Why there's so much AI in this museum

MindStretched was built with AI, very deliberately.

A few years ago I had to journal in coaching school because apparently writing down your thoughts is good for you. There are studies, therapists recommend it, coaches recommend it, millions of people swear by it. I tried it and thought, this is dumb. It felt like being my own stenographer, except more painful. The thoughts were already perfectly clear in my head, why am I writing them down?

Later I asked AI about it and got all the reasonable explanations for why journaling helps people - processing your thoughts, organizing them, putting feelings into words, gaining perspective. Fine, but none of that explained why something that supposedly helps humans think felt completely pointless to me.

So I kept asking questions. How do thoughts actually show up for different people? Is writing things down different for some individuals than for others? Does the established explanation actually survive what we currently know about how different brains process information? That was much more interesting because human minds are way more unique than most people realize. There is no one size fits all best thing to do at all. 

And this basically became how I used AI. I would notice something in real life, ask about it, get an answer and very often say, But that doesn't make sense. Or, that can't explain this other thing. Or my personal favorite, But is that actually what the latest neuroscience says? Then we'd look at the evidence, which would create another question, which would turn up a counterexample, which would wreck something we thought made sense twenty minutes earlier, and now there were three new questions and WHY ARE WE STILL AWAKE?

Observation → AI → objection → evidence → new question → AI → counterexample → revision → another observation over and over and over.

Eventually the questions started connecting to each other. Then the connections started looking like machinery, and somewhere along the way we ended up with this whole ridiculous contraption and apparently needed to build a museum to put it in.

There’s no doubt I would not have built MindStretched without AI. Not more slowly or twenty years from now. I wouldn't have built it at all. I'm one layperson who happens to have a ridiculous number of questions about humans, and no individual person has enough lifetime, expertise across neuroscience, psychology, evolutionary biology, decision science and everything else we wandered into, or frankly enough patience to spend the rest of their life tracking down every question that appeared along the way. I certainly don't.

AI changed what was possible. I could ask a question about something I noticed at lunch, find out which fields of research might have something to say about it, get pointed toward evidence I didn't know existed, challenge the answer, follow the contradiction somewhere else, then ask a completely different question because now something else didn't fit. To be honest, there is an addictive quality to having an answer machine sitting next to you at 2:00 in the morning. One answer doesn't finish anything, it gives you five better questions.

Which is also why I've never understood the argument that AI sometimes makes things up, therefore you can't use it to learn. Of course it gets things wrong, so do people (but people do it way more often). If something looks suspicious, argue back, ask where it came from, look at the evidence. Ask the question another way, ask another AI, find the research paper. Sometimes the answer survives and sometimes it falls apart, which is useful information either way. I didn't ask AI for the answer, I used AI to keep the question going.

The long in-depth Q&As eventually generated the Machine we currently have. We only had to crash and burn about a dozen other versions of the machine to get here. We (me and AI) kept asking whether our explanations survived what we were observing and what current scientific evidence could support.

I know there is no view from outside the machine. Everything I notice goes through my own machinery, and AI isn't floating above humanity in some pristine state of objectivity either. It was built from the enormous pile of things humans have discovered, written, argued about, gotten right and gotten spectacularly wrong. My real advantage is that I often see things differently than most people, which brings on some fantastic arguments between human and machine. 

So yes, AI fingerprints are all over this museum. They're supposed to be.