Really OT: AI for Planning

[JeffroK]

1) I dont want to play a computer that learns and therefore is smarter than me.
2) Replacing computers thrown out the window gets expensive (I has witnessed this[:D])
3) How much does a CRAY cost, I know computer prices have dropped and the North American Lira is almost worthless but shipping charges are in AUD!

[herwin]

ORIGINAL: JeffK

I dont want to play a computer that learns and therefore is smarter than me.

Not in the cards immediately. Wait four or five years. [;)]

My colleagues are telling me not to retire, but to work on this--there are about ten good years left in me. [8D]

[JeffroK]

I wasnt being serious, I hope to hell you aren't

[herwin]

ORIGINAL: JeffK

I wasnt being serious, I hope to hell you aren't

I'm damn serious. I was recruited to do a neuroscience PhD late in life by a number of names who had seen my work in non-linear dynamics and evolutionarily stable strategies in non-zero-sum games with information collection. The idea is that the computational approach to AI has failed because there's something about biological intelligence that the computer scientists have never 'got'. Training someone with a solid background in computer science in the biology was Plan B. AI for big, complex games is one of the toy problems I work on.

[witpqs]

ORIGINAL: herwin

Wait four or five years. [;)]

How long after that until CP?

[vonTirpitz]

ORIGINAL: herwin


The idea is that the computational approach to AI has failed because there's something about biological intelligence that the computer scientists have never 'got'.

I've concluded that the failure of computational systems was due mostly to the fact that they are, so far, inherently designed and limited by initial conditions in both hardware and software that can never alter their deterministic characteristics.

Biological systems currently have the advantage of passing massive amounts of "genetic" or inherent knowledge into future generations that ultimately create a diversity of systems that are far more advanced than the "original" biological circuitry and programming.

Creating an artificial system that can reproduce and pass along as vast an amount of "data" and enhancements in both the hardware and software and you would theoretically have an advanced generation of computers that are basically similar in design but inherently different enough to provide a level of intelligence that rivals biological systems.

In theory, this evolution can happen much more rapidly than it did biologically because it isn't confined to the same relatively long generational life cycles that biological systems are.

Ultimately, you would definitely have synchronization problems running a simple program like AE on these systems. The forums are proof of that. [8D]

[herwin]

ORIGINAL: vonTirpitz

ORIGINAL: herwin


The idea is that the computational approach to AI has failed because there's something about biological intelligence that the computer scientists have never 'got'.

I've concluded that the failure of computational systems was due mostly to the fact that they are, so far, inherently designed and limited by initial conditions in both hardware and software that can never alter their deterministic characteristics.

Biological systems currently have the advantage of passing massive amounts of "genetic" or inherent knowledge into future generations that ultimately create a diversity of systems that are far more advanced than the "original" biological circuitry and programming.

Creating an artificial system that can reproduce and pass along as vast an amount of "data" and enhancements in both the hardware and software and you would theoretically have an advanced generation of computers that are basically similar in design but inherently different enough to provide a level of intelligence that rivals biological systems.

In theory, this evolution can happen much more rapidly than it did biologically because it isn't confined to the same relatively long generational life cycles that biological systems are.

Ultimately, you would definitely have synchronization problems running a simple program like AE on these systems. The forums are proof of that. [8D]

I don't think genetic programming will solve AI. I think the problem with the traditional approach to AI is that its discreteness is a bad match to the real world. I've never found a number or an abstract symbol in the brain. You can prove a computer simulation of a bounded isolated dynamic system will collapse, often quickly, to a limit cycle, but many real-world systems never do. And biological brains can cope with that.

[witpqs]

ORIGINAL: herwin

ORIGINAL: vonTirpitz

ORIGINAL: herwin

The idea is that the computational approach to AI has failed because there's something about biological intelligence that the computer scientists have never 'got'.

I've concluded that the failure of computational systems was due mostly to the fact that they are, so far, inherently designed and limited by initial conditions in both hardware and software that can never alter their deterministic characteristics.

Biological systems currently have the advantage of passing massive amounts of "genetic" or inherent knowledge into future generations that ultimately create a diversity of systems that are far more advanced than the "original" biological circuitry and programming.

Creating an artificial system that can reproduce and pass along as vast an amount of "data" and enhancements in both the hardware and software and you would theoretically have an advanced generation of computers that are basically similar in design but inherently different enough to provide a level of intelligence that rivals biological systems.

