[ExI] GPT-4 on its inability to solve the symbol grounding problem

Jason Resch jasonresch at gmail.com
Sat Apr 8 09:52:13 UTC 2023


On Sat, Apr 8, 2023, 3:01 AM Gordon Swobe via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> On Fri, Apr 7, 2023 at 10:42 PM spike jones via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
> back around the mid 90s, the chess players tripped over a similar
>> question, as the software was getting so good at our favorite game.  There
>> were those who argued the software wasn’t really playing chess in the sense
>> we understood it, but was only trying a bunch of possible moves and
>> calculating.  They thought merely calculating wasn’t really playing chess
>> as we understood it.
>>
>
> In 1999, I led a team of computer chess enthusiasts in a global effort to
> beat Kasparov. The game was one move per day online at MSN. Moves were
> decided by vote. I had about 12 guys on my team from all over the world,
> each of us running his own favorite chess software. I was running Fritz.
> Each night, in email, I would assign a possible line to each member of the
> team to analyze overnight, then in the AM we would look at all the
> evaluations and decide which move to recommend to the world team.  The
> world lost, but those were some of the most fun and intense four months of
> my life.
>
> You can see my name mentioned here in the 4th paragraph.
>
> Kasparov versus the World
> https://www.chess.com/blog/ThummimS/kasparov-versus-the-world
>

That's fascinating I never heard of that. It's quite incredible on
Kasparov's part.

Perhaps it reveals something about the inefficiency of committees, or the
lack of a cohesive strategy, or the weak additivity properties of ELO
scores?

I'm interested to know why you think 50,000 people working together could
not beat him. Perhaps humanity will face a similar problem facing even an
only-slightly-superior (1-2 orders of magnitude smarter) AI.

Jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230408/102648a4/attachment.htm>


More information about the extropy-chat mailing list