Forums before death by AOL, social media and spammers... "We can't have nice things"
|    comp.ai    |    Awaiting the gospel from Sarah Connor    |    1,954 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 1,226 of 1,954    |
|    jonesrob@emporia.edu to BryanJacobson    |
|    Re: Managing computational complexity fo    |
|    03 Nov 06 22:48:54    |
      BryanJacobson wrote:       > jonesrob@emporia.edu wrote:       > > . . .       > > Computational complexity appears to be the biggest       > > stumbling block standing in the way of efforts to build       > > more capable AIs. In my "Asa H" project I seek to control       > > complexity using some of the same methods that humans       > > use . . .       > > 1. category/concept formation/maintenance by clustering/       > > vector quantization       > > . . .       > > 4. avoiding search when the previous match remains       > > adequate       >       > Rob,       >       > Interesting note.       >       > Is computational complexity the biggest stumbling block? Perhaps a       > bigger stumbling block is inadequate knowledge representation which       > then results in inadequate semantic algorithms (such as computationally       > intensive search).       >       > For example, I'm hungry so I decide to warm up some leftover spagheti.       > I don't think I searched through all the things I can do (read, write,       > watch TV, call a friend, etc.) to find something that "solved" hunger,       > and then searched through all the things I might eat (steak, Big Mac,       > bugs, applesauce) to finally settle on spagheti.       >       > Unfortunately I don't know exactly what I do. But I don't think it       > involved searching through billions of alternatives.       >       > Of course, until we have better knowledge representation, we may be       > able to accomplish a lot with what we have, and good luck on your       > system.       >       > Cheers,       >       > -- Bryan              Asa H. assembles a kind of semantic network as it learns. Spreading       activation mechanisms in this network accomplish something like       you have in mind.              There is, however, the question of simple universal elements vs       rich and special purpose mechanisms. I think this tradeoff remains       to be fully explored.              R. Jones              [ comp.ai is moderated ... your article may take a while to appear. ]              --- SoupGate-Win32 v1.05        * Origin: you cannot sedate... all the things you hate (1:229/2)    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca