[Scarabin]

mike777, on 2013-February-09, 02:09, said:

you seem to want to not discuss point one.....


agree there is a agreement for and to measure.....all else in nonsense


If we cannot agree on how to measure ...rest is silly.


You keep claiming no agreement =with nothing so silly to discuss.

Sorry, your posts are just too cryptic for me and I really don't know what you want to say. At a guess you may merely be saying that there is no point in prolonging the discussion if we cannot agree a definition of intelligence and a means of measuring it. If this is indeed what you are saying then I agree completely. So ave atque vale.

[Scarabin]

Been reading up on robot swarm technology but it does not seem to add to AI, rather a reversion to mindless robot behaviour using evasion technology.

[mike777]

http://games.yahoo.c...-232905827.html

Baile Zhang, an assistant professor of physics at Nanyang Technological University in Singapore, showed off his "invisibility cloak" on Monday at the TED2013 talks in Long Bea
--


btw new huge focus on brain research over next ten years. One step towards computers gaining intelligence.

any event think 2030 not 20,000 years from today when we can measure and compare intell. computer with a human.

[barmar]

If Moore's Law continues unabated, computer power in the 2030's will be 4,000 times current performance. With that kind of power, brute force algorithms should be able to reproduce many more qualities of human/animal intelligence. For instance, if a current facial recognition algorithm takes 5 minutes to recognize a face, the same algorithm run on a 2030 computer will take 75 ms (and improved algorithms plus specialized hardware will likely improve on this significantly). I suspect this is comparable to human face recognition, but haven't been able to find actual timing info.

[Scarabin]

My knowledge of AI is totally amateur, unschooled and pretty basic. I first discovered AI while trying to defeat it in the old Cinemaware classic "Lords of the Rising Sun" written for the Amiga.

Having got that disclaimer off my chest, may I make 2 points. I think we should distinguish between brute force and AI, and I think AI can have 3 elements: brute force, human intelligence replication, and illusion.

Brute force is machine intelligence: the things machines do better than humans, like math calculations, and which humans usually try to replicate with shortcuts. This may be very impressive but to my mind is not real intelligence. An example of this is a computer simulation of Scrabble, which any existing computer should be able to play to perfection.

Human intelligence replication is the process whereby a computer tries to reproduce human reasoning. To some extent, Monte Carlo simulates this based on using frequencies to supply a priori probabilities, but because it's an approximation, in practise its success is flawed by inexplicable artefacts and inadequate sampling. The only possible attempts to actually attack this are all programs which I have been unable to obtain: examples are Fred's Base lll, Ian Frank's Finesse, which both seem to be abandoned.

Illusion ranges from cheating,double dummy play or hidden advantages to your opponent, to rituals which give the illusion of intentional movement. Examples of the latter abound in "Age of Empires".

[mike777]

I only want to compare and measure intelligence of a computer in accepted ways.


Again does a submarine swim or is it only an illusion. Please let us only measure and compare it to a human in an accepted/standard way.

in 2030 lets compare and measure a computer's intellingence to a human in a standard way. My guess is it will measure to be equal.
If you want to claim the computer cheated or created an illusion/replication, ok.


For fun we can predict that a computer may obtain a measured level a billion times the sum of the entire human race in 2050. Again this may be cheating or creating an illusion/replication.

[dwar0123]

Scarabin, on 2013-February-28, 23:15, said:

My knowledge of AI is totally amateur, unschooled and pretty basic. I first discovered AI while trying to defeat it in the old Cinemaware classic "Lords of the Rising Sun" written for the Amiga.

Having got that disclaimer off my chest, may I make 2 points. I think we should distinguish between brute force and AI, and I think AI can have 3 elements: brute force, human intelligence replication, and illusion.

Brute force is machine intelligence: the things machines do better than humans, like math calculations, and which humans usually try to replicate with shortcuts. This may be very impressive but to my mind is not real intelligence. An example of this is a computer simulation of Scrabble, which any existing computer should be able to play to perfection.

Human intelligence replication is the process whereby a computer tries to reproduce human reasoning. To some extent, Monte Carlo simulates this based on using frequencies to supply a priori probabilities, but because it's an approximation, in practise its success is flawed by inexplicable artefacts and inadequate sampling. The only possible attempts to actually attack this are all programs which I have been unable to obtain: examples are Fred's Base lll, Ian Frank's Finesse, which both seem to be abandoned.

