quote:

---

Original post by UlfLivoff

quote:

---

The basic actions that the humans can perform are JUST and ONLY move the muscles of the body. And not more, you cannot perform any other action, nor learn any more action during your life.

---

This is a widespread understanding and also what I learned in the class "Brain Physics"

---

Popolon''s statement that I disagreed with was "I don''t agree with considering the use of a tool as learning a new action." He says that using the muscles was the _basic_ actions that the human body can perform then went on to say that the basic actions we perform were the only actions we can perform and that all tool use, complex or not was still just the action of moving our muscles. Our genotype maps onto our phenotype in many epigenetic and epistatic ways. Richard Dawkins argues that the extended phenotype, i.e. tool use, beaver dams, certain soft shelled marine creatures using shells etc. was a natural extension of this and that you could not split one from the other without drawing arbitrary lines in the sand.
As you''ve stated, whatever actions we can perform to change our own state in a vacuum (i.e. without environment) might be considered our basic actions. But tool use and environmental interaction are certainly extended actions as much as they are part of our extended phenotype.

Ah,I see.
But that pretty much boils that discussion down to the definition of the word action, right?

In some situations, using a hammer is a action (for example in a computergame), but in a biological way it's not an action..

I just felt like pointing out that I can manipulate my enviroinment without using my muscles

Anyone having views on my other points? I'll be glad to hear your opinions...


[edited by - UlfLivoff on July 10, 2003 9:35:47 AM]

Cyril:

quote:

---

IMHO, the first step to "learning AI", is to make it generate code on the fly, meaning that for each new thing discovered, it should create some new actions, and thus, some new code. Hard-coding AI is a terrible mistake, once again, IMHO.

---

You''d be surprised how many things that are hard-coded in our brains.

Face-recognition, reacting to a screaming child and fear of snakes just to name a few.

Psychology books are full of examples of ''hardcoded'' stuff

Ulf

The original poster apparently hasn''t tried out the new counter-strike bots. Granted, this isn''t in the realm of general AI or "real AI" as he puts it, but the bots do exhibit (an illusion of) some learning behaviour.

For example, when the bots are being defeated in a certain area of the level they will be less likely to go there in the future. They will also shift their overall style of play between defensive and offensive depending on a morale system. The more skilled bots don''t only aim better, they''re also more aware of the surroundings and "know" which areas are good to hide in (both by themselves and the enemy).

Another important point is that they have the same sensory inputs as the player, rather than typical computer-like omniscience. This alone makes them act more human because they can be surprised or ambushed.

Does anyone think the original poster even bothered reading the responses to the claim made? I don''t see evidence of that. However, the discussion among the rest of the thread has been interesting and enlightening.

quote:

---

Original post by UlfLivoff
You''d be surprised how many things that are hard-coded in our brains.

Face-recognition, reacting to a screaming child and fear of snakes just to name a few.

Psychology books are full of examples of ''hardcoded'' stuff

Ulf

---

Well, if you''re telling us you read about it in a book somewhere, then it must be true!

Unfortunately, I am not afraid of snakes.

good for you. And now for a little excercise:

Consider how many people that are afraid of spiders.
Then think of how many people that are afraid of cars.
Compare the number of people killed by spiders each year to the number of people killed by cars each year.

One fear (stronger or weaker) is 'hardcoded' in the brain in most people, the other is not.


[edited by - UlfLivoff on July 10, 2003 6:51:18 PM]

quote:

---

Original post by MikeD
A power of ten out again and I even own that book

---

Hey, an order of magnitude error isn''t that bad... it''d certainly be acceptable in astrophysics!

quote:

---

Original post by MikeD
Out of interest Timkin, do you, off hand, know how much biology I would need to know to take a taught Masters in Neurobiology?

---

How does one quantify a volume of knowledge? The main issue would be your understanding of the literature and your ability to a) identify the relevance of your research; and b) place your research in the context of other research in the field.

That really depends on you. Personally I have no doubt that you could learn the requisite background material... the question is, do you really want to???

Cheers,

Timkin

Within the academic AI community at least, action is defined to be any cause of a state change initiated by an agent that isn''t explained by the transition laws of the environment. Typically the environment transition would be described by s(t+dt) = f(s(t)). An action is anything that causes a different state s''(t+dt) if the environment started in state s(t). This might be represented as s''(t+dt) = f(s(t),a(t)).

In terms of probabilistic representations, the environment transition function would defined by the conditional probability density function, p(s(t+dt)|s(t)), at least for Markovian processes. If you understand probabilities, you''ll recognise that p(s(t+dt)|s(t),a(t)) is a very different beast.

Remember though, this is just the academic (AI/scientific/engineering) definition of action and may differ from the psychological or physiological definition of action... which I don''t believe I''m qualified to comment on.

Cheers,

Timkin

quote:

---

Original post by UlfLivoff
good for you. And now for a little excercise:

Consider how many people that are afraid of spiders.
Then think of how many people that are afraid of cars.
Compare the number of people killed by spiders each year to the number of people killed by cars each year.

One fear (stronger or weaker) is ''hardcoded'' in the brain in most people, the other is not.


<SPAN CLASS=editedby>[edited by - UlfLivoff on July 10, 2003 6:51:18 PM]</SPAN>

---

That has an easy explanation, people are used to see cars and drive them, but people aren''t used to deal with snakes or spiders. That''s definately not ''hardcoded'' in our brains.