The area of the wall that gets into my retina is too small.
Yeah, you move your eyes around a bit…but you know what you're actually doing?

You're anticipating the color of the wall.
You're anticipating that the whole wall is yellow.

Now, stop there and think:

But if I'm anticipating it, I'm actually not directly detecting it…which means (in some sense) I'm imagining it.

Here's another one:

You're running across the floor. Not just walking, you're running.

When you run the sensory motor signal is too slow for you to rely on it to guide you when you're running.

So what are you doing?

You're anticipating the floor, which means you're actually imagining it.

That’s really weird.

Now one more point, which is derived from the neural networks (deep learning…the thing that's in LLM’s).

What we discovered (way back when in 1996 by Geoffrey Hinton) was that how deep learning works is: Neural networks, they’re trying to detect a pattern in the world. They don't pick up the whole pattern. They take the data, they crunch it down to what's in common for all of the data, and then what they do is they generate a bunch of fantasies.

What the brain does is it fantasizes, and then most of those fantasies don't actually help you detect new patterns (so they get killed off) but some do, and so they stay…and you keep cycling.

That's how deep learning works.

You use the processes by which you generate behavior to recognize behavior in others.
You use the machinery by which you gesture to interpret the meaning of other people's gestures.

This is why you have a capacity for empathy and sympathy.
You use the very things that generate in order to recognize.

But do you understand what this means?

Empathy and sympathy are acts of imagination.
Acts of fantasizing.

And so now not only are you worried about “Am I just fantasizing the wall and the floor?” but “Am I fantasizing you when you nod your head and me taking that to mean that you’re tracking what I'm saying?”

But I want to stop here.

Look, science depends on this fact that a property of predictions is that they can be true.
What distinguishes predictions from other statements is that they can be true or false.

And this is why I don't like people saying, it's just a controlled hallucination:
Hallucination means it's not true, but when you run across the floor and you anticipated it being flat and it was true, and so you succeed…

So let's remember that immediately talking about hallucination is missing something here.