Robust Local Synchronization - Research Notebook Video

Discussion and demos about synchronizing the asynchronous robustly in computing systems.

The T2 Tile Project:

Software:

Living Computation Foundation:

0:00 Introduction
0:45 Sync of Computers Past
3:05 Sync of Computers Present
4:34 Sync of Computers Future
5:58 Sync over Asynchronous Cellular Automata
8:25 Demo 1: Waiting for the neighbors
10:56 Simulating synchronous updating
12:30 Demo 2: Conway's Game of Life
15:00 It's the end of the universe
18:30 Bottom Line: Global sync fails
19:14 Local programmed sync is different
21:28 Demo 3: The jerk and the empath
24:25 Related work: Ring Oscillators
27:50 Demo 4: Generalized Software Ring Oscillator
34:13 Discussion: Against object-orientation
36:53 Conclusion: Fight the master of the universe

[jayp]

Please make regular videos if possible! Very interesting stuff. Always love to see people rethinking from first-principles. Every paper I have read on brain oscillations suggests that the brain is a distributed, emergent computer. Very excited to see more of your work.

[isaacmceachin]

Most interesting man on the planet! Really enjoy the content and innovation.

[ptrckqnln]

I would've happily watched for much longer than 40 minutes - this is riveting stuff!

[Pheonix]

It's nice to see a new video again!
This seems like a really cool idea. It's like you're getting the best of both worlds, like you said.
I can't speak for anyone else but I could listen to you talk for hours.

[mario.caseiro]

11:47 Nice master.... Nice to be in touch again. Mather did resist for 21 months. It was the best moments of my life teacher...
Grate to be back. Here in Brazil they already using the concept of AL with 5ghz and UDP packages for all IOT (tv, mobiles, notebooks ipads routers and so on)
I'm Glad to see our studies comes to concrete life...
New news .. Just got married again. Now I have time to give some help in what ever is needed master .

God bless we all and its good to be back

;)

[rickwalters]

This was mind-blowing. I've been messing around with abusing BEAM for some very rudimentary versions of these ideas and this one video answered a number of persistent questions… and better yet, raised about 10x more!! My introduction to your work (and your fantastically clear presentation style) can't wait to discover more

[murilopolese]

Best 40 minutes of my life, you nailed it! I'm totally a LCF nerd! 🤩

[MrEstrax]

Excellent video, watched all the way through.

[ZeDlinG]

this is super cool, if I had teachers like you, I would've never left uni

[sean...]

Fantastic solution Dave, love your mind :-)

[endofsilence]

wonderful! thank you for your content and I'm very much looking forward for more!

[Pygochelidon]

So how do we go about building a robust first society?

[mgetommy]

this was awesome. watching the pattern persist even after that large splodge of wall was super cool. kinda creeps me out reminds me of grey goo for some reason, like computer system thats robust to anything

[BlackAhorn]

I think your ideas are revolutionary. I'm having a lot of trouble sharing my excitement with fellow humans though, as it's very hard to put your ideas into a, well, graspable practical context. How do i tell people, in metaphors maybe, what the great benefit of decentralized, asynchronous computation will be for us? I can barely describe why it will be more efficient in large, or rather huge, systems, but especially the x-ray, nuke or wall disturbances you throw at it to prove the stability are difficult to explain. All i can think of is the rare occurence of electromagnetic errors in our hardware, but sure there's more to it? Or maybe the problem is bigger than i realized so far?

[danielgordon]

add another layer between the RGB-layer and the Life-layer that checks for locking patterns and rotate them randomly around the point with the lowest-energy state using observer-local delta time to reset a dead universe like the rest of us.

[jcasaubon]

I have a question about your note “copy when up|down stream all agrees”. Does this mean each node has to receive a signal from all up|down nodes and essentially tally to ensure all voted yes prior to flipping its own state? If I’m understanding that correctly the communication overhead increases as you move across the board as each flipped node becomes a node that needs to be checked by all other nodes in whatever direction we are moving in. So for example in a 4x4 grid moving upstream the first move is just checking one corner.. but the last move is checking the entire 3x3 grid below the upper corner.. 9x the number of checks.. if the grid was 100x100 we would have to check all the 99x99 grid or 9801 checks before being able to change the last node. Am I understanding the concept correctly? Great videos BTW.. not long enough really! 👍

[michaellittman]

Have you ever looked into the "actor" model by Gul Agha? (He was a visiting professor at my university back in the 80s and shared his gospel for how to think about computing.) He talked a lot about the important ways that synchrony and asynchrony co-exist. I can look for references if you want.

[drmedwuast]

20 minutes in and incredibly easy to follow. I can’t believe I actually seem to understand what’s going on.

Just one thing: how does the black and white universe relate to the redbluegreen universe?

[animowany]

I don't think you've considered one possible failure mode, what if a single compute element freezes? That is, its output state gets stuck? All of your synchronization methods fail in that case, the neighborhood of the frozen cell will never reach consensus for the next tick, but the cell won't appear broken, because you can't locally tell if it's frozen or waiting for upstream.
There's also the performance issues, this architecture is terrible if you want to do lots of operations quickly - you have to wait for the signal to fully propagate for computation to occur, which kills the advantage of asynchronous computation.

[hansbrackhaus]

Gotta give the video a dislike because dislikes are hidden and y'all need to complain about it.