Andy Halliday

It's all online with MIT professors and so on.

And one of the terms that we very, very rarely have used

on this show that was really in the core of understanding how AI is trained is epochs.

I don't know if you've ever run across that term.

Well, when you do training runs through a deep neural network, you set, in the code language that you use to do this, you set something called an epoch, which is, OK, I want to run 100 epochs of this thing, which means it's going to do the same thing 100 times.

And so that gets to the point of, how are AIs trained?

How do you do training with an AI?

And it's not as simple as, oh, just dump all the, you know, dump all the data in, give it access to all this data.

There's actually a complex routine of steps that are involved in doing that with a real neural network and training the weights, if you will, the representations that happen in the model.

This thing that we're talking about right now is not going to touch on that kind of jargon and terminology.

It's much more basic and much more.

It's easier for you to understand the essentials of AI.

So I think it's a great thing.

And I'm looking forward to doing lesson number two today.

that's like what Beth will remind, you know, when she's on the, on the, it's interesting because the AI will behave in its responses.

When you note to it and draw to its attention that it's made an error, it will seem embarrassed.

It'll be very apologetic, but you know, that's not covering and papering over a lie.

It's just, Oh yeah, that that's just what came out of my routine.

And I, you know,

Sorry, not sorry.