Superhuman AI Cracked An ​"​Impossible​"​ Game

6 views
Skip to first unread message

John Clark

unread,
Dec 23, 2022, 12:03:44 PM12/23/22
to 'Brent Meeker' via Everything List
John K Clark    See what's on my new list at  Extropolis

Lawrence Crowell

unread,
Dec 23, 2022, 5:12:21 PM12/23/22
to Everything List
I played Stratego as a kid. The play pieces have a value only the player knows, and some of them are bombs. This is a game where probability plays a role. This relies upon Nash equilibrium. This makes me speculate that a Bayesian regression should converge to a value correspond to the Nash equilibrium. I suspect this system uses something of that sort.

LC

Brent Meeker

unread,
Dec 23, 2022, 7:49:11 PM12/23/22
to everyth...@googlegroups.com
One of the problems with using AI neural nets to solve problems like this is that we don't know what method it uses or even whether it's possible to find something we would call "a method" within the weightings of its network.

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 everything-li...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/everything-list/0cc9ae65-23bb-428f-bea5-633908e89f33n%40googlegroups.com.

spudb...@aol.com

unread,
Dec 23, 2022, 7:54:23 PM12/23/22
to everyth...@googlegroups.com
Ah! The Waters of Thought!

Is it something to invest in? I'd say no, because it's not like something that would ever fit in a phone. It's akin to the super-cooled physics that quantum architecture requires to function. Still, may be a quantum competitor some day?



Lawrence Crowell

unread,
Dec 23, 2022, 9:04:17 PM12/23/22
to Everything List
On Friday, December 23, 2022 at 6:49:11 PM UTC-6 meeke...@gmail.com wrote:
One of the problems with using AI neural nets to solve problems like this is that we don't know what method it uses or even whether it's possible to find something we would call "a method" within the weightings of its network.

Brent


In part the method is something that is emergent. The AI or deep-learning AI neural network settles into some extremal condition that is or approximates a "method."

LC

John Clark

unread,
Dec 24, 2022, 8:34:53 AM12/24/22
to everyth...@googlegroups.com
On Fri, Dec 23, 2022 at 7:54 PM spudboy100 via Everything List <everyth...@googlegroups.com> wrote:
 OK I admit it, that link was good, I wish all your links were that good.

  John K Clark    See what's on my new list at  Extropolis
tlg


Lawrence Crowell

unread,
Dec 24, 2022, 8:52:49 AM12/24/22
to Everything List
On Saturday, December 24, 2022 at 7:34:53 AM UTC-6 johnk...@gmail.com wrote:
On Fri, Dec 23, 2022 at 7:54 PM spudboy100 via Everything List <everyth...@googlegroups.com> wrote:
 OK I admit it, that link was good, I wish all your links were that good.


While the link is good, it has nothing to do with the subject here.

LC

spudb...@aol.com

unread,
Dec 24, 2022, 5:56:37 PM12/24/22
to goldenfield...@gmail.com, everyth...@googlegroups.com

The subject here is the liquid switch to utilize terahertz radiation versus traditional semiconductors. That's all and nothing else. 

--
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 everything-li...@googlegroups.com.
To view this discussion on the web visit
Reply all
Reply to author
Forward
0 new messages