Illusion ranges from cheating,double dummy play or hidden advantages to your opponent, to rituals which give the illusion of intentional movement. Examples of the latter abound in "Age of Empires".

In practice, human intelligence is flawed by inexplicable artifacts and inadequate sampling. Really just a flaw in human intelligence, not so much a flaw in computers simulation of it.

What we really need is a Monte Carlo simulation of Monte Carlo simulations of human intelligence.

As for Age of Empires, I know a disgustingly large amount about that game. I played it an obscene amount and worked on the test team of one of its successors(age of kings).

There are really two aspects of AI in that game. Unit AI, which is mostly about pathing. Finding the best way to get from point A to point B. In Age of Empires it was done individually for each unit. In later installments they did it for groups of units, which made it a lot more efficient. Need to do a lot fewer pathing calculations when you are dealing with much fewer items to move and not only that but when you order 30 individuals to all move to the same location, they have a tendency to get in each others way, much better to order 30 individuals to move together to that spot.

Then there is Computer player AI, which never cheated by knowing anything a human player wouldn't know from their position, but had enormous advantages in resources. This was just a script to do certain things based on the situation and was readily editable. Zero town centers, build a town center. Have an army, send it to enemy location. Don't know where enemy is, scout. Low on food, build farms, etc.

[barmar]

Scarabin, on 2013-February-28, 23:15, said:

Having got that disclaimer off my chest, may I make 2 points. I think we should distinguish between brute force and AI, and I think AI can have 3 elements: brute force, human intelligence replication, and illusion.

This is a false distinction. It doesn't really matter how a device achieves a goal, so long as it achieves it. Birds fly by flapping their wings, airplanes fly by spinning a propeller or forcing exhaust through a pipe, their all ways of getting from here to there through the air.

The brain uses brute force -- thousands of synapses fire in coordinated ways to cause perceptions, memories, and thoughts. They're just connected in different ways than the circuits of typical computers (there are computers called "neural networks" that are designed more like brain networks, but they don't have as many cells and connections as real brains, so they can only perform simple tasks so far).

With enough computing power, you don't need to emulate the actual mechanisms in the brain. Chess playing programs don't think about the game at all like humans do, but they're now better than any human.

This is what I talked about earlier, where the bar keeps moving. A few decades ago, chess was a target of AI researchers. But eventually computers got fast enough that brute force was good enough to play well. Clever algorithms weren't necessary, and it became uninteresting to the AI people. Over time, this will keep happening.

Watson, the computer that played Jeopardy last year, is another case in point. It has a few interesting tricks, but for the most part it's just a fancy search engine.

[Scarabin]

dwar0123, on 2013-March-01, 13:55, said:

In practice, human intelligence is flawed by inexplicable artifacts and inadequate sampling. Really just a flaw in human intelligence, not so much a flaw in computers simulation of it.

What we really need is a Monte Carlo simulation of Monte Carlo simulations of human intelligence.

As for Age of Empires, I know a disgustingly large amount about that game. I played it an obscene amount and worked on the test team of one of its successors(age of kings).

There are really two aspects of AI in that game. Unit AI, which is mostly about pathing. Finding the best way to get from point A to point B. In Age of Empires it was done individually for each unit. In later installments they did it for groups of units, which made it a lot more efficient. Need to do a lot fewer pathing calculations when you are dealing with much fewer items to move and not only that but when you order 30 individuals to all move to the same location, they have a tendency to get in each others way, much better to order 30 individuals to move together to that spot.

Then there is Computer player AI, which never cheated by knowing anything a human player wouldn't know from their position, but had enormous advantages in resources. This was just a script to do certain things based on the situation and was readily editable. Zero town centers, build a town center. Have an army, send it to enemy location. Don't know where enemy is, scout. Low on food, build farms, etc.

I really enjoyed the various "Age of Empires" right up to No 4. The computer AI seemed to have several advantages: extra resources, better management of several tasks at once, and sometimes impossible goals for the hiuman player. I did realise it had a relatively simple standard strategy

I also enjoyed the Gettysburg,Waterloo - Austerlitz series, and even tried to mod these. Eventually I discovered I was more interested in AI than in graphics and decided to go back to my first-love, bridge simulation.

As regards your first point, I accept this but think human and machine artefacts are significantly different.

