larger structures. I think it would be really cool to have those stored in one place and That could be a place for it. I would love to see like everyone Having a having a place to exchange and contribute ideas. We have the the getter but you know Building that up is is important like as a community. I think so, too It's totally new to me as well. I mean Right. So so write down your confusions And So Where are we here? Um, do we want to take one more question from the peanut gallery and then and then sign off Sure. Okay. Well, thanks. Thanks everyone for for joining us. We're gonna take one more question. Um Follow t2 updates On tuesday, we've got the the live chat on getter just kind of an irc kind of thing Got the subreddit. We've got mfm rocks. We got probably too much stuff that you can follow But we're working on that but um, but this is this is home base um Final question for the day Tyler Coleman when when people all have their own t2 system. Do they connect up or have borders? I'm assuming he's at if if we're talking about like the internet and the boundary gateway protocol or something Or I mean my impression was it was just like whatever just add tiles and It tiles a tile right And part of it is again. How far in the future are we imagining? um You say If people have their own real estate How do they join up at the edges? And it's going to be the same thing because their own tiles essentially is going to be this computational real estate and Good fences make good neighbors. I mean and the to me the thing that Traditional cpu and ram abstracts away Which is the fact that these things actually occupy space and something is actually next to something and not next to something else Is going to be where we're going to go back to answer these kind of questions You know Yes, this this part of physical space is mine If I choose to fill it up with t2 tile computational fabric version nine running my wacky physics That's up to me Now if we want to have it connect at edges to somebody else's or some you know Public utility fabric or something like that run by the local government Well, then we'll have to deal with that. We'll have to figure out, you know impedance matching interfaces that different kinds of physics can exchange and so on and so forth My fundamental goal is to say we're never going to offer pointers Into our actual space everything that crosses a frontier everything that crosses a spatial boundary Is going to have to pass by message protocols And if we then have some, you know zombie code word that someone can say to us that causes us to reprogram all of our t2 tile fabric to Send spam or whatever. Well, that's our fault But it's not going to be easy to make that happen If that's an answer Well, you know, let me say this um I think this was super great Hey, andy, andrew a j Thank you so much for being willing to to give this a try Perhaps, you know, we could do it again in a month or two or something like that. I don't know I think it could be a helpful other thing Maybe other folks who are in the peanut gallery would be willing to actually join in On the hangout and and contribute face to face or whatever. Um I do believe That having the getter and reddit and all these different things which seems scattered And it does make the it is a bit of a barrier to entry But where I'm coming from it's much more welcoming than no entry Which is basically where we were so That doesn't seem wrong to me right now And the instinct of saying there needs to be one place where everything goes and so on That's the that's the urge to empire that leads to cpu and ran And and you do want that but you want it late So right now we're trying to develop a lot of different things the wiki and so on and so forth Let things develop as they do and let's learn from what actually does work and so forth and then go forward to that But you know I'm completely terrified about the the t2 tile build which is coming up and all of these components that are in flight and so on and so forth And I'm completely petrified that I'm going to spend all this money and then I forgot Oh, you have plus connected to minus and it's never gonna work Like that that's what we'll do the crowdfunding. No problem. Well, that's what happened Yeah, that's why I'm not quite ready to say I'll take your money For the first build of the t2 tiles if we if we succeed Spectacularly, then maybe we could think about a second build and people could sponsor a tile Or maybe we could even funding agencies might sponsor an entire ring or a lotus And my idea is is right if for people that sponsor it They could contribute a ping file that would show every time the thing boots up where it currently says mfmt2 It would say micon or or you know, uh, whatever it is as long as it was in, you know, not horrible taste And if you buy a whole lotus you could have an intel logo or whatever you want Showing on the entire little spatial region that that you sponsored, you know, but again We got to walk before we fly Well, how about back to you? Hey, Andy since you're our official guy I'm happy. Let's let's close it up. Thanks everybody. Thanks everybody. Thank you so much. Thank you. It was great. Bye