Introduction to the T2 Tile Project 

Never has an idea excited me so much since I learned about the Newton’s laws of motion. I am a mechanical engineer who Worked with little GPU programming in 2008 when there was not much support. So I understand what Dave is proposing in bits and pieces. Yet it excites me so much. I’m so happy that I’m a monthly donor to this project. I wish I get a really high paying job soon and donate hundreds every month to this project

Very exciting. Keep us updated!

Ahh. Finally I think I can explain it to someone. 😊. I never understood the link between robust first computing and creating an indefinitely scalable computing architecture

Whenever I come on Dave's channel and watch a video or two, it always warms my heart to see someone talk about such a cool topic with such enthusiasm. His talks have given me a lot to think about, and have allowed me to put a finger on and express more succinctly ideas I had been grappling with before. Indefinite scalability and ALife are so fascinating to me, and I want to see them take off in the near future. But I also wonder whether these types of ideas are already being put into practice, albeit in different forms and names. I've heard of supercomputers being upgraded and adding more nodes, and while certainly not indefinite, I imagine something like several magnitudes of scaling could be considered reasonably close to "indefinite" depending on one's point of view. And perhaps programs, containers, networks and entire arrays of hardware and software are already performing living computations, even if we can't easily comprehend them. One of my main questions is, how and when would the ideas Dave talks about be widely adopted? For instance, in supercomputers and data centers. How could we make that happen? Would it be worthwhile to try to engage a broader audience in the tech world, pitching projects to IBM and Google? I feel like I'm so ignorant about so many things that perhaps these questions aren't even worthwhile somehow. But if anyone would care to shed some light on them for me, I'd be completely grateful.

This is amazing!

I've been watching this for a like 3-4 years now, and I still don't know if Dave is actually insane or a genius

We need an update Dave, you have been making so much progress but its hard to follow you every week. Would be great to see the current status of this project in a bigger scope

A very interesting and ambitious project; I'd be excited to see what kind of results you get! Im curious about you think about the emerging blockchain technology as it's currently tackling similar scalability issues. The idea of life-like computational systems is crazy. I've had a similar thought when trying to imagine the next stage that may replace our current deep learning paradigms like convolutional neural nets. The issue with current deep learning architectures I think is that they don't actually learn like a brain. They require many more training examples than an animal would, and they are not complex (in the way that the brain is a complex system) but just highly optimized machines.

Cool

How does the saying go: “go big or go home?” Have you thought of implementing your T2 Tile Project using ASIC elements? I went down this exact same road a number of years ago when attempting to build a large asynchronous array of simple processors. The wall I hit designing such a system is the cost involved. No matter how I tackled the hardware side, I could not come up with any design options better than about $30 per cell. At this rate, for example, 100 cells would have cost about $3K. In other words, any array small enough to be affordable could be quite easily simulated in software at no cost at all; so why bother? On the other hand, there was a chip, the GA144, from a company, Green Arrays, that was a 8x18, 144 element, asynchronous array of simple processors that each independently clocked about 1 GHz. All this on a single die approx. 5mm x 5mm. And you could pick one up for about $20. It was a tempting part to try to use, except for some serious flaws. The first was that the simple processor elements were too simple to actually implement anything useful with and second, the I/O pins were not brought out in a way that would allow these parts to be incorporated into a larger tiled array. As far as I know, Green Arrays is now a defunct operation. But I’ve always wondered what could have happened if the engineers behind this chip had designed it is a tiling element rather than a SOC part? But, if you think the T2 Tile Project is worth $50K to $100K, then going the ASIC route, taking the GA144 design as a starting point, makes the most sense. Not only would you be able to build asynchronous arrays on the order of 10K’s of cells, you might actually end up with a part that you could sell commercially too.

Use FPGAs ?

How is this really all that different from serverless architecture?

So like a beowulf cluster?

How can I help?

So, basically, you want to make a supercomputer out of self-sufficient processors, that can be detached and attached in the runtime? I am not really a computers' person, you see. I don't see any benefit in what you propose. To me it all looks right now like a gimmick, that only reduces processing power. Though speaking of determinism, I agree that it would be nice, if we had more ways of trading off accuracy for speed on a hardware level.