vu0208@bingvaxu.cc.binghamton.edu () (06/17/91)
With so many much advance in computer architectures and AI, are we heading to any progress towards the development of computer machines, that in turn ON-THEIR-OWN design new computer architectures. This could be a PIPE-DREAM but if can achieve that in distant future -then I guess we have created a new intelligent life. :}
gdtltr@brahms.udel.edu (gdtltr@limbo.org (The Befuddled One)) (06/17/91)
In article <1991Jun16.214929.26509@newserve.cc.binghamton.edu> vu0208@bingvaxu.cc.binghamton.edu writes: =>With so many much advance in computer architectures and AI, are we =>heading to any progress towards the development of computer machines, =>that in turn ON-THEIR-OWN design new computer architectures. => =>This could be a PIPE-DREAM but if can achieve that in distant future =>-then I guess we have created a new intelligent life. :} It seems to me that the main problem along these lines is describing the problem and the desired goal. Ok, so we give a computer a program to fit so many transistors into a certain space and let it design away. How do you evaluate the performance of the newly designed machine, and for what? It could certainly attempt to optimize for a benchmark, but it is my understanding that benchmarks are more marketing tools than anything else. However, even if we can come up with a "state-evaluation function" for a processor, even for a restricted case, certainly it would take enormous amounts of time to run the thing. It seems that human designers rely heavily on philosophical grounds in deciding on which paths to take. (obvious e.g. RISC vs. CISC) It might be possible to add "assumptions" (where here I actually mean "biases") to the program to simplify things, but I don't think that would produce much more than a very advanced CAD system. I'm no AI expert (or necessarily an expert in anything,) but last I checked, we didn't have creative AI systems yet. On the other hand, strictly in pipe-dream terms (:-), I don't see why such a thing could not eventually be possible. Before we can expect to have a fair, intelligent computer design computer, we will need A) more advanced human-designed computers to run it on, and more importantly, B) a better understanding of what exactly we are designing and what we are designing it for. Gary Duzan Time Lord Third Regeneration -- gdtltr@brahms.udel.edu _o_ ---------------------- _o_ [|o o|] To be is to be networked. [|o o|] |_o_| Disclaimer: I have no idea what I am talking about. |_o_|
mshute@cs.man.ac.uk (Malcolm Shute) (06/17/91)
In article <22157@brahms.udel.edu> gdtltr@brahms.udel.edu (gdtltr@limbo.org (The Befuddled One)) writes: > It seems to me that the main problem along these lines is describing >the problem and the desired goal. This is the sort of problem which G.A. (genetic algorithms) are supposed to tackle. So, like you, I'd agree that computers do more and more to design computers. As you also hint, they are already doing it... today's sophisticated CAD does more to design computers than yesterday's did. In the same way, today's compilers do more to write their own assembler programs (from the abstract ideas which the human sketches out to them) than yesterday's did. Are these really 'designing' off their own bat, though? Of course not! What good would a machine be to mankind if it refused to do what the human operators commanded it to do... a machine, afterall, is built by humans to do the tiresome work *for* the humans. We will always want the machine (the computer which designs computers in this case) to act under the direction of human guidance. Returning to the thread at the top of this posting, though: If GA is to be used, the potential, new computer designs need to evolve in some sort of environment in which natural selection is allowed to work. This is already provided, of course... we all know that 'the market place' represents the ideal jungle in which to test out the survival of the fittest. The GA program therefore needs to be connected to a good behavioural simulator, and a good economics simulator, and left to run. (The problems with this, are how to connect the behavioural simulator to the economics one (what makes a good, marketable processor?); and the slow operation of GA). If you subscribe to the Richard Dawkins view of the universe, then this process is already well underway. The cycle admittedly presently contains some pesky ant-like creatures (called humans), organised into structures which they call Intel, MIPS, Motorola, Inmos, etc... But gradually newer and fitter processors *are* evolving. (See a thread which is currently underway in sci.nanotech on the subject of 'meme's). As I have said above, I don't believe that the human