Analog computers display characteristic problems as well. I don't know for sure but the problem is that in complex decision making many of the approximations to a good result will lead to a weaker decision making and this kind of error can accumulate (rapidly) in analog systems as well as digital systems. I believe that digital systems can be better at many detailed processes which base branch decisions on run time evaluations because analog systems are able to maximize their capability only by using powerful approximations. So analog computers are good at making simple estimates where the error can be kept within bounds but they are terrible at more complex decision making processes.
Using multiple independent systems can be more efficient but I do not think that there is an exponential or log time efficiency that holds for parallel systems. There are exceptions but the point is that once people became aware of the exception they can rewrite their serial programs to render the advantage of using a parallel system to polynomial time. (Typically it is just a linear time advantage because highly effective interactive massively parallel systems need so much management.) The most efficient parallel programming is to separate simple problems which can be done almost independently to be run on separate threads. More elaborate systems which, for example, might use multiple copies of memory might be more effective, but it would obviously not be memory efficient. Other kinds of mechanical devices can be used to run more efficient simulations but because mechanical devices cannot rapidly assume different symbolic values (different 'conceptual' symbol references) that means that they are not as adept to working on a variety of different kinds of problems, and that is one of the assumptions of AGI. Jim Bromer On Thu, Jun 28, 2012 at 10:05 AM, Peter Voss <[email protected]> wrote: > > This issues has bothered me for a long time, and I’d like to explore it a > bit: > > While digital computers obviously can be set up to solve equations, there > still seems to be a significant difference in efficiency of simulating/ > calculating versus physical analog ‘doing’/ execution – like for example in > solving an n-body problem. Real systems system just produce the result by > interaction of all the forces (electro/ mechanical), while computers have to > approximate/ iterate. > > Key question: Are there AGI common problems where digital/simulated > approaches need hyper-exponential amounts of computing power compared to > physical systems? Is this kind of equation-solving core to AGI? I don’t > think so, but… > > Other may be able to formulate this better. > > What has bothered me is the glib assertion that a digital computer an > calculate to any arbitrary level of precision (true)… but does the cost > become unworkable in practice, even with Moore’s law. > > Peter While the human mind does not solve the n-body problem for complex systems, it probably does something similarly complex. But that does not mean that the solutions that it might come up with would necessarily be perfect. We might think of it in the terms of a rough solution method that somehow resolves most of the most cruder errors. That is where contemporary AGI seems to fail. We have a lot of techniques which can make good local estimates but which seem to have almost no capacity for error correction. ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
