SFI Talk: Four the hard way: Computer design and living software 

hello, as usual thank you for posting. Nice job with the captions of audience comments! I've watched enough of your videos where I think I know where the talk is going. But, having the unpredictable questions from the audience is nice! I'm only 15 minutes into the video right now, so I'll keep going. I work on web applications on a day-to-day basis, so it's refreshing to have the assumptions challenged and consider, what if there were a better way?

it's nice how this talk was robust and re-constructed itself after each interruption from the public :)

Another suggestion I like to give, is that in your pervious video comments, and one of the audiences in this take, asked what kind of "demos" can be run and show future audiences, what living computing can do. I think besides our "localized/distributed scheme" in sensor network is one, but also a lot of embarrassingly parallel algorithms/schemes existed should be the easiest to be converted into this, and they will provide some obvious results to show the robustness and self-repairing features living computing excels. Personally I would love to see some type of Artificial Neural Network and artificial life hybrid based upon this to be implemented and demo. It's quite obvious that it's a very flexible structure to scale up neurons and make the parallel potential maxed out. (Or even in signal process with DFFT, perhaps?) However, I believe the real hardware needs to be testable before some performances comparison with existing systems can be made. (But for demos, it doesn't have to be) Finally, I think due to the nature of human perception and engineering issue, demo and first generation of testbed in 2D is OK enough, but I think the true potential with jump up significantly if going to 3D grids, and something very fundamental engineering problem will need to be addressed crossing to 3D. I wonder what's your opinions about these?

The commenters drove me crazy for some reason

First of all, This is a very interesting talk, and the interaction with the audiences gave this talk much more context, and certainly will help us one step toward the paradigm shift toward living computing (You don't have to convince me, I am already converted). In my own field of researches of sensor network (or what the media/industry like to generally them IoT/IoE), we face the exactly same issue of unreliable/unpredictable hardware/nodes, even unreliable connection between nodes, and these nodes are generally very simple in structure (what make things even worse, is that nodes might not just have errors, they might even be mobile, leave and join back unpredictably). And for a long time we already tried to devise what we called "localized schemes" in order for a group of mobile sensor/nodes to work together and build useful applications on top of these heterogeneous units (we already used a lot of bio-inspired algorithm to help solving issues we faced, like use ACO to discover routes in ad-hoc network). Concepts from this paradigm will be extremely helpful and interesting to be integrated into ours (like one of the new techniques we like to call Virtual Objects, which is essentially like your simulator but in a distributed fashion. running across multiple nodes dynamically, and a lot of time adaptive)

Another homer. Even I, a civilian, feel I'm understanding the contours of the problem. Now I'm just worried about those x-rays. When the Russian mob starts hacking our shiny new best-effort machines, what's to stop Boris and Natasha from irradiating them?!

Dave, this is awesome! With a large gradient of levels of operation, the small cpu + kernel does not need to know it all, that can be fixed in a level above. knowing that a box should be a box so it can repair itself. i imagine a vast array of risc systems. Paralella is such an idea to get started. and from a high perspective, does my local machine have to do it all ? can i outsource cpu usage to and from a cluster ? while ipv4 ran out, i am still waiting on my provider to rollout ipv6. but we do need a more flexible communication protocol. my computer could do processing for my phone.

Find this stuff very interesting, but surely questions should be held off until the end?

👍