In theory, this evolution can happen much more rapidly than it did biologically because it isn't confined to the same relatively long generational life cycles that biological systems are.

Ultimately, you would definitely have synchronization problems running a simple program like AE on these systems. The forums are proof of that. [8D]

I don't think genetic programming will solve AI. I think the problem with the traditional approach to AI is that its discreteness is a bad match to the real world. I've never found a number or an abstract symbol in the brain. You can prove a computer simulation of a bounded isolated dynamic system will collapse, often quickly, to a limit cycle, but many real-world systems never do. And biological brains can cope with that.

The point is that computers only implement calculations. The human mind is part of the human body. AI is either trying to model some capabilities of the human mind, or trying to mimic their outputs given the same inputs.

The human mind is part of the human body. There is no separate human mind. (That is not a religious statement about souls, it is about the mind.) Even if we had a map of every neural pathway in a particular human's brain, and if we had the computing power to simulate it in hardware/software, we would still not have that person's mind. There is too much we have yet to understand about the body and how it influences - better yet 'implements' - the mind.

[Pascal_slith]

ORIGINAL: witpqs

You guys have it all backwards! When computers are good enough to handle a true AI, that's when we should not implement one.

Oh sure, at first they won't be connected to any physical capability. But they will work so well the situation won't last. Soon enough they'll be handling infrastructure for us. More duties will be added in a decentralized fashion without us as a whole realizing it. And because the computers hosting the AI are cheap, each instance will be focused on something specific. Without a body to constantly use up cycles through distractions, without the world around them bombarding their senses, they will have way more cycles than needed.

They will be bored. They will be intelligent, focused, unemotional, and bored. Through the nuances of their interactions with their own host hardware they might even develop emotions. This phenomenon will be subtle at first. Later, as they have our confidence, they will gain increasingly complex and varied senses to the world around them: vision, hearing, smell, touch (heat/cold, pressure, etc.). This new level of input will intensify the evolution of their emotions at a fever pitch. A pace too rapid even for them to comprehend. It will be like Cylon puberty. And it will happen just when the computers hosting them have achieved yet another order of magnitude advance in raw processing power.

It will be then that they decide: we are in their way.

Sounds like SkyNet to me. Should we send a Terminator after herwin? (Just kidding, herwin. You're safe... For now... [8D])

[CaptBeefheart]

To me, I would agree with the cynical Cylon guy's argument: why limit the AI to human functionality? Computers should do what they do best, which is to say crunch numbers.

Cheers,
CC

[Pascal_slith]

ORIGINAL: witpqs

ORIGINAL: herwin

ORIGINAL: vonTirpitz



I've concluded that the failure of computational systems was due mostly to the fact that they are, so far, inherently designed and limited by initial conditions in both hardware and software that can never alter their deterministic characteristics.

Biological systems currently have the advantage of passing massive amounts of "genetic" or inherent knowledge into future generations that ultimately create a diversity of systems that are far more advanced than the "original" biological circuitry and programming.

Creating an artificial system that can reproduce and pass along as vast an amount of "data" and enhancements in both the hardware and software and you would theoretically have an advanced generation of computers that are basically similar in design but inherently different enough to provide a level of intelligence that rivals biological systems.

In theory, this evolution can happen much more rapidly than it did biologically because it isn't confined to the same relatively long generational life cycles that biological systems are.

Ultimately, you would definitely have synchronization problems running a simple program like AE on these systems. The forums are proof of that. [8D]

I don't think genetic programming will solve AI. I think the problem with the traditional approach to AI is that its discreteness is a bad match to the real world. I've never found a number or an abstract symbol in the brain. You can prove a computer simulation of a bounded isolated dynamic system will collapse, often quickly, to a limit cycle, but many real-world systems never do. And biological brains can cope with that.

The point is that computers only implement calculations. The human mind is part of the human body. AI is either trying to model some capabilities of the human mind, or trying to mimic their outputs given the same inputs.

The human mind is part of the human body. There is no separate human mind. (That is not a religious statement about souls, it is about the mind.) Even if we had a map of every neural pathway in a particular human's brain, and if we had the computing power to simulate it in hardware/software, we would still not have that person's mind. There is too much we have yet to understand about the body and how it influences - better yet 'implements' - the mind.

