BEAM Newsgroup? [Tilden; bottom-up vs. top-down]
Arthur T. Murray
uj797 at victoria.tc.ca
Sun Mar 18 15:34:03 EST 2001
eskwired at SPAMBLOCK.shore.net writes on Sun, 18 Mar 2001:
[ big juicy controversy cut to the chase: ]
>|> | There is no such thing, except among deluded BEAM-ites. Even Tilden
>|> | wouldn't say such a stupid thing, he would look stupid to too many
>|> | people. Being his little believer minions you, however, are not so
>|> | limited by your prestige and can afford such a cheap belief!
>|> Rodney Brooks says something quite similar, Steve, and I'd tend to trust
>|> his conclusions more than I trust yours.
>| Quote him. Betcha can't!
>"I developed this behavior-based subsumption approach to robots ... I
>really think that this is all there is in animals - and people."
>Q: This lower-level processing?"
>"What's happened is, AI has sort of accepted this:
>'Oh, OK, that's how the low-level parts work, but then
> there must be this other high-level stuff.'
> But I don't think there is that high-level stuff.
> I think that's pretty much all that there is..."
Tiresome, meddling Mentifex Arthur:
Having written http://www.scn.org/~mentifex/jsaimind.html --
with "eskwired" and with Rodney Brooks, because I was
amazed at how simple the building blocks of mind were,
in the form of neronal aggregates joined by associative tags,
when I deciphered a way of embedding Chomskyan superstructures
on top of a simplistic neuronal mindgrid as decribed at
http://www.geocities.com/mentifex/theory5.html "Know Thyself!"
> So Brooks said that he rejects the notion of "higher-level"
> stuff, and posits tht complex behavior, and intelligence,
> is made from levels of simple behavior-based inputs and reactions.
> Bart said: "it is feasible to use this type of logic
> for a more powerful robot as the research continues.
> It is extremely simple stuff, but I do believe it will evolve."
Who's got time to wait around for blind bottom-up evolution?
> So Bart too posits that simple technology can be the basis
> of more complex behaviors.
> I find it interesting that you reject Bart's POV
> on a wholesale basis, while indicating below that
> your attitude towards Brooks POV is not rejected.
> And yet, Brooks goes well beyond Bart, and posits
> not only that simple reactive circuits can be combined
> to do usefull or interesting stuff, but also that
> the thought processes of higher animals are composed
> of nothing more.
> Brooks is saying that abstract thought, and that emotion,
> and all cognative abilities are an illusion,
> if one thinks that they are different in type
> from yanking your hand off a hot surface.
> All that Bart is saying is that one method of
> controlling low-level processes may yield higher-level
> processes when combined. This combination is currently
> effected using layers, just like Brook's subsumption architecture.
> Both fields use non-Von Neumann computational models.
> Both seek to build complex behavior with little or
> nothing more than simple building blocks.
> Both have little in the way of central awareness.
> Both use the world as a model. Neither use "representation"
> of the external world as a basis to perform tasks.
> Neither use "reason" as a necessary component for intelligence.
The Encyclopedia Cybernetica -- in machine-grokkable form
More information about the Neur-sci