[Scarabin]

barmar, on 2013-March-01, 17:55, said:

This is a false distinction. It doesn't really matter how a device achieves a goal, so long as it achieves it. Birds fly by flapping their wings, airplanes fly by spinning a propeller or forcing exhaust through a pipe, their all ways of getting from here to there through the air.

The brain uses brute force -- thousands of synapses fire in coordinated ways to cause perceptions, memories, and thoughts. They're just connected in different ways than the circuits of typical computers (there are computers called "neural networks" that are designed more like brain networks, but they don't have as many cells and connections as real brains, so they can only perform simple tasks so far).

With enough computing power, you don't need to emulate the actual mechanisms in the brain. Chess playing programs don't think about the game at all like humans do, but they're now better than any human.

This is what I talked about earlier, where the bar keeps moving. A few decades ago, chess was a target of AI researchers. But eventually computers got fast enough that brute force was good enough to play well. Clever algorithms weren't necessary, and it became uninteresting to the AI people. Over time, this will keep happening.

Watson, the computer that played Jeopardy last year, is another case in point. It has a few interesting tricks, but for the most part it's just a fancy search engine.

Mine is a purely personal view and limited to game AI, but I honestly do not think the distinction between brute force and human intelligence is a false distinction. Let me try to make my point with two examples.

1) We could write a bridge simulation based on double dummy bidding and play, and overlay this with a pragmatic rule-based interface. Where a rule existed the simulation would follow the rule and where no rule existed it would fall back on double-dummy bidding and play. I venture to suggest this would give a very high level of simulation, virtually indistinguishable from expert play, and yet I do not think you would argue that it matched or exceeded human intelligence?

2) Similarly Scrabble playing simulations exist which are based on searching a built-in dictionary and performing anagrams. Apart from possible time constraints this would outplay any human competitor, and it is pure brute force or if you prefer machine intelligence so I guess you might say its AI matches or exceeds human intelligence?

I would not agree, because human intelligence goes far beyond this (a point which I think you made previously). The program might well explain the meaning of words better than I could but it would not be able to form those words into intelligible sentences.

[kenberg]

It seems to me that one should ask what computers can do and what they will be able to do, and at least partially put aside the somewhat semantic issue of what qualifies as intelligence.


Example: Drones were mentioned earlier. They can do a lot of things. Now let's ask some questions. In the tv series Homeland, a soldier held captive befriends the son of one of the captors. The son is killed in a drone strike, and when the captive soldier is released his subsequent behavior is much influenced by his experience of this death of a child. Can computers assess the likelihood of a soldier responding in the manner portrayed? Can computers assess the likelihood of citizens accepting the use of drones by our government in the control of our own citizens? Can computers decide whether it would or would not be a good idea? Can computers decide what the phrase "good idea" means?

It seems to me that in everyday life the mechanical part of planning is pretty simple. But what should we plan? What should we aspire to? These questions can be harder. And more interesting. And, I think, less amenable to robotic answers.


All that being said, I still think that my new Honda is talking to other cars and they all have a good time laughing about my inability to properly get my cell phone in sync with Blue Tooth.

[y66]

kenberg, on 2013-March-02, 08:45, said:

It seems to me that one should ask what computers can do and what they will be able to do, and at least partially put aside the somewhat semantic issue of what qualifies as intelligence.

Example: Drones were mentioned earlier. They can do a lot of things. Now let's ask some questions. In the tv series Homeland, a soldier held captive befriends the son of one of the captors. The son is killed in a drone strike, and when the captive soldier is released his subsequent behavior is much influenced by his experience of this death of a child. Can computers assess the likelihood of a soldier responding in the manner portrayed? Can computers assess the likelihood of citizens accepting the use of drones by our government in the control of our own citizens? Can computers decide whether it would or would not be a good idea? Can computers decide what the phrase "good idea" means?

It seems to me that in everyday life the mechanical part of planning is pretty simple. But what should we plan? What should we aspire to? These questions can be harder. And more interesting. And, I think, less amenable to robotic answers.

All that being said, I still think that my new Honda is talking to other cars and they all have a good time laughing about my inability to properly get my cell phone in sync with Blue Tooth.