component will ever be removed from this system. By giving the tiresome work to the machine, the human designer is freed to soar higher into the design space, devoting his/her braincells to higher levels of the design (at each new generation of CAD, the fontier rises one stage further, the machine takes on a higher level of the design, and the human is freed to rise one stage higher too). No commercial organisation could afford to exclude the human member from the human-machine partnership, not least because the former must always be the team leader! I'm sorry for the long and rambling nature of this posting. I'm in the middle of marking a pile of examination scripts, and needed to hit back at the rest of the world in general! -- Malcolm SHUTE. (The AM Mollusc: v_@_ ) Disclaimer: all
thomae@eos.ncsu.edu (DOUGLAS ALAN THOMAE) (06/18/91)
In article <2705@m1.cs.man.ac.uk>, mshute@cs.man.ac.uk (Malcolm Shute) writes: |>In article <22157@brahms.udel.edu> gdtltr@brahms.udel.edu (gdtltr@limbo.org (The Befuddled One)) writes: |>> It seems to me that the main problem along these lines is describing |>>the problem and the desired goal. |>This is the sort of problem which G.A. (genetic algorithms) are supposed |>to tackle. So, like you, I'd agree that computers do more and more to Not just genetic algorithms. GAs are an optimization technique that could be used for CAD tasks, including placement, partitioning, and logic synthesis. Other approaches can be used as well, both iterative improvement methods (like GAs) and constructive methods. Iterative improvement methods include simulated annealing, variations on the Kernighan and Lin algorithm and force directed methods. Examples of constructive methods include liner and nonlinear programming approachs, seeding and clustering. Constructive methods may be used by themselves or to generate initial solutions for iterative improvement methods, for example generating an initial population for GAs. Using constructive algorithms in this way can improve performance over randomly generated starting solutions in iterative improvement methods. |> |>Are these really 'designing' off their own bat, though? |>Of course not! What good would a machine be to mankind if it refused to |>do what the human operators commanded it to do... a machine, afterall, |>is built by humans to do the tiresome work *for* the humans. |>We will always want the machine (the computer which designs computers in |>this case) to act under the direction of human guidance. |> This is a pretty slippery point. I would agree that anything that CAD systems do today or are likely to do in the near future can't really be considered design in the same sense that a human does design, but I'm not sure that basing the decision on whether or not the computer does the design on its own initiative or not is a good criterion. After all, when the typical human designer works is the primary reason because he wants the product of his efforts for himself or is it because his boss asks him to do it for some reason? I would say that the difference between what CAD systems do today and what human designers do lies in the level of the specification. Humans can deal with fairly loose specifications, while even 'high level' synthesis systems have to be told about the design in relatively low level nuts and bolts terms. |>Returning to the thread at the top of this posting, though: |>If GA is to be used, the potential, new computer designs need to evolve |>in some sort of environment in which natural selection is allowed to work. |>This is already provided, of course... we all know that 'the market place' |>represents the ideal jungle in which to test out the survival of the fittest. |>The GA program therefore needs to be connected to a good behavioural simulator, |>and a good economics simulator, and left to run. (The problems with this, |>are how to connect the behavioural simulator to the economics one (what makes |>a good, marketable processor?); and the slow operation of GA). |> Building good models of the thing that you want to optimize, and coming up with appropriate cost functions is one of the problems that stands between the current state of the art and being able to build machines that can do design in the same sense that humans can (assuming that this is possible). Problems with the large number of CPU cycles that would be required might be attacked by trying to find simpler models that still contain the essential elements of the system you're trying to model. For example, it should be possible to model the economic factors that influence the success or failure of an architecture without modeling the whole economy (taking the most extreme interpretation of the statement above).
