[Don Patrick]

... You may want to read a biology or neurology book, because it's well known that stimulating the pleasure center of the human brain functions as behaviour enforcement. And there have been many AI experiments where this principle was applied by the various geniuses we've had since the computer age. It may be a revelation to you, but it's common knowledge among scientists.

[LOCKSUIT]

---

[ivan.moony]

Not everyone has the same taste... One likes blonds other likes brunettes, there is a bunch of different tastes. I think something magical is happening, some life force that we don't know much about, especially with girls. A touch of life, I'd call it.

[LOCKSUIT]

---

[ivan.moony]

The problem is: how are actions defined? And how to measure which actions can compare and which is better? It is not that easy when you imagine a tasteful browny in your mouth to construct a set of actions that make the bowny melt on your tongue in reality. Especially if there is no browny in the house and all baker stores are closed.

Actually, there is a way, but it is a bit complicated.

[LOCKSUIT]

---

[ivan.moony]

... - something favoring a shape of balls is a trigger, to spike and save actions just done & its linked-senses just gotten.

How to produce those actions in a first place? And how to tell the computer what's attractive and what's not?

P.S.
Sometime is attractive even to hit the wall, when we don't want to i.e. fall into volcano

[LOCKSUIT]

---

[ivan.moony]

How do you define "pain"? What the algorithm should like and what should avoid? This is a very important question and I don't have an answer. How to universally define what to like and what to dislike. The world is a lot more complicated than just a set of walls.

[LOCKSUIT]

---

[ivan.moony]

But what to pre-install? That is not such a trivial question. If we knew that, we would have a plenty material to build on the future work.

[LOCKSUIT]

---

[ivan.moony]

But to build a robot you need to program into it everything you want it to exhibit. There is not magical programming command "feel what humans feel". You need to tell it:
* when you hit the floor, that's negative;
* when you break a wire, that's negative;
* when you hit the wall, that's negative.

But these are just three examples and there is an infinite number of them, one combined with another. So we need a universal function that takes values from senses as parameters and gives yes/no answer, depending what the parameters are. It is a very complicated function and no one knows it.

And there is more than yes/no, Look at this:


Where to map what input from what senses? It looks impossible to do such a task, yet we are trying to do it. Ok, I'm in, but we need details: what maps where and what to do after what.

[LOCKSUIT]

---

[Art]

OK...with all due respect, kindly show someone or all of us this AI / AI Animation so that an opinion of it might be formed.

It's one thing to talk about something but as the expression goes, The proof of the pudding is in the pie!


Thank you!