You might enjoy this part of Yarden Katz's November 2012 Noam Chomsky interview titled Where Artificial Intelligence Went Wrong:

Quote

Chomsky: I think we learn a lot of things from the history of science that can be very valuable to the emerging sciences. Particularly when we realize that in say, the emerging cognitive sciences, we really are in a kind of pre-Galilean stage. We don't know what we're looking for anymore than Galileo did, and there's a lot to learn from that. So for example one striking fact about early science, not just Galileo, but the Galilean breakthrough, was the recognition that simple things are puzzling.

Take say, if I'm holding this here [cup of water], and say the water is boiling [putting hand over water], the steam will rise, but if I take my hand away the cup will fall. Well why does the cup fall and the steam rise? Well for millennia there was a satisfactory answer to that: they're seeking their natural place.

Like in Aristotelian physics?

Chomsky: That's the Aristotelian physics. The best and greatest scientists thought that was the answer. Galileo allowed himself to be puzzled by it. As soon as you allow yourself to be puzzled by it, you immediately find that all your intuitions are wrong. Like the fall of a big mass and a small mass, and so on. All your intuitions are wrong -- there are puzzles everywhere you look. That's something to learn from the history of science. Take the one example that I gave to you, "Instinctively eagles that fly swim." Nobody ever thought that was puzzling -- yeah, why not. But if you think about it, it's very puzzling, you're using a complex computation instead of a simple one. Well, if you allow yourself to be puzzled by that, like the fall of a cup, you ask "Why?" and then you're led down a path to some pretty interesting answers. Like maybe linear order just isn't part of the computational system, which is a strong claim about the architecture of the mind -- it says it's just part of the externalization system, secondary, you know. And that opens up all sorts of other paths, same with everything else.

[kenberg]

It's always good to hear that someone I knew of in my youth is still alive and active! Go Noah!
I'll read the whole article. After lunch. And a nap.

[Scarabin]

I think Mike777 makes a valid point when he suggests defining what "computers matching human intelligence" means before estimating when it may be achieved, and that's what I am trying to do. After all, if we lower the bar too much then it has already been accomplished, if we raise it too high then it never will.

That said I am not sure there is much point in predicting computer progress in a vacuum. There must even be a significant chance PC development will stagnate while attention is diverted to mobile phones,etc.

Like the reference to the "We think we know but do we" concept, but puzzled by Ken's "somewhat semantic", a bit like saying we communicate in prose?

Returning to Barmar's "It doesn't matter how a device achieves a goal so long as it achieves it", I think the "how" makes a huge difference. Many years ago the Rand Corporation were hired to determine the best way to deliver an atomic device to its target. They considered many factors: cost, speed, reliability being the most obvious. Enough said?

[mike777]

Scarabin, on 2013-March-02, 21:22, said:

I think Mike777 makes a valid point when he suggests defining what "computers matching human intelligence" means before estimating when it may be achieved, and that's what I am trying to do. After all, if we lower the bar too much then it has already been accomplished, if we raise it too high then it never will.

That said I am not sure there is much point in predicting computer progress in a vacuum. There must even be a significant chance PC development will stagnate while attention is diverted to mobile phones,etc.

Like the reference to the "We think we know but do we" concept, but puzzled by Ken's "somewhat semantic", a bit like saying we communicate in prose?

Returning to Barmar's "It doesn't matter how a device achieves a goal so long as it achieves it", I think the "how" makes a huge difference. Many years ago the Rand Corporation were hired to determine the best way to deliver an atomic device to its target. They considered many factors: cost, speed, reliability being the most obvious. Enough said?

You raise excellent points.

I am assuming the learning curve of understanding the brain is convex rather than linear(can a learning curve be linear? ) or concave. Many disagree on this point and I frankly think this is the most interesting debate, the shape of the learning/knowledge curve for this topic. I am also assuming that the growth of computer power/memory/speed/etc is convex.

Perhaps I have not been clear.


1) Use standard accepted methods to measure and compare. I dont suggest a new method.
2) More research into the software and hardware of the brain is needed.
3) one goal post is understanding the hardware of the brain by 2020
4) understanding the software of the brain in 2029
5) computer intelligence equals that of an average human in 2030
6) by 2050 we will not be able to predict/understand what computers can do or achieve.

Of course all of this assumes it is not against the laws of science that a computer can have a standard measured level of intelligence that can be compared to a human.

[kenberg]

Scarabin, on 2013-March-02, 21:22, said:

puzzled by Ken's "somewhat semantic", a bit like saying we communicate in prose?

