From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Fri Oct 04 2002 - 05:26:10 MDT
Lee Corbin wrote:
> People seem smart enough that most adults lie on the
> positive side of an interesting watershed: I think
> it likely that most adults and some children have
> reached the threshold of universal intelligence.
>
> I will call an entity universally intelligent if it
> is possible for that entity to understand *anything*
> if provided enough time.
In the Turing sense, a simple CPU with infinite memory and running time is
"universally intelligent" because it can simulate a mind that understands
something. But - and this is the confusion on which Searle's Chinese Room
fallacy rests - the CPU does not possess those cognitive patterns that we
associate with high-level understanding. A human in a Chinese Room
simulating a Chinese-speaking upload might not ever have the vaguest
comprehension of the high-level intelligence that lies above the 1s and 0s
being simulated. The human might never even catch a glimpse of the neural
level.
Are there limits to what a human can *understand*, as opposed to what a
human can Turing-tarpit? It seems to me that the answer is obviously yes.
There's a limit to the complexity and class of patterns that human
neurons will learn to recognize, and a limit to the number of abstract
elements that can be simultaneously held in short-term memory. If you go
over that limit, the human might be able to simulate the pattern by
simulating the low-level patterns one element at a time, but would never
be able to abstractly understand the higher-level regularities.
ObShotAtProof: For any pattern, there exists the shortest Turing machine
that describes that pattern. If that shortest Turing machine is larger
than the human brain, then while the human might be able to simulate the
Turing machine by swapping low-level elements in and out of consciousness
using an infinite paper tape as auxiliary memory, the human will never be
able to understand the system as a whole.
-- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:17:24 MST