Slashdot Mirror

Uses a port of minix that runs in user mode on Solaris called Solaris Minix. Currently (last time I heard) it's used in the OS component of COSC301 and used to be used in COSC403.
It uses the solaris calls for context switching. The year I did the course we traced through a system call. For a trivial call (getpid() off the top of my head) we went through the minix user mode libraries that effectively trapped to the solaris kernel and out into the interrupt handler for minix (which was a wrapper for SIGUSR1) up into the kernel, through several context switches into the memory manager and back, etc.
It worked well as a teaching system (after all, that's what Minix was designed for) and has the advantage of running hosted in user mode and not taking out an entire computer to run it. Emulating things like a MMU involved some rather hacky solutions, but all the basic functions of an OS could be implemented in user mode. In fact, examining some of the emulation techniques used was quite educational.

Uses a port of minix that runs in user mode on Solaris called Solaris Minix. Currently (last time I heard) it's used in the OS component of COSC301 and used to be used in COSC403.
It uses the solaris calls for context switching. The year I did the course we traced through a system call. For a trivial call (getpid() off the top of my head) we went through the minix user mode libraries that effectively trapped to the solaris kernel and out into the interrupt handler for minix (which was a wrapper for SIGUSR1) up into the kernel, through several context switches into the memory manager and back, etc.
It worked well as a teaching system (after all, that's what Minix was designed for) and has the advantage of running hosted in user mode and not taking out an entire computer to run it. Emulating things like a MMU involved some rather hacky solutions, but all the basic functions of an OS could be implemented in user mode. In fact, examining some of the emulation techniques used was quite educational.

I also prefer the three-letter ones, just in consistency.

Erm http://news.bbc.co.uk/ , http://www.canterbury.ac.nz/ ...

"The Internet is American. English only, please." - saweeeet@aol.com in Cuban newsgroup

This implies that a non computational device could be built that is properly intelligent. Turing foresaw such devices, and refuted Penrose in the process, with his B-Machines (this is a poor reference but the only online mention I could find quickly, the full one is "Alan Turing's Forgotten Ideas in Computer Science" B. Jack Copeland and Diane Proudfoot, Scientific American April 1999.

Apparently Turing couldn't get this work published, while Penrose has sold tens of thousands of copies of his book; another piece of proof for my theories about peer review and publication (dark glower, bitter mumblings).

Roger Penrose is a great physicist, and an amazing mathematician; he is truly important to science in the way that Turing or Einstein are. It is a puzzle as to why he has got such a bee into his bonnet about AI, and rather flattering that he considers it so important as to divert his attention from his real (and probably more fundamental and important work).

2) Creationists argue that there are instances of "irreducible complexity" in biological organisms, and that this means that they could not have evolved, but were created instead. This is clearly bunk, however even if we accept it, and all its axioms, then it is actually an argument in support of the AI project. If we are created then there exists an existence proof for creation of intelligence.

``...two sets of functions are of special--although certainly not exclusive--interest. These are the functions that are computable by an idealized human being who is unaided by machinery, and the functions that are in principle computable in the real world, which is to say, are computable by machines, or organs, or in general entities, that physically could exist, given the resources on offer in the real world, even if they do not actually exist, nor ever do so. Turing argued, we think persuasively, that the first of these sets is coincident with the set of Turing-machine-computable functions. We believe that the extent of the second set is an open, empirical question. [...]

...it would--or should--be one the greatest astonishments of science if the activity of Mother Nature were never to stray beyond the bounds of Turing-machine-computability.''

Copeland believes, like Penrose, that the mind is NOT a Turing machine! On the other hand the best analysis of modern physics indicates that any finite physical process, understood by modern physics, can be simulated efficiently, i.e. with a number of gates polynomial in the size of the system, by a Quantum network. Copeland believes that Deustch is wrong not because he is too radical but because he is too conservative. (A Turing machine can perfectly simulate a Quantum computer although it is VERY UNLIKELY it can do it efficiently-- if it could we could crack RSA in real-time on your linux box.)

A. Wait.

Cats are more independent. You could start making software for a cat by making Oneko, the X cat, more responsive to its environment and give it more emotions than boredom. Not that its boredom can't be useful, as PURR-PUSS uses boredom as a trigger to try a more creative action, while learning [Andreae] by trial and error.

There is also a lot more stuff on adaptive behavior and machine learning out there.