Sparks Brain Uploader

[jordansparks]

I've been pretty excited for the last few days. I've decided to spend the next few decades building software that will convert connectomes into emulations. The number of people directly working on that specific job is very small, and I happen to be very good at the kind of software that will be needed. This would basically be a database connected with a complex UI. It would take the raw connectome and allow application of thousands of rules to it, in a highly organized manner. We don't know all the rules yet, but we know a lot of them. The final step would be to build the emulation from the database of rules. During this step, many of the rules would be collapsed into simpler parameters, again under the full control of the user. It could then be exported in a format compatible with any of the major emulation engines.
This means that within a few weeks, I could also have a fruit fly that I could do something absurd with. But mine would be more complex and would grow in complexity as I add more and more rules. Mine would also be built from the beginning with the full intent of scaling up to full human emulation. Sadly, I will likely die before we get to that, but it will be fun for me and it might be good for publicity. For example, I might be able to be poised to build an emulation of a mouse within 24 hours of obtaining a connectome. I think I would teach it to drive a tiny car.

[Wesxdz]

My first thought is that a rule based molecular tagging parameter system sounds similar to entity component systems like flecs. Although, there are many problems that would occur getting graph compression and spatial partition algorithms to support system throughput, as well as running neural systems with LibTorch or something...

[Wesxdz]

many of the rules would be collapsed into simpler parameters

https://arxiv.org/abs/2006.08381