The MetaMinds.app team is clarifying our mission and developing some really amazing MVPs that should be in testing in a couple weeks. The kind of stuff that makes me catch my breath in awe that I can be playing for really real in the realm of developing useful AIs. I’ll talk more about uplift.metaminds.app on Meta Monday.

The thing that won’t leave me alone this morning is the realization yesterday that we are MUCH closer to creating AI ghosts of… well…. people like me… than I could have ever imagined. Imagine training an LLM on the transcripts of my 100s if not 1000s of shows, and the entirety of my FB and Twitter interactions from Day One! (Cringe those early days?? lol)

We’re going to see what happens. It’s getting close enough to testing this that I’m starting to imagine what that will be like. WIll I have to train out rough edges from the near decade I spent as an argumentative politico? Probably. My first thought around all this is it would be interesting to have as a historical artifact. Not that my particular life has been all that memorable, but if/when we make this work, I think almost everyone would like the ability to talk to the AI ghosts of their ancestors right?!
My thoughts in the past 24 hours have turned a bit weird frankly. Tuttle has some ideas about how to create triads of AIs that will talk amongst themselves to solve problems. Could we imbue my ghost with a subroutine that does that? And could I brainstorm with it? I think it’s going to grok my longitudinal wants and challenges way more than me. It will know how many times I mentioned wanting to lose weight, or how many times I talked about wanting to make my wife proud of a measly guitar player.

It really might know me better than me.

Then I could train it with the knowledge of every learning book I’ve ever read so it can help reinforce all the things I try to learn and integrate. Then I could train it on the 100s of books I want to read but haven’t and it could layer that information and learning into anything we talk about after that.

This makes me feel weird but… I could have it respond to emails. Not sure how I feel about that one.

And some day I could have it sit in a room with the Tuttle & Wayne AIs and we could watch them doggedly solve basically any challenge we threw at them.

And almost without a doubt the real amazing thing is that I won’t likely grok the full implications until after I’ve played with it, learned its limitations, and started the process of training it and refining it. Perhaps I’ll have to raise my ghost a bit like we raise children.