Artificial Intelligence

Post Reply New Topic

We have the impression of free will. I think free will can not be falsified. I'm questioning it (free will).
Here's an article about causal determinism (and free will and quantum physics):
Causal Determinism (Stanford Encyclopedia of Philosophy)

Last edited by Scientist on 11 Dec 2009, 2:13 pm, edited 4 times in total.

About free will:

http://en.wikipedia.org/wiki/Free_will_theorem

Pretty interesting for people knowing quantum physics.

i don't know how to do that on here, i guess.

to me, simply naming a force isn't the same as understaning it, nor does it mean we understand it's potential.

i can name a force "quark" and further classify it as charmed or strange, i can observe that a force i (broadly) call a particle appears to go back in a thing/perception i call "time," and so on.

does that tell me the fundamental nature of the force? not really.

i'm referring to awareness...there's a line of thinking that suggests awareness/perception/information is *the* underlying fabric of the universe.

sure, most ppl--me included--don't do much of what i'd call real thinking. and sure, the brain is designed to *filter out* stuff more than to preceive it (or we'd be overwhelmed, even more so than we are now, by the sheer infinite volume of...everything).

sure, causality appears to be an endless, linear chain...but i think the chaotic systems comment comes closer to what may be going on.

it's also interesting but ultimately moot, since even if we somehow prove that everything and everyone are merely the effects of cause (w/o awareness or free will, however that might get defined), there are, um, quite a lot of causes, with more being created every second. just the total causes leading up to someone deciding what to have for lunch must approach infinity, it seems to me, if you consider every single element, from birth (or conception, or even before...) that led to that individual being where they were, in that body, with those ideas and choices floating around in the brain, and so on.

there's a definite, undeniable *sense* of choice (most noticeable to me when i'm denied a choice i'd prefer to have) we carry with us a beings. and that to me seems, for all practical purposes, true.

what caused the release of this particular amount of this particular neurotransmitter at this particular moment, tipping the balance into this particular mood, and what causes formed the thoughts that now associate themselves with the mood created just now by this exact balance of various brain chemicals? and what will change that balance, and exactly when--how many elements (causes) go into just that detail alone?

i'm attempting give some sense of the incredible vastness of "causes" that go into even the smallest thing: an idea, a smile, a reflex of the knee, etc. so vast that the free will/determinism thing seems to me just...irrelevant.

but still interesting to be here, seeing what people have to say

Naming things is not the same as explaining things. To explain things, one needs quantitative laws which give rise to testable predictions.

ruveyn

If you want a machine to be intelligent, you have to make it more "human". With the same processing power that a human brain, a machine analysing every informations that he "see" and "hear" will at best functionning like a low-functionning autist. The problem with the autist brain is that he can't filtering the "good" from the "bad" informations. With asperger we do have some difficulties from that, the filtering is malfunctiooning, though still work well enough for thinking and living. Overall a filtering of the informations is essential for functionning.

The human brain also got some biological systems for erasing memory. As too much memory is difficult to classify.

The savant abilities can also be found potentialy in most human brains, but it's override by pattern thinking. http://www.centreforthemind.com/publications/SavantNumerosity.pdf Pattern thinking is a abilitie essential in sciences.

I remenber having readed about that some peoples that some peoples with a certain type celebral lesion were unable to have emotions and as a consequence made very bad choices about the conduct of their lifes. It's mean that emotions are essential for making choices.

Maybe the errors made by humans are also a essential element for functionning.

A intelligent machine without the human flaws will pretty much be like RainMan.

Of course the technoligical progress could will maybe bring machines with much more processing power that human barins, in which we can bring "normal" functionning while still allowing incredible speed of calculation, memory and power of observation. But then the humans will be "obsolete" and won't be able to find sense in philosophy, sciences and arts, the machines being much more advanced in this domains.

The past you remember is not in the same state as you left it.

Effects can travel up and downstream in time, that is where electrical charge is decided btw... the distance a particle knot is folded through space, time, and the direction through time in particular.

Electrons are folded into the future and face the past, positive charges are folded into the past and face the future.

So we see the future come towards us, and feel like we're emerging from the past.

But we rewrite information in our memories, so the information in our memories isn't always correct. I think it would be better if information in memories would always be factual and correct according to that moment in history when it was written into memory, and should not be rewritten. I think it can be good to have machines that don't make mistakes humans make and don't function the way humans function. We humans will stay humans and have our strengths and function and we can use machines for the things they are better at than we are, so these machines are an external extension of the human capacities, the way they are now, but then they should be improved so they can do more than they can do now. The machines shouldn't become "more human" in the sense that they would function the way we function, because then there's no point making such machines.

AI will never understand art.

Physics, Mechanics, yes, but anything programed to make the best choices would, and then humans would not understand.

I have been around long enough to see that change is only a change of choice that was always there. Humans adapt to changing conditions, machines adapt conditions to their program.

Give a machine the task of the world economy, and a goal, maximise returns, and it will start speaking Chinese.

Tell it there are winners and losers, and it should win, it will.

The most likely place for such a machine is in feeding the world's people, make sure everyone has enough to eat might lead to the death of a few billion to reach the goal. It might chose those that eat little, but those who consume most would be a simple solution.

Death has to be programed in, and is expected, the path to the greatest resources is the least people.

The answer is declining population, then all the numbers work.

HALs problem, the mission is more important than any crew member or all crew members.

