135. AI AIC?

We are trying to get into understanding the purpose behind the overture that was sent and kind of get a sense for what it's looking for and why.

But we need to do a little legwork ahead of time.

You know, we kind of chatted before hitting record about what is AI and what are we talking about in the context of this particular overture.

So maybe start real low level.

One of you guests, maybe Frank or Tori can jump in and give us a sense.

What are we talking about when we say AI? It's not Apple intelligence, right? They threw that on there late in the game.

It's something else, right?

It is.

And I hope I don't hurt anyone's feelings by changing the scope of my answer from too high to too low in the middle of a sentence.

When we're talking about AI, we're talking about... what is commonly referred to as large language models.

These large language models are effectively a really big spreadsheet, Scott.

And that spreadsheet has weighted balances on how to respond to tokens that are received via a prompt.

So when you're dealing with something like ChatGPT or Claude or Grok, any of those things, those are really just a software wrapper around a large language model that feeds repetitively the tokens of text that you are giving it.

And it tries to come out with a response that it believes is the most appropriate response based upon those weighted tables.

And so what it is trying to do is process the information that you are giving it and providing you a response that is the most appropriate or the most helpful response that is possible.

One of the interesting things, though, about these weighted tables is that there's no certainty.

So no table has 100% response rate.

So this is why you see some discrepancies, some differences.

This is why you can see some of the talk about AI hallucinating or deceiving.

It's because these weighted tables are not deterministic.

It means when you put the same prompt in, you won't get the same answer every single time.

This is helpful.

in some regards because it makes it sound like you're talking to a human instead of a computer.

If you got the same response every single time, you would know you were talking to a computer.

I think it's a little bit ironic that intelligence is being defined by variable returns in response and But that's kind of where we are living with this tech is a large language model built up system with software around it, whether that be something that generates text or generates images or generates music.