J. R. Molloy wrote:
> From: "Robert J. Bradbury" <bradbury@aeiveos.com>
>
>>Now, in the future, as scientific literature hopefully becomes freer
>>*and* as it becomes directly accessible by our brains, then we will
>>be able to think things out carefully backreferencing to the scientific
>>literature as nessary. And that will be a marvelous day indeed.
>>
> Autonomous thinking machines of the kind that can figure things out for
> themselves, and that can do it in nanoseconds, increasingly outdistance
> humans (who remain mired in marvel). Machines have already demonstrated the
> ability to solve problems that have stymied humans. To think for yourself
> (autonomous thinking) means to solve problems on your own, without resorting
> to the literature, without looking up answers in reference books, without
> using cheat sheets. To reach correct conclusions by responsive reason, rather
> than by reactive recitation, is the mark of intelligence, rather than the mark
> of regurgitation (as in parrots and script kiddies).
This reflects my ideas about "intelligence" as of about 1995. I was
wrong. If "intelligence" is pure logic, then a perfect intelligence
would always operate from first principles and observed fact, and never
from pre-computed stored knowledge, which you think of as "cheating." In
the real world, intelligence is really the ability to efficiently access
and manipulate the information you have stored. This ability to
quickly pick the correct starting points in the knowledge base minimizes
the amount of reasoning you need to do. What I really want is a "perfect
search engine" that will deliver relevant information and which is
highly adaptive to current context. Such a tool would allow me to
reason from a huge knowledge base.
This archive was generated by hypermail 2b30 : Fri Oct 12 2001 - 14:40:24 MDT