 We're trying to build a whole new computational stack based on these principles. First be robust, then as correct as possible, and finally third as efficient as necessary. Efficiency usually gets put up at the top, here we put it at the bottom. Top three. But take a look at this. I actually saw that, but it happened again right there. The arm going across in the new nano-arm technology. It's unbelievable. You see there it is, back in the previous one, it's still poking along. Just we've already had kids and this one is about to begin copying the code from the left go to the right go to the right. So it's a big difference. It's a big efficiency gain. So you know, I'm not a big computer science theory fan, even though I'm a computer scientist and a professor and all that sort of thing. But every so often, the good old big O, the analysis of algorithms comes up and says, you know, hey, here I am. And this is one of those examples, the microarm, abbreviated UA. The way it works, the source of the code is at the tail and it sends a command down to the head of the arm, which does it and sends a confirmation back and so on. As a result, the time per command grows with the arm length because it has to do a round trip each time. So if you have N commands and N time per command, you get quadratic N squared time. The new one, the nano-arm NA, stores the program in the head of the arm and executes it directly out from the head. So the time per command doesn't change depending on how long the arm is. So it takes linear time. So it's a perfect example of efficiency of linear time versus quadratic time. Now anybody who knows anything about this project would legitimately say, why, Dave, are you caring about efficiency now? And the answer is, I don't. I don't care about efficiency. But the stuff I was working on this last month, I needed to grow an arm that didn't have a controller at the back end. And the easiest way to do that, or the way that I thought of to do it, was to put the program in the head so that it could just use it and we could in fact tear down. What if there was a single worm head I thought on May 11th that carried its whole pattern so we could actually burn up what's happening behind it and still have the program there? That was the reason I started into doing the nano-arm but it had these other effects. And just to finish the story, why is it called the nano-arm? Well, because it has two bits per instruction, whereas the micro-arm has three bits instruction. So here are the eight instructions of the micro-arm, go west, north, south, west, it's got a bunch of different things. And the nano-arm is absolutely minimal. Go forward, turn left, turn right, and done. It's like the turtle geometry, it's like logo. So and it works great. So that's why not and big O. The schedule. This was officially artificial life creation T plus five, which seems to me a really bad marketing idea. So at the very end here I've turned it around, it's now the multicellular challenge at T minus one and counting because the next T Tuesday update in a month is the culmination of this six month journey. So what I was supposed to have done for today for right now is I submitted a paper to the artificial life conference that's coming up in Japan at the end of July. I think that's right. I had to set it in camera to copy, I did that, supposed to get plans for the second half of this year. I thought about it, not really done. And I was so beat down by failing at multicellularity over the last months and months that I was just asking for plausible or not actually literally plausible necessarily multicellularity. I think I got that and have some big fun. Well, you know, I still don't know if this is going to work. I still don't know what we're going to have a month from now, but, you know, certainly over the last month, I've been just pulling new design out of thin air and just making up stuff and, you know, going to Wikipedia finding out how nature does it and says, okay, I can implement that. So yeah, but we had some big fun. So the schedule continues this month was supposed to be extended germline stretch goal, not even exactly clear what that is, but it's been retroactively redefined to be morphogenesis, the process of building the shape of an organism, in particular, building the shape of a multicellular organism, which is what we've been trying to do. We're still holding out for the multicellular menagerie a month from now. Now that is actually July 4th is the next tea Tuesday, which is not going to be that great necessarily for a live audience in the United States anyway. But YouTube tells me that one of my key demographics is the UK. Anyway, folks have good taste. OK, that's the schedule. Meet M.C. Slash. OK, here we go. Let's take a look at this clip growing there. There again, that was the nano arm going fast, splitting. And now they're sort of bouncing off each other and they almost completely miss each other. But look at that. There it is. They have locked in and they're really pretty good. So I started a bunch of other of these little things and it did not. Well, a bunch of them. So these are now the lower parts of three different M.C. Slashes, which have now teamed up to form a big backslash. But that's not really real. It's only doing it because it's jammed up against the edge of the universe. So meet M.C. Slash, the multicellular slash. It's got two cells. I have kept simplifying and simplifying and simplifying. But this actually kind of works. Multicellularity is hard yet. It's still hard. Yes, it's still hard. So just to make it clear where we are at and where we are not at, here are some of the bigger bugs that are currently in the code base. I'm not going to go through any of this, but you get the idea. So it's working a little bit and OK, I will take it. All right, so that's M.C. Slash. We'll talk about it more in the next point. Body plans and body models. So yes, the big talk for today is morphogenesis. Literally the generation of form, biological process causes a cell tissue organism to develop its shape. Yeah, that's what we're talking about. It's not good enough just to split and make multiple cells. We could do that when they were being single cell creatures. But they have to actually organize themselves in space in order to become a multicellular organism. And there's going to be, we need to be, I have now implemented a body plan concept that says, well, these cells are going to split and one's going to hear, one's going to hear, you know, that kind of stuff. Taking into consideration that there's no absolute geometry, there's no go to X, Y in the whole universe. Everything