I'll explain, as I often take this view on issues.

Can computers play a good game of chess? No doubt.
A good game of bridge? Not as good, for the moment, as chess but yes, quite good.
Can computers do a good job of assessing the value of a stock? Probably, but there are issues.
Should President Obama nominate a computer as ambassador to France? No, not a good idea I think.

We need all of the intelligence we can get in international relations. But I doubt we should turn to computers except as very useful adjuncts. The type of judgment required for a good ambassador certainly involves intelligence but I think it is the kind of intelligence that computers are not so good at. Of course you might say that the human track record here is not so great either, true enough. But with sense and experience, a person comes to feel "This person I can trust, that person I cannot". Or "This would be a good idea, but the time is not ripe for presenting it". Or etc.

So what I meant is that we could discuss, and maybe agree, maybe not, on what sort of things computers would be good at doing and what they would not be good at doing without ever using the word "intelligent". If, later, we could also agree on which tasks represent intelligence and which do not, that could be nice, but it is comparatively unimportant. Any disagreement would, perhaps, center on what is the correct usage of the word "intelligent". If we agree on what tasks computers can be trusted with and what tasks should be left to humans, then I call questions about the applicability of the word 'intelligent" a semantic issue.

[barmar]

The basic problem is that "intelligence" is a very vague term.

We shouldn't say that whether a computer can play a good game of bridge, chess, or scrabble is a measure of its intelligence. One of the best ways we've come up with or gauging this is the Turing Test: if the computer can fool humans into thinking that it's a human, then we should declare it to be intelligent. If it does it entirely through brute force, so be it.

[mike777]

There is also the issue of what does it mean to be human.

Clearly there will be enhancements to humans in the near future.


It is the “first authoritative and comprehensive survey of the origins and current state of transhumanist thinking, according to the editors, and the anthology includes a roster of leaders in transhumanist thought.

“The rapid pace of emerging technologies is playing an increasingly important role in overcoming fundamental human limitations,” say the editors.

Featuring core writings by seminal thinkers in the speculative possibilities of the posthuman condition, essays address key philosophical arguments for and against human enhancement, explore the inevitability of life extension, and consider possible solutions to the growing issues of social and ethical implications and concerns

http://www.kurzweila...tm_medium=email


The book is edited by the internationally acclaimed founders of the philosophy and social movement of transhumanism, The Transhumanist Reader is “an indispensable guide to our current state of knowledge of the quest to expand the frontiers of human nature

[Scarabin]

barmar, on 2013-March-04, 09:29, said:

The basic problem is that "intelligence" is a very vague term.

We shouldn't say that whether a computer can play a good game of bridge, chess, or scrabble is a measure of its intelligence. One of the best ways we've come up with or gauging this is the Turing Test: if the computer can fool humans into thinking that it's a human, then we should declare it to be intelligent. If it does it entirely through brute force, so be it.

My understanding is that the Turing Test got whittled down to the point where it could be met by "Liza" clones. If we return to a more general test where the computer demonstrates human performance on several fronts then I think it mjght be more convincing.

[Scarabin]

mike777, on 2013-March-04, 10:05, said:

There is also the issue of what does it mean to be human.

Clearly there will be enhancements to humans in the near future.


It is the “first authoritative and comprehensive survey of the origins and current state of transhumanist thinking, according to the editors, and the anthology includes a roster of leaders in transhumanist thought.

“The rapid pace of emerging technologies is playing an increasingly important role in overcoming fundamental human limitations,” say the editors.

Featuring core writings by seminal thinkers in the speculative possibilities of the posthuman condition, essays address key philosophical arguments for and against human enhancement, explore the inevitability of life extension, and consider possible solutions to the growing issues of social and ethical implications and concerns

http://www.kurzweila...tm_medium=email


The book is edited by the internationally acclaimed founders of the philosophy and social movement of transhumanism, The Transhumanist Reader is “an indispensable guide to our current state of knowledge of the quest to expand the frontiers of human nature

Thanks for the link which is interesting and clarifies your previous points. I fear however, Mike, that I am too old and too sceptical to take Transhumanism seriously. Indeed I have trouble freeing myself from the feeling that it's an elaborate hoax.

I fear that Ken's suggestion of evaluating things computers do well or badly is more my speed. I am sorry about this because I really enjoyed our interchange of ideas.