vu0208@bingvaxu.cc.binghamton.edu (06/19/91)
In article <2705@m1.cs.man.ac.uk> mshute@cs.man.ac.uk (Malcolm Shute) writes: >In article <22157@brahms.udel.edu> gdtltr@brahms.udel.edu (gdtltr@limbo.org (The Befuddled One)) writes: >> It seems to me that the main problem along these lines is describing >>the problem and the desired goal. >This is the sort of problem which G.A. (genetic algorithms) are supposed >to tackle. So, like you, I'd agree that computers do more and more to >design computers. As you also hint, they are already doing it... >today's sophisticated CAD does more to design computers than yesterday's did. >In the same way, today's compilers do more to write their own assembler programs >(from the abstract ideas which the human sketches out to them) than yesterday's did. > Doing smart CAD is not enough! In my original posting I meant "Can we (humans) build Computers which in turn build new computer architectures without any human intervention at any stage.(period) All these CADs/GAs/AIs etc are ofcourse the building blocks and tools that will be available to human-designed-computers. >Are these really 'designing' off their own bat, though? >Of course not! What good would a machine be to mankind if it refused to >do what the human operators commanded it to do... a machine, afterall, >is built by humans to do the tiresome work *for* the humans. >We will always want the machine (the computer which designs computers in >this case) to act under the direction of human guidance. that's exactly my point. >As I have said above, I don't believe that the human component will ever be >removed from this system. Here I will disagree, Assume (atleast in a fantasy) we have built such a smart-self-reliant-computer (or h-m-c human-made-computer) that will only take the specification of functions/tasks and the evaluation standards from it's creator (the human) for CREATING a new computer architecture (I call it c-m-c ie. computer-made computer)! You can think of all the fantastic tools/libraries/databases/algorithms/self evaluation schemes etc.. that are availabe to this h-m-c. Then I don't see any reason why it cannot build a machine (c-m-c) that equals or exceeds the specs. we provided to it without any human intervention at all!! Now you may argue that what if the specs. we provide are so tight & tough that it may not even come close to it. That's where we have to ponder and do research, and I am sure such an h-m-c will help us to solve these problems and lead us to the point where a h-m-c will be completely autonomous and would only require specs/problems from a human not their solutions.!!! -- a Dreamer.
daq@hpfcso.FC.HP.COM (Doug Quarnstrom) (06/20/91)
>Doing smart CAD is not enough! >In my original posting I meant "Can we (humans) build Computers which >in turn build new computer architectures without any human >intervention at any stage.(period) Then the answer to your question is NO. We are nowhere near that. It would really suprise me a great deal to see anything approaching true machine intelligence in my lifetime, and I am only 31. >>As I have said above, I don't believe that the human component will ever be >>removed from this system. >Here I will disagree, Assume (atleast in a fantasy) we have built such >a smart-self-reliant-computer (or h-m-c human-made-computer) that will >only take the specification of functions/tasks and the evaluation >standards from it's creator (the human) for CREATING a new computer >architecture (I call it c-m-c ie. computer-made computer)! Ok, well this is simpler, and you MIGHT be able to do this without real intelligence. Certainly without (crige) sentience. It is dangerous to say this kind of thing will never happen. It may, but it is probably not too close. >You can think of all the fantastic >tools/libraries/databases/algorithms/self evaluation schemes etc.. >that are availabe to this h-m-c. Then I don't see any reason why it >cannot build a machine (c-m-c) that equals or exceeds the specs. we >provided to it without any human intervention at all!! What is your experience in computer design? I ask because you might be suprise at how little really gets done by the computer as far as the creative process of design goes. Right now I wouls say that it is little or nothing. >Now you may argue that what if the specs. we provide are so tight & >tough that it may not even come close to it. That's where we have to >ponder and do research, and I am sure such an h-m-c will help us to >solve these problems and lead us to the point where a h-m-c will be >completely autonomous and would only require specs/problems from a >human not their solutions.!!! Yeah, but if the specs are so hot that a computer can design based on them, it is little more that a really good CAD system and the 'magic' or creativity is in the spec. The machine you describe probably would not be the new 'intelligent life form' you describe. >-- a Dreamer. At the very least. In my opinion, we are likely to see effective computer interfaces to the human nervous system long before we create a new intelligence. doug
