madmax2976

The "issue" is whether or not we actually make choices or whether we are fated/forced to do what we do due to the laws of the universe. In other words, do choices really exist in a deterministic universe?

I have my own thoughts, but was curious what others thought on the issue.

DRFseven

So are you saying you don't think a mechanism that chooses describes a real choice? Isn't a bean sorting machine choosing beans? I mean there has to be a mechanism by which things get chosen, or no choosing would ever occur. I don't see why inevitability of outcome would mean a choice was not made; we're going by criteria so we're bound to have to go through a process of elimination which will "automatically" eliminate everything that doesn't fit.

crocodile deathroll

The Problem is a Human Brain may only run at a mere 40hz and the fastest personal computer may run at around 2+gigahertz , But the big difference is the brain has billions these processes all running in parallel. We need this because a the most sophisticated robot will not have a clue how to react if it encounters a tiger of a cow, it may well end up bolting from the cow and and attempting to milk the tiger <img src="graemlins/boohoo.gif" border="0" alt="[Boo Hoo]" /> . But will still kill us at chess.

crocodile deathroll

excreationist

You can also get neural computers that work in parallel as well but I think you can only get about a million artificial neurons in one at the moment.

they can tell the difference between the two by comparing the colours of them. The main problem with AI is that they can't understand natural language (e.g. English) properly. But in the future I think that if they teach a massive artificial neural network language then it will know it pretty well. I think there was also a link that someone gave about a computer that had been taught english for about 10 years and is as good at language as a very young child.

Well a couple years ago Kasparov was still beating the Deep Blue chess computer sometimes... maybe it's a different story now though...

[ December 11, 2001: Message edited by: excreationist ]</p>

Morte

Artificial HUMAN minds ?
How about attempting to make an artificial insect-mind first, then we will see about the 'higher' minds.
Maybe we need some artificial evolution to construct a mind similar to ours... or we might as well just copy a real one to harddrive.

- morte

excreationist

<a href="http://www.newscientist.com/hottopics/ai/whizzkid.jsp" target="_blank">Hal - who has the language skills of a toddler</a>
I found a link to the AI I was talking about. It autonomously learns, similar to a human, except that its only input is typed words, like children's stories. It hasn't actually seen the world though, but it learns about the relationships between different words so that it can use them somewhat competently.

<a href="http://www.newscientist.com/hottopics/ai/gasbrain.jsp" target="_blank">gasnets and artificial insects</a>
This uses artificially evolved creatures that act like insects (to some degree).

Yeah, artificial evolution or "genetic algorithms" are often used to set up autonomous artificial neural nets.

<a href="http://www.newscientist.com/hottopics/ai/clever.jsp" target="_blank">A "kitten" with 40 million artificial neurons</a>

"Neural networks must be fine-tuned to perform particular tasks. But no human programmer could write the software needed to refine a network as complex as the CAM (Cellular Automata Machine) brain. Instead, this will be generated using an approach that simulates biological evolution. Through random mutations and breeding of the "genetic material" that describes the structure and connections of the network, the program will be evolved over many generations to get the optimum design. Robokoneko will not be built until this work has been completed on a computer simulation of the robot cat."

<a href="http://www.newscientist.com/hottopics/ai/godofthenorns.jsp" target="_blank">Steve Grand Interview</a>
He created the "Creatures" games that had norns. (They're a bit like the creatures in "Black & White")

"COMPARED WITH ANIMALS, WHERE YOU WOULD PLACE THEM (the "norns")?
They're like ants. They don't have thoughts of their own and they are trapped in a sensory loop, where the environment tells their senses what to receive, the brain generates behaviour, and, much like us, that behaviour changes the environment, and so on . . ."

Now he's working on a neural net controlled robot called <a href="http://www.cyberlife-research.com/about/anatomy.htm" target="_blank">Lucy</a> which he is trying to give <a href="http://www.cyberlife-research.com/about/brainintro.htm" target="_blank">an imagination</a>. It will have to be taught how to do things, like an infant - it isn't programmed with advanced behaviours.

I think we'd make huge neural networks that have human-like intelligence before we can copy a human brain.

[ December 11, 2001: Message edited by: excreationist ]</p>

madmax2976

Choice seems to involve volition. The ability to think ponder and evaluate actions based on their potential outcomes. A bean sorting machine would not have the ability to choose as it just directs beans to a particular bin based on size or whatever. It would not be able to deviate from its programmed task.

Thus the applicable question for this thread would be: are we no more than bean sorting machines that are fated to do what we do because of the "beans" that come our way?

I course I believe there are several answers to this puzzle, one of which is that the laws of nature are not prescriptive but descriptive. The laws of nature do not "force" anything to happen. They describe what does happen.

Devilnaut

Laws of nature describe what happens as opposed to forcing things to happen? Huh? This makes no sense to me.

The answer you are looking for is that yes, we are no more than bean sorting machines in the sense that our thought process also uses real imput. I honestly don't see why this is a problem.


I think that you see it as a problem because you have some unnecessary baggage attached to your concept of the word choice. The best solution would be to drop it.. i suggest reading through BD-from-KG's posts again..


devilnaut

DRFseven

But what else is there to be?

We've got to either use criteria (bean holes) to categorize or just not categorize. Unless you think consciousness is some kind of a force separate from ourselves, there has to be some kind of a neurological mechanism to feed the input into the hopper and, in fact, that is what imaging techniques are beginning to show us.

Synaesthesia

That article on Hal is very interesting. I like the developmental approach to artificial learning systems, I think that has to be integrated into AI philosophy on a much broader basis. Still, it is possible, no, likely that we are centuries away from developing computers with the cognitive skills approaching that of toddlers.

“In contrast to children, who can take years to learn the basics of language, Hal can be trained in just a few days. This is due both to the intensity of the training and because the algorithms have no distracting inputs, says Hutchens.”

In other words, this system is still highly specialized. That is necessary but it makes humans very susceptible to the so-called Eliza effect, our tendency to anthropomorphize cognitive skills that superficially resemble our own.

I’m not trying to downplay the accomplishments of such research projects. Sophisticated studies of restricted domains is vital to the development of AI. However, it’s very easy to be mislead when descriptions like “understanding” and “the language skills of a five year old” are used. It makes a lot of people forget just how much conceptual ground has yet to be broken.