[FRIAM] AI art

[Jon Zingale]

Chess tends to have a pretty specific culture relative to other similar
games. Often whenever I find chess happening in public spaces I will stop
to watch a game and occasionally a player will ask if I play. I don't play
chess, but I know enough of the rules that I enjoy speculating as to what I
might do in a given board position or what the players might be thinking
themselves. Typically, my response is that I do not play, that I would love
to learn and I would love a teaching game. Players almost never take me up
on the offer. I get the feeling that teaching games are not part of the
culture, at least not here in the United States. I get the strong feeling
that this is because chess players tend not to see the game as beautiful,
something to be intimate with and share. The only teaching game I have
received to date was from a Georgian who I believe does see the game as
beautiful. While I am not a chess player, my love of go gives me an
appreciation for strategy games and I find that the audience for public
displays of these games are typically others who engage in speculation
similarly.

It really doesn't matter to me whether or not I am watching a human game or
not. My go server, for instance, is deep in the Turing challenge. The
server offers not only the opportunity to play mostly anonymous games with
others, but also to be a spectator to live games on the server. It is often
completely unclear as to the ontological status of the players and lines of
differentiation can be drawn nearly everywhere. There are degrees of
cyborg, degrees of experimentation versus repertoire, degrees of deception
at nearly every level. My go playing friends and I will sometimes attempt
to guess the nature of the bot we are witnessing, the degree to which it is
MCMC or DCN or simply someone's idea of an entertaining and completely top
down rules based engine.

When I watch games between strong professionals online (sometimes on
servers, NHK, or Twitch) there can sometimes be a significant difference in
the rankings of both players. The stronger player is in effect giving a
teaching game to the weaker. Often both players are part of the same study
group within their organization and while both are interested in winning
the match, they both have a dedication to a kind of scientific discovery of
the game. They are helping each other to see further. I have no hope of
seeing what they see, but in my engagement with their game I am hoping to
also see further.

Perhaps a year ago now, I mentioned on this forum a discussion I had with
Michael Redmond 9-dan on his twitch stream, late one night. He made it
clear to me that while the strongest AI bots on the planet are very good,
they likely can only see 10-15% into the game of go. At the time of Lee
Sedol's retirement games (in which he chose to play a specially made AI),
the strongest players on the planet were 30 points weaker than AI. Today,
with AI study and related narrative construction, humans have reduced the
gap to 10 points. Further, AlphaGo discovered new joseki by exploring
directions long thought (200 years or more) to be deadends. Strong players
have since learned to understand these openings and those that play them
tend to win more often than those that don't. This suggests to me that the
AI is capable of finding large scale optimizations that we can leverage
beyond being simply local, tactical and narrowly defined computational
advantage.

The Go community (and here I mean strong amateurs to top professionals)
study with AI, play with AI (competitively and collaboratively), and seem
to accept AI as both a partner and a tool. I sometimes watch MassGo on
Twitch play games where each player chooses a particular AI engine and uses
their engine to suggest three top moves. Then the players choose for
themselves the move that they find most interesting. Once the game is over
they review, co-constructing narratives alongside a third AI analysis tool.
I am not sure this kind of thing happens in the chess world, but it does
remind me a lot of the kinds of human-computer interactions that do happen
in art.

I suspect that in the long run, for those communities open enough, purity
will matter less and less, while a refinement for what is novel and
interesting will become more diverse and specific. In many ways, I believe
that it is what we want from studying a game and the agency our tools
afford us that determines the excitement we feel in engaging those tools.
At present, I am happy with the new directions my community is advancing
alongside these AI tools.

Last and tangentially, I assume many here have already listened to the
recent Ezra Klein podcast with Holly Herndon. I appreciate the sensibility
Holly brings to not only uses of AI in art, but also the clarity with which
she seems to understand her own relationship to art in general. The podcast
begins with Ezra highlighting that mimicry is the present and dominating
state-of-affairs for AI art, but that there are some who are pushing to
create something we can more honestly call generative.

https://www.youtube.com/watch?v=4MJ2D9uCLLA&t=2374s&ab_channel=NewYorkTimesPodcasts

Jon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20240624/86bb1093/attachment.html>