has to be done relative to where we are and where we think we are. And so here it is, engineered morphogenesis. So, you know, biology and evo-devo evolutionary developmental biology studies how life does it and life does it. Unbelievable, a little different way. So, this of course is coming out in the other direction. We are engineering morphogenesis under a strict set of rules of indefinite scalability, which means we can't cheat. We can't use pointers and so on and so forth. We have to keep everything super local and figure out how to make these big things get where they need to get. So it's something like we have the original ancestor, which is going to split into two kids that we call two and three. So I'm giving each cell in a developmental sequence of a multicellular organism a number, a diamond sequence number. One, two, three, two splits into four and five, three splits into six and seven. There's this neat trick in computer science where you can use a data structure called a heap that can be implemented in an array. And the way you do it is the root is sitting at some index, some number. And you say the left kid of the root is at double that index and the right kid of that root is at double the index plus one. Two times K and two times K plus one, and that's what we do here. So the ancestor is one, the first left kid is two, the next the second, the right kid is three and so forth. Two splits into four and five and so on. It works nice, it's a direct analog. But we need more than just the tree structure cuz we have to lay them out in space. So one goes to split something like this, but now we want to get it into MC slash. We want to get it on a northeast, southwest axis. So what I decided to do is to make some information. Now in this one I'm recording the actual sequence numbers of the other ones. I eventually changed it so I'm just using kind of arbitrary tags that are supposed to match up. So if this northwest has tag five and this southeast has tag five, that means they're supposed to stay close together and that's what MC slash uses. And eventually it turns into this body plan, a series for die sequence number one, that's the ancestor number two. That's the one that's supposed to end up down on the left. And die sequence three is gonna end up on the right, up and on the right. And so there's tag five that's supposed to made up with tag five and so on. And it actually kind of works. Now, in order to actually do all this, we actually have to go through this whole sensory motor interaction where we get information from the world. We make assumptions about it and we take actions and we see what's happening. And almost every AI paper you ever look at is gonna have something like this as sort of figure one, the basic layout of it. Here's the creature, here's the world and there's this loop going back and forth around it. This is from the first scientific paper I wrote pretty much in 1983. And I'm just gonna focus on this part, the real world, which some, you can't see all of the real world, you can just see a little bit of it. And that gets turned into a world model that is actually part of the space that you own, and you deal with the real model, the world model. You make inferences and decide what to do and so on. And all of that has actual physical layout roles in terms of, in this case, MC slash. So I mean, so here's the real world, all this stuff out here, it cannot see at all, there's anything that happens out there. In particular, actually, this other diamond is not visible at the moment because these antennas would be white if they were actually sensing each other. They're not at the moment, so that's all part of the hidden world. Only the stuff that is very close to the thing is actually capable of being picked up because it's in the event window of the antennas or the membrane and so forth, and it gets signals, gets sent inside, which so the edges of the diamond pick up the information coming from the antennas and make decisions about it based on what they know, that I'm looking for tag five, is there a tag five there? No, there isn't, so we better get away from that. All of that stuff happens at the edge, but it has to get brought to one place so that we can do inference on it. Okay, we shouldn't go northwest, but we shouldn't go southeast either. Can we go northeast? And that's where the inferences in the world model are done or the body model, in this case, the world is the body for the single cell creature. These four, well, there's five atoms there, but that yellow one is an empty one, are the body model. And each one of these is meant to control one edge. This one is supposed to decide what the northwest is doing, the northeast, the southeast, and the southwest of, in this case, MC slash. And this came from the genome. It was copied as part of the initial deployment where there's now a new chunk of the ancestor at the beginning called the body model prelude that gets copied over here and it just sits there waiting until the mother loop goes away and then it drives around and encircles itself. And that's what the nano arm was originally for. So that once it was sitting there all by itself, we'd go broop and close the loop for the body model to be there. And all of the decision making about let's go north, southeast, west, let's grow, all that stuff happens based on the body model, not based on reality, not even based on the perceptions, but on the stuff that's all been centralized so a decision can be made. I mean, this is obvious stuff, but making it all physical, making it all real, there are the actual implementation, it's code, it runs for all of its troubles. So that's pretty cool. And finally, goals and plans. So for a multicellular challenge T minus zero, goals is MC slash lives and can actually reproduce so that we have a two guys, two beings. And then we have another MC slash over here and so forth. And if possible, we should get it, it's mystical, mysterious buddy, MC backslash, little die, which is actually a two by two arrangement of diamonds. That's what I started working, but that was too hard. So I went back to MC slash, come up with a schedule for the rest of 2023 and have tons of fun. And that we'll see, we'll see where we go. I mean, the chances are for the rest of the year, it's going to be not about T2Tile project so much, although maybe we'll continue the live streams if people want to, but about making videos on the Dave Ackley channel, so we shall see. Folks, thanks so much for stopping by and checking out and having an interest in the project whenever you can manage to do it. And I hope to see you next time.