Aaaahyou are the guy who proposed HelixNet, the names sometime blurr on reddit. But reading your prompt it is clear you are very smart and not afraid to explore new ways.

I know my praise seems generic and there is lot great people around but I really think you are among top people.

Did you used GPT4 to generate dataset or other models?

Yes such solutions already exist, but compounded with less intelligence than humans have results are somehow underwhelming but you might try.

Biggest problem with ai now is it lack real memory. And nobody including openai despite their claims cracked this problem. So in essence your ai would be only “actors” doing what you tell them to do each scene. Anyone can do this today.

To make it work at least partialy you would need to outsmart whole world in building at least partialy working external memory and perfect ai oriented search in it. Projects for this exist, they use vector databases and such but still nothing really working. You upload to openai “gpts” 1Mb of text and it already start confabulating and halucinating instead of really searching it for answers.

I mean this is trolling right? :)

But if you actually have money what about sponsoring with little of it some opensource guys who do real cool stuff as datasets or finetuning, can point you toward them.

This is thing I was talking about some months ago after I first saw GPT4. Glad someone actually worked in this direction as ideas are somehow cheap these days :)