New Project

[Rober-t]

Not an issue per se so feel free to close it out but thought I would share this project https://github.com/Rober-t/apxr_run which is an updated and refactored version of this repo. The credit for the good stuff still goes to the original author but perhaps others will find the project above easier to work with :)

[CorticalComputer]

Various life events took me away from my work in this field for a while. It is indeed a great idea with regards to using something like Gemeni or Claud to convert it to another language, I recently did something similar with a C library, and it worked rather well.

Due to there still not being substantial optimization and ability to run Erlang on a GPU, training is rather slow as compared to more optimized languages when using 1 neuron per node. On the other hand, we now have things like attention networks, transformers, autoencoders, substrate NNs... and those can be their own nodes, and thus allow one to evolve networks of autoencoders, transformers, attention networks, substrate subnetworks... this might allow one to augment and evolve further SOTA existing open source MLLMs. That would be interesting. But to do evolutionary or any other style reinforcement learning with such vast systems, would probably require running on something that is more optimized for speed and memory, unfortunately, something that runs faster than pure Erlang allows at this time...

The world is now moving towards agent based systems, where such agents can interface with software and hardware (sensors/actuators = MCP?) and even monitor themselves (exoself).

[CorticalComputer]

Have you seen the work of Sean Moriarity https://github.com/seanmor5 ?

• Unfortunately I have not. But I am looking into it now. I've been doing work in a different domain for the last few years unfortunately, and am not up to date.

Are you open to consulting work or collab?

• Sure, it's kind of what I'm doing these days in some sense anyway. I've developed and developing various UxV platforms, vision and laser guided, fire control systems... And I did use DXNNv2 to evolve a swarm controller that beat DARPA swarm challenge smart shooter algorithm, though that was now half a decade ago.

Will you be working on a new book?

• Yes. It will deal with developing a new TWEANN system, and autonomous platforms. But I must admit it is going very very slowly, due to work taking a substantial part of my time, which is not directly related to TWEANN or AGI unfortunately.

Do you teach/can I study any more of your material?

• Sure, I am always happy to discuss various topics, brainstorm ideas... you're welcome to email me. Though again, my skill sets these days do veer heavily into UxV development... I'm hoping that is going to change in about a month, and I might be able to go back to doing primarily R&D in computational intelligence and related applications.