I thought about this this morning (January 2, 2024).

First, caveat that I am basically speaking entirely out of my wheelhouse and area of any vague sense of knowledge, so there’s a really good chance that what I write here could be fundamentally incorrect.

That said…one issue I see with how the Mega Corp is approaching AI is that, like everything else, they’re taking a monolithic approach. Which, like, awesome…never interrupt your enemy when they’re making a mistake or something.

As an example, ChatGPT or Baird is like one absurdly huge program that requires massive amounts of data and computing power to function. Which, the end result is pretty cool, but do we necessarily need a singular AI to know everything? And by everything, I generally mean stuff posted on websites that were used to train it, so like a gravel brained version of everything.

Instead of having one massive model, my perspective is that we would likely be better suited by approaching this in…drumroll please…a decentralized model! Which, like, I’ve said a lot, what a surprise!

So, imagine if you had some type of meta AI agent that you would send your query to…what’s a really good cake recipe or whatever. Then, that meta AI agent would then consider its own knowledge for all the various narrow AI agents, perhaps one designed just to be the best fucking cake recipe AI agent ever, and then go and bring that query to them. Or, heck, maybe there is more than one AI cake model and they can chat with one another to improve their offerings.

So, basically taking the Public Cloud concept where AI developers and model creators can plug in their own specialized AI which could in turn be available to this larger network, with its usage rewarding its contributors rather than whatever Jekyll Island Frankenstein monster we get from the soulless losers presently in charge of the world’s closed source AI nonsense.