mshute@cs.man.ac.uk (Malcolm Shute) (06/20/91)
In article <1991Jun18.202403.9592@newserve.cc.binghamton.edu> vu0208@bingvaxu.cc.binghamton.edu writes: >Doing smart CAD is not enough! But at what point does CAD cease to be "Computer Aiding the Human to do the Design" (CAHD), and turn into "Computer doing the Design at the Human's Request" (CDHR)? I agree entirely with thomae@eos.ncsu.edu (DOUGLAS ALAN THOMAE) when he replied in message <1991Jun18.144225.21906@ncsu.edu>: %I would agree that anything that CAD %systems do today or are likely to do in the near future can't really be %considered design in the same sense that a human does design, but I'm not sure %that basing the decision on whether or not the computer does the design on %its own initiative or not is a good criterion. After all, when the typical %human designer works is the primary reason because he wants the product of his efforts %for himself or is it because his boss asks him to do it for some reason? He goes on: %I would say that the difference between %what CAD systems do today and what human designers do lies in the level %of the specification. Humans can deal with fairly loose specifications, %while even 'high level' synthesis systems have to be told about the design %in relatively low level nuts and bolts terms. Surely we'd all agree that there is a whole spectrum between CAHD and CDHR, above? That at no point in that spectrum would you be able to say that the machine could take the credit for the design. Even in Douglas Alan Thomae's simile, the boss normally takes the credit for the design of systems which were overseen by him... and with good reason. He might not have specified the nuts and bolts, but he had the highlevel insight into how to structure the design, and/or the design effort. With a boss, though, the human underlings can fight back if they feel that too much of the credit is being stolen from them. When a computer is the underling... I can't imagine it fighting for its rights for some centuries yet! >In my original posting I meant "Can we (humans) build Computers which >in turn build new computer architectures without any human >intervention at any stage.(period) In which case my answer is no. What use is a machine which does not act only to obey the commands of its human owner? However, if what you mean is "Can we build Computers which can embark on CDHR?" the answer must surely be yes, since it is a sliding scale, and we are already slid quite a way already in that direction. Each year will see us slide a bit further. >>We will always want the machine (the computer which designs computers in >>this case) to act under the direction of human guidance. >that's exactly my point. Yes... I don't think that we are disagreeing with each other... just shifting the wording around. >Assume (atleast in a fantasy) we have built such >a smart-self-reliant-computer (or h-m-c human-made-computer) that will >only take the specification of functions/tasks and the evaluation >standards from it's creator (the human) for CREATING a new computer >architecture (I call it c-m-c ie. computer-made computer)! >You can think of all the fantastic >tools/libraries/databases/algorithms/self evaluation schemes etc.. >that are availabe to this h-m-c. Then I don't see any reason why it >cannot build a machine (c-m-c) that equals or exceeds the specs. we >provided to it without any human intervention at all!! Except at the beginning and the end of the design. I agree with you, this fantasy would seem to be attainable one day. I would use my CDHR label on it (pinching your idea to give it a c-m-c label!). -- Malcolm SHUTE. (The AM Mollusc: v_@_ ) Disclaimer: all
w1cowan@watson.ibm.com (Crispin Cowan) (06/21/91)
In article <1991Jun16.214929.26509@newserve.cc.binghamton.edu> vu0208@bingvaxu.cc.binghamton.edu writes: >With so many much advance in computer architectures and AI, are we >heading to any progress towards the development of computer machines, >that in turn ON-THEIR-OWN design new computer architectures. At this point, we don't even have a computer system that can UNPACK a computer without assistance. Crispin ----- Crispin Cowan, CS grad student, University of Western Ontario Phyz-mail: Middlesex College, MC28-C, N6A 5B7 E-mail: crispin@csd.uwo.ca Voice: 519-661-3342 Temporarily at: w1cowan@watson.ibm.com "If you want an operating system that is full of vitality and has a great future, use OS/2." --Andy Tanenbaum