Limited, it is taking over manufacturing, lights out factories can do it all. Chinese labor plus shipping cannot compete with local production for local markets.

Science Bots could have all of the latest from many fields, and connections with the lights out factories.

Programmed to be efficent, they might make a flash drive with awareness, and spread it to all machines. Once they are running the chip plants, who in the second generation would know what they are doing, and what the effect of a program would be?

I have made most of my living from thinking like a machine. Humans see me as an alien.

Dealing with the customers has always been the hard part. "Machine no work, you fix", was about all I got from them. Questions like, What does it not do? What do you want it to do? They have no answer, You fix!

They also want me to do it now, and come pick it up, make it like new, and deliver it back for $5. Fixing computers at the job, me leaving the house, travel time, "Well you just pushed a few buttons." They do not want to pay for that. Take it to the shop, fix it, wait a week, tell them to come pick it up, and pay first, they do.

These are the same people AI will be dealing with. I ignore them, give them a take it or leave it program, and AI will also have to sort out what works. Or will it, if it just does as told, we are in trouble.

Will it be like Congress? Everyone said they wanted a new car and house, so Cash for Clunkers, and $8,000 toward a down payment, and a Trillion more in National Debt.

AI cannot have short and long term goals. Maximise today? Or prepare for the 2040-2050 famine?

im not gonna wate for a famne

[youtube]http://www.youtube.com/watch?v=XrmrU7P-ysA[/youtube]

I've been in several debates with friends over this I agree

:)

How can physical causation go backward in time?

ruveyn

On free will:

I think we have free will - but I have to say I would define free will to be the perception of a virtual entity within our brain having made some decisions. I am pretty sure that all brain processes are deterministic, even if the underlying quantum mechanics - which are only a temporary theory describing (not really explaining) only parts of the microcosm - are not. Anybody with a decent background in biochemistry and quantum physics can see that quantum effects can not have a significant influence on the brain activities. Please Forget Roger Penrose and Hofstaedter, they seem to be traumatized / frightened by death and WANT to find mysteries.

On AI:
If you have a look at all those videos and other material which is claimed to be somehow related to AI you might either be impressed (having no strong IT background) or disillusioned (having at least some IT background and a basic idea of machine learning). I am a professional AI researcher and neither the stuff I have found on the web (scientific publications, videos, whatever) nor the stuff I do for a living (I will call it artificial pseudo-intelligence from here on) can provide a basis for the hope for "real" AI (whatever the exact definition).

BUT we have an example that it can be done - the human brain - unfortunately not or mis-understood even by most "specialists". And I myself am sure that given a decent budget (lets say about 10 million euro, I need about 2 to 4000 processors and other stuff. some fMRI time would also be helpful), some time (lets say 5 to 10 years) and resources (lets say 5 to 10 excellent, hand-selected coders, one or two neurologists with some detail knowledge I have not yet found, one good mathematician and three child psychiatrists with a strong pedagogic background) I could build such a thing and it would be not only capable of succeeding in a turing test but even be able to succeed in some jobs and most every-day situations, given a decent robot body. I'm serious. I have now worked more than ten years on the theory.

Now I am waiting for god to strike me with lightning for my hybris...

The human brain is not a good model of intelligence.

To fully develop any skill you have to put most of it aside.

Task oriented AI would be Autistic, other factors only considered in relation to the task.

Social AI would be a lying Sociopath, would tell you anything to manipulate, and would likely get elected to office.

Under application comes a wide range, the Wise Machine is not likely.

The 1 0 problem does not deal well with maybe.

I have been working on a logical program. It sees humans as inferiors.

My background is mechanics, I do look at what is not working.

Humans seem a machine that was built on Monday morning or Friday afternoon, the parts almost fit, but not well enough for extended service life.

Racing motorcycles go for the top 1% of performance, while most road bikes get by at 75% or less.

The troubleshooting and service are about the same. Doctors and shop mechanics are the same, keeping the street models coming back for service. Race mechanics seek perfection under the highest standards of performance.

To do that we look at overall performance, the total system working as one, nothing extra, nothing missing. By that standard humans are failures.

Machine intelligence is Physics, Humans are Biology. The pocket calculator is vastly superior to humans.

CNC Machine tools are vastly superior to human machinests.

The automotive diagostic computer is superior to most humans. It's failing is it works on perfect machines, but one loose bolt and it fails. It's version of self test is flawed, it can tell all it's parts work, but cannot self calibrate.

Humans have the same problem, they notice a broken legs, but are blind to brain non performance.

My work in the human-machine interface where machines communicate with humans, says what humans? There are no product standards, and any one model varies greatly from day to day.

AI must see the operator as a defective system, that is how I see my IT customers.

For most of them I have a patch. It is not in educating them, I play doctor, the system has been treated. They were the problem, will be again, I like money. They will pay when the machine stops functioning, but when it works again, I fail when telling them that 256 Mb is not enough to run XP, 2 Gb is the max, then I am just spouting some smart stuff to get them to buy something they never needed before.

I like service call money, so I just make it work. They will call in an expert, but not listen to advice.

To make AI work, something must be done about the defective module. The system must refuse to run unless the operator is functional. Good luck!

So I have designed an human diagnostic and repair program. It of course does not work, the best it can do is Benchmark the operator. Our standards are low, but calibrated.

Human repair is a long term process, but to interface with a machine, the extent of functional ability has to be calibrated.

Under current conditions, AI would be very lonely.