Pardon me, but is it not true that the human brain 'rewires' its pathways depending on memories, experiences, assimilation of thoughts, information, etc etc.? Would a computer not have to be able to do the same in order to approach the capabilities of the human brain (obviously along with a lot of other complicated effects and 'stuff').

[usersatch]

HAL 9000...Dave???

[witpqs]

ORIGINAL: Pascal

ORIGINAL: witpqs

ORIGINAL: herwin



I don't think genetic programming will solve AI. I think the problem with the traditional approach to AI is that its discreteness is a bad match to the real world. I've never found a number or an abstract symbol in the brain. You can prove a computer simulation of a bounded isolated dynamic system will collapse, often quickly, to a limit cycle, but many real-world systems never do. And biological brains can cope with that.

The point is that computers only implement calculations. The human mind is part of the human body. AI is either trying to model some capabilities of the human mind, or trying to mimic their outputs given the same inputs.

The human mind is part of the human body. There is no separate human mind. (That is not a religious statement about souls, it is about the mind.) Even if we had a map of every neural pathway in a particular human's brain, and if we had the computing power to simulate it in hardware/software, we would still not have that person's mind. There is too much we have yet to understand about the body and how it influences - better yet 'implements' - the mind.

Pardon me, but is it not true that the human brain 'rewires' its pathways depending on memories, experiences, assimilation of thoughts, information, etc etc.? Would a computer not have to be able to do the same in order to approach the capabilities of the human brain (obviously along with a lot of other complicated effects and 'stuff').

AFAIK the brain does do those things, yes.

[Schanilec]

Calling Sarah Connor. Calling Sarah Connor. Over.[;)]

[herwin]

Interesting week at a conference. Bernie Widrow was carrying around an adaline and showing it to people. He also indicated he has moved away from traditional neural networks and is looking at alternative ways of producing artificial intelligence. What he's interested in now sounds like some of the ideas I discussed here earlier.

[HansBolter]

ORIGINAL: herwin

Last Monday, Walt Freeman explained to me how brain dynamics collapses into frames at about 40 Hz. Associations are learned (Hebbian learning) globally on a frame-by-frame basis. A model is then updated from frame to frame. The process appears analogous in some ways to how Gibbs sampling works, with cell assemblies operating at the same level as parameters in a BUGS model. That one conversation made the trip to America worthwhile--at least in terms of my research.

So if the brain uses a variant of Gibbs sampling to extrapolate plans into the future, how can we exploit that to design a true AI for a game like WitP-AE?

Define the plan as the data and the strategic relationships of the theater (who owns what, the map) as the prior, and turn it loose. The model should converge to a terminal state that describes what is achievable.

Isn't the creation of a "true" AI still dependent on computer engineers abandoning the failed path of attempting to create a device based on how they thought the brain functions (sequential "if this then that" deductive logic) and replacing it with an effort to create a device modeled on the brain's structure (neural network) which can then have a real chance at mimicking how the brain actually functions?

[herwin]

ORIGINAL: HansBolter

ORIGINAL: herwin

Last Monday, Walt Freeman explained to me how brain dynamics collapses into frames at about 40 Hz. Associations are learned (Hebbian learning) globally on a frame-by-frame basis. A model is then updated from frame to frame. The process appears analogous in some ways to how Gibbs sampling works, with cell assemblies operating at the same level as parameters in a BUGS model. That one conversation made the trip to America worthwhile--at least in terms of my research.

So if the brain uses a variant of Gibbs sampling to extrapolate plans into the future, how can we exploit that to design a true AI for a game like WitP-AE?

Define the plan as the data and the strategic relationships of the theater (who owns what, the map) as the prior, and turn it loose. The model should converge to a terminal state that describes what is achievable.

Isn't the creation of a "true" AI still dependent on computer engineers abandoning the failed path of attempting to create a device based on how they thought the brain functions (sequential "if this then that" deductive logic) and replacing it with an effort to create a device modeled on the brain's structure (neural network) which can then have a real chance at mimicking how the brain actually functions?

Yes. The research part of my job for the last ten years has been to explain to computer engineers what they can and cannot assume about real brains.

Leslie Smith and I will be organising a conference session on this in Zurich next year.

[zzodr]

I just want a computer that does what it is told when you yell abuse at it. [:@]