On 12/25/2025 4:40 AM, John Clark wrote:
On Wed, Dec 24, 2025 at 8:40 PM Brent Meeker <[email protected]>
wrote:
*I want to start by saying that I wish we were having this
conversation on the list.*
*>> Whenever AI makes an advancethe goal post moves. At one
time making a brilliant chess movewas considered a preeminent
example of human creativity and abstract thought, but 25 years
ago when a computer made a brilliant chess move people
suddenly decided that chess had nothing to do with creativity. *
/> It was never considered and example of human creativity. /
*That is incorrect. In his 1979 book "Gödel Escher Bach" Douglas
Hofstadter argued that a computer to beating a chess grandmasterwould
require General Artificial Intelligence. He said chess mastery is a
byproduct of high-level cognition.It would need to understand symbols,
abstractions, and the world at large. To his credit much more recently
Hofstetter admitted he was wrong and said that the enormous success of
modern LLMs has "overturned" many of his core beliefs about the unique
nature of human consciousness and intelligence. *
/> The fact that humans have proposed several false tests for AGI
doesn't prove that there is no such test; only that it has yet to
be proposed./
*Many** tests for AGI have been proposed, but for some strange reason
whenever a computer beats a human at one of those tests the test
suddenly becomes obsolete and means nothing. I see no evidence that
humans possess some sort of secret sauce that a machine could not
emulate. *
Have you tried to think of one or have you just assumed that there can't
be one?
*>> And a great human grandmasterhas developed a specialized
chess program in his head, that's why he may be able to beat
any human being on the planet at the game of chess but he's
not especially good at anything else. There is however one big
difference between the human and the computer, the human
developed his skill after years of watching other grandmasters
and reading books about chess, but the computer program
AlphaZero needed no help from anything except simple
instructions that told it (or him or her) which moves were
legal and which were illegal. And just 24 hours later, after
playing millions of games of chess against itself,
AlphaZero was able to beatthat human chess grandmaster. *
/> But would that work at Poker?/
*Yes. No-limit Texas Hold'em is the most popular form of poker and
it's played in the World Series of Poker, and in 2017 the AI program
"Libratus" defeated the best human poker players in the world. It got
so good at the game by using something called Counterfactual Regret
Minimization. It played trillions of hands of poker against copies of
itself and after each hand it in effect looks back and asks "what if I
had played differently?" For every decision pointit calculates regret
for not taking alternative actions and gradually adjusts its strategy
to minimize this regret. This processconverges to a Nash equilibrium
strategy which is unexploitable even by a perfect opponent who knows
your strategy.*
I didn't know that; very interesting. And I'll bet it had an excellent
poker face.
/>>> The difference is that human specialization emerges
from a single, unified system/
*>>If that was true then Einsteincould've used language to
explain exactly how he got such wonderfully good ideas*
/>How do you know he couldn't? /
*I know that because I have read some of the stuff he has written
butI'm still not as smart as Einstein.*
I've read stuff he wrote too, but I don't recall him explaining how he
got good ideas except in few instances. I doubt you even know how you
get good ideas. It seems to me they just come into my mind as a think
of a problem.
Have you read the book that describes Einstein's patented ideas. He
patented quite a few inventions and in some cases, with the help of an
engineering partner, tried to make them into products. For example he
patented an airfoil. The only one that worked was a gas refrigerator.
Even geniuses don't always have good ideas.
Brent
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/everything-list/b0a8f285-5bf7-4c67-b6ce-2917810f523a%40gmail.com.
