Dr. David Eagleman

I don't think the AI could learn that about visual images because when it generates the pixels, it's doing this, you know, this magical stuff under the hood where it's deciding which pixels and how they diffuse together and, you know, mix the image.

But it doesn't know how to read that image like, oh, yeah, the way this is and blah, blah.

That'll really appeal to Steven.

It's not seeing the image except as a bunch of pixels.

Okay, so I totally agree with you, but let me just mention one thing, which is fascinating, which is that often the way it's doing it is not at all the way that a human would do it, which might be fine for our purposes, but the data and the way that it's picking up on it, it might be something about, you know, how much, I'm making this up, how much green was in the YouTube thumbnail image or how much red or whatever, whatever the thing is, or just noticing that there's big font versus smaller font or whatever, right?

The next time you try it, it says, oh, yeah, this thumbnail is going to be great.

And it's some ridiculous thumbnail that doesn't make any sense to you as a human, nor to your fellow humans.

But it might say, oh, yeah, this would be great because it's judging things on very weird dimensions that we can't always see.

Quite possibly, but here's the interesting thing.

Human art constantly evolves and all AI is trained on is what has been done before and what has worked.

And so if I asked it, let's say we composed five different songs and said, hey, AI, which song is going to be better?

It's going to say something that's right in the middle of the distribution of popular songs.

But that's not what actually makes it next year and the year after.

It's new things.

It's new twists that nobody has seen before.

That's what we love.

That's what we seek as consumers.

And so because AI can only be trained up on what already exists, it's never going to get the new thing at the edge.

But that's the human selection part that we're seeing over there.

If you asked, okay, out of all these dots, which do you think AI is going to be best?