Posted by
David Eric Smith on
Feb 11, 2020; 10:31pm
URL: http://friam.383.s1.nabble.com/Eric-Smith-s-interview-on-Jim-Rutt-s-podcast-tp7594472p7594483.html
Thanks Steve,
Yes, there is a certain thread of literature and experience that is always in my mind on this topic.
It began some years ago when David Krakauer waved a copy of Richard Novikovsky’s book Games of No Chance in front of me, and that was my introduction to the modern work that had been done at MSRI Berkeley on combinatorial games. All during my time at SFI, I was interested in finding some good merger of the low-dimensional but nonzero-sum games the economists almost-exclusively deal with, with the literature on combinatorial games that does much better at capturing what makes a problem “hard” to solve, as opposed to “hard” to clearly frame.
Somehow — and I don’t remember who played what role in it — we had Elwyn Berlekamp out to SFI for a couple of days, and I hosted him. It was like a visit from Erdos. After about 48 hours everyone else was worn out and Elwyn was going strong. But it started me reading a string of books by him and other MSRIers on mathematical Go, other work on mathematizing games, and also Conway’s On Numbers and Games, which I never put in the effort to understand but which Cris Moore understands through and through.
All that stuff is delightful, and I have wished to get to understanding it well enough to have new ideas in that space. But have not done so yet. Most interesting to me from Elwyn’s visit, and not contained in the books, was the experimental work he had done with world-champion-level Go players, to determine whether the values of game positions that he had derived from the mathematicization were pertinent to those that guided the players choices, and he claimed there was good agreement between the two. I haven’t tried to track down any literature that may ever have come out of that, and I didn’t get references from Elwyn at the time to even write down. From high-wattage people like that, there is just so much that comes across so fast, you can’t live enough lives in parallel to keep up with it.
But, it would be great to see some of those interests revived and developed further. They seem to have so much promise not yet realized. I wish there were anything concrete I could offer you by way of either ideas or references, but sadly I have none.
All best,
Eric
Eric-
Great interview!
Eric: You know what I like on this
though, I think back to the, I guess it was AlphaGo
competition with Lee Sedol in the computer human contest for
Go playing. I really loved Lee’s comment at the end of it,
where he was saying that of course those had been the most
difficult games he had had to play, but that he had never
enjoyed playing Go more than in those games because before, he
was the best in the world in a style of play that was
essentially established and playing the machine, it was
opening [inaudible 00:06:46] of play that no human would have
opened against him. It was giving him an insight into the game
that had not been available to him from anyone before. Apart
from the superb character that that demonstrates in the man, I
think that’s a good way to look at human-computer interactions
that we have all of these big branching structures. The
question is when will computational solutions open [inaudible
00:07:11] of play that human conventions were not exploring.
I really appreciated this
point/perspective. I distinctly remember two moments related to
this. The first was when the 4 color theorem was proven by
machine and there was a LOT of discussion about the implications
of that. The smallest of the conversation seemed to be the kinds
of *insights* that such a method of proof could elicit. I'm not
clear that any such thing came of this or any other automated
proof, but it seems possible? Surely you or someone else here has
a better handle on that.
At the 1983 Cellular Automata conference at LANL, there was
fairly widespread discussion of the problems of automated Go play
with speculations/assertions about just how hard the problem was
and whether it could ever be approached at the "atomic" level.
It warms my heart to hear Lee Sedol's anecdote about feeling like
he was obtaining a new insight into a game he had obviously
already dedicated a lifetime to understanding.
My own dabbling in the area of human-in-the-loop ensemble
steering is based on the assumption/hope that the coupling of
automated generation/analysis and human insight is in some way
transcendent of either approach alone.
- Steve
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe
http://redfish.com/mailman/listinfo/friam_redfish.comarchives back to 2003:
http://friam.471366.n2.nabble.com/FRIAM-COMIC
http://friam-comic.blogspot.com/ by Dr. Strangelove
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe
http://redfish.com/mailman/listinfo/friam_redfish.comarchives back to 2003:
http://friam.471366.n2.nabble.com/FRIAM-COMIC
http://friam-comic.blogspot.com/ by Dr. Strangelove