In a previous article the way to make computers / software super-intelligent was considered insofar as we probably can't do it if we don't know what super-intelligent things will do, how they will think, etc. (Ithaka: They Don't Know What Higher Intelligence Means - Science)
In general, we regard computers as intelligent when they're fast, remember large volumes of data, and can run lots of different programs to do many different things. In that context, the CERN computing facility is smart insofar as it can find a Higg's Boson but it's not particularly intelligent since it can't be used for anything else. In other words, intelligence is more than being a killer at chess and we saw what happened with Bobby Fischer. He was one of the best players who ever lived and yet he went all-out cracker NAZI at the same time.
Ed: you're mixing up intelligence and abnormal psychology!
Perhaps not as the question is actually whether high intelligence makes us vulnerable to aberrant psychology. In computing, you will encounter many people with high intelligence and yet they're not capable of tying their own shoes. You would have to work in the field to really grok the truth of that and I worked in it for longer than was healthy.
From that time I offer one example of a maths wizard who only did laundry for his socks annually because he thought it was more effective for time management ... or who knows what. He was also in regular communication with a Russian academic friend who was one of the few in the world who understood Tschebyscheff polynomials. (Digital Commons: Tschebyscheff fitting with polynomials and nonlinear functions)
Note: among other things, he supported the mainframe security for the University of Cincinnati ... and, amazingly enough, it still worked.
The general premise is things can get rocky when humans push into high intelligence and the Rockhouse maintains that's in part because we don't know what high intelligence means, what it should do, how it should behave etc.
Ed: are you considering getting to the actual article?
This one needs an opening act, mate. Roll with it.
Also in general, we are teaching computers to get super intelligent but we do that by teaching them to emulate things our own minds do in terms of neural networking and there multiple approaches of this nature since the predominant philosophy appears to be super intelligent computers will come if we make computers think like us ... but better ... and faster.
Now we can get to the current science in which the researchers are trying to teach computers how to 'play a hunch' since human problem solvers with high skills can take short cuts in analysis because, among other things, such people can make 'educated guesses' with a high probability of accuracy. (Science Daily: Human intuition added to planning algorithms)
What that graphic has to do with it is your guess but it came with the article from MIT and here's the annotation.
Researchers from MIT's Computer Science and Artificial Intelligence Laboratory are trying to improve automated planners by giving them the benefit of human intuition. By encoding the strategies of high-performing human planners in a machine-readable form, they were able to improve the performance of competition-winning planning algorithms by between 10 and 15 percent on a challenging set of problems.
Credit: Jose-Luis Olivares/MIT
- SD
Here's further support for the idea the only way to make computers super intelligent is to do things we do but better.
"In the lab, in other investigations, we've seen that for things like planning and scheduling and optimization, there's usually a small set of people who are truly outstanding at it," says Julie Shah, an assistant professor of aeronautics and astronautics at MIT. "Can we take the insights and the high-level strategies from the few people who are truly excellent at it and allow a machine to make use of that to be better at problem-solving than the vast majority of the population?"
- SD
The researchers did just that but the result wasn't that spectacular.
The researchers discovered that the large majority of the students' strategies could be described using a formal language called linear temporal logic, which in turn could be used to add constraints to the problem specifications. Because different strategies could cancel each other out, the researchers tested each student's strategies separately, using the planning algorithms that had won their respective competitions. The results varied, but only slightly. On the numerical problems, the average improvement was 13 percent and 16 percent, respectively, on the flight-planning and satellite-positioning problems; and on the temporal problems, the improvement was 12 percent and 10 percent.
- SD
The MIT researchers offer a relatively dramatic improvement to computer problem solving but it really doesn't have such a high yield.
Ed: indicating use of a human model for radical increases in computer intelligence may not be such an effective path for the long term?
Quite so, Watson.
The Rockhouse wants a Lady Gaga show every time in science. We want to know what cool and bizarre creatures live on Enceladus; we want to know what they can do with a Higg's boson now that they found one; we want to know or at least read about what fantastically intelligent research can bring us. We want to know what a computer with an IQ of 500 or 1000 or more will do.
Ed: but the premise remains the same that we can't design fantastical intelligence unless we know what it means (i.e. not us).
That's a roger, Cap'n.
There may be some thinking in you that we will get computers to figure it out but they don't know what it means either; they're really not any more intelligent but they're much faster than we.
Ed: I can't tell if this is doom and gloom?
It's neither since this is when science crosses over into philosophy in asking what is intelligence anyway. It's not a trivial question since it's not likely we will meet that IQ 500 computer until we figure it out.
Since you may have your own hunch of something cheap coming because, well, this is the Rockhouse, the Clintons claim they have high intelligence so how do they solve this annual problem?
- Doug Sneyd, cartoonist for Playboy magazine in the Sixties and Seventies.
In general, we regard computers as intelligent when they're fast, remember large volumes of data, and can run lots of different programs to do many different things. In that context, the CERN computing facility is smart insofar as it can find a Higg's Boson but it's not particularly intelligent since it can't be used for anything else. In other words, intelligence is more than being a killer at chess and we saw what happened with Bobby Fischer. He was one of the best players who ever lived and yet he went all-out cracker NAZI at the same time.
Ed: you're mixing up intelligence and abnormal psychology!
Perhaps not as the question is actually whether high intelligence makes us vulnerable to aberrant psychology. In computing, you will encounter many people with high intelligence and yet they're not capable of tying their own shoes. You would have to work in the field to really grok the truth of that and I worked in it for longer than was healthy.
From that time I offer one example of a maths wizard who only did laundry for his socks annually because he thought it was more effective for time management ... or who knows what. He was also in regular communication with a Russian academic friend who was one of the few in the world who understood Tschebyscheff polynomials. (Digital Commons: Tschebyscheff fitting with polynomials and nonlinear functions)
Note: among other things, he supported the mainframe security for the University of Cincinnati ... and, amazingly enough, it still worked.
The general premise is things can get rocky when humans push into high intelligence and the Rockhouse maintains that's in part because we don't know what high intelligence means, what it should do, how it should behave etc.
Ed: are you considering getting to the actual article?
This one needs an opening act, mate. Roll with it.
Also in general, we are teaching computers to get super intelligent but we do that by teaching them to emulate things our own minds do in terms of neural networking and there multiple approaches of this nature since the predominant philosophy appears to be super intelligent computers will come if we make computers think like us ... but better ... and faster.
Now we can get to the current science in which the researchers are trying to teach computers how to 'play a hunch' since human problem solvers with high skills can take short cuts in analysis because, among other things, such people can make 'educated guesses' with a high probability of accuracy. (Science Daily: Human intuition added to planning algorithms)
What that graphic has to do with it is your guess but it came with the article from MIT and here's the annotation.
Researchers from MIT's Computer Science and Artificial Intelligence Laboratory are trying to improve automated planners by giving them the benefit of human intuition. By encoding the strategies of high-performing human planners in a machine-readable form, they were able to improve the performance of competition-winning planning algorithms by between 10 and 15 percent on a challenging set of problems.
Credit: Jose-Luis Olivares/MIT
- SD
Here's further support for the idea the only way to make computers super intelligent is to do things we do but better.
"In the lab, in other investigations, we've seen that for things like planning and scheduling and optimization, there's usually a small set of people who are truly outstanding at it," says Julie Shah, an assistant professor of aeronautics and astronautics at MIT. "Can we take the insights and the high-level strategies from the few people who are truly excellent at it and allow a machine to make use of that to be better at problem-solving than the vast majority of the population?"
- SD
The researchers did just that but the result wasn't that spectacular.
The researchers discovered that the large majority of the students' strategies could be described using a formal language called linear temporal logic, which in turn could be used to add constraints to the problem specifications. Because different strategies could cancel each other out, the researchers tested each student's strategies separately, using the planning algorithms that had won their respective competitions. The results varied, but only slightly. On the numerical problems, the average improvement was 13 percent and 16 percent, respectively, on the flight-planning and satellite-positioning problems; and on the temporal problems, the improvement was 12 percent and 10 percent.
- SD
The MIT researchers offer a relatively dramatic improvement to computer problem solving but it really doesn't have such a high yield.
Ed: indicating use of a human model for radical increases in computer intelligence may not be such an effective path for the long term?
Quite so, Watson.
The Rockhouse wants a Lady Gaga show every time in science. We want to know what cool and bizarre creatures live on Enceladus; we want to know what they can do with a Higg's boson now that they found one; we want to know or at least read about what fantastically intelligent research can bring us. We want to know what a computer with an IQ of 500 or 1000 or more will do.
Ed: but the premise remains the same that we can't design fantastical intelligence unless we know what it means (i.e. not us).
That's a roger, Cap'n.
There may be some thinking in you that we will get computers to figure it out but they don't know what it means either; they're really not any more intelligent but they're much faster than we.
Ed: I can't tell if this is doom and gloom?
It's neither since this is when science crosses over into philosophy in asking what is intelligence anyway. It's not a trivial question since it's not likely we will meet that IQ 500 computer until we figure it out.
Since you may have your own hunch of something cheap coming because, well, this is the Rockhouse, the Clintons claim they have high intelligence so how do they solve this annual problem?
- Doug Sneyd, cartoonist for Playboy magazine in the Sixties and Seventies.
No comments:
Post a Comment