[03:28]  <emmab>  test
[03:28]  <akash>  i have a 9 of hearts
[03:29]  <jujujujustin>  congratulations
[03:53]  <akash>  whaddup @hexwab
[03:53]  <emmab>  how are you hexswab
[03:54]  <hexwab>  I'm fine, just watching out of curiosity :)
[03:55]  <emmab>  welcome
[03:55]  <jujujujustin>  niiiice
[03:56]  <rangerscience>  we're trying to figure out why OpenAI chose this as their next game
[03:56]  <jujujujustin>  hexwab!
[03:56]  <rangerscience>  and why it's hard for AIs
[03:58]  <hexwab>  you're not on discord? most people tend to hang around there for post-game discussions &c.
[03:58]  <hexwab>  also there are a bunch of papers at https://github.com/Zamiell/hanabi-conventions/tree/master/academic-papers that explain the state of the art wrt bots
[03:59]  <rangerscience>  interesting
[04:00]  <hexwab>  I am not aware of any bots that attempt to play the conventions that people commonly use here (there are a whole bunch)
[04:00]  <rangerscience>  absolutely
[04:00]  <rangerscience>  but
[04:00]  <rangerscience>  openAI doesn't make bots
[04:01]  <rangerscience>  they make neural nets that learn to play games
[04:01]  <rangerscience>  chess, Go, DOTA so far, probably a few others? Starcraft, too, I think
[04:01]  <hexwab>  they're explicitly not wanting human intelligence in programming them, AlphaZero-style?
[04:03]  <emmab>  thank you for the papers, hexswab
[04:05]  <hexwab>  as in, the problem they're trying to solve is "make a machine that can learn", and hanabi is just a fun microdomain to use as a playground?
[04:05]  <rangerscience>  more than "fun", but yes
[04:05]  <rangerscience>  I think they've chosen at the next thing because it requires a theory of mind
[04:06]  <hexwab>  yes, that is what makes hanabi interesting for humans, and is also something machines are currently terrible at too
[04:07]  <hexwab>  e.g. the "intentional" AI at https://github.com/yawgmoth/pyhanabi that attempts to model this is pretty bad
[04:10]  <hexwab>  Zamiel, who made this server and who standardized the conventions used here, is also the one who collected the papers. sounds like he may be worth talking to; he certainly knows way more than I do
[04:11]  <hexwab>  (how did you run into this server? did you turn up as a group just to try the game out?)
[04:12]  <hexwab>  (hmm, others left, here may be a poor place to chat?)
[04:14]  <rangerscience>  we'll be on ABC tonight (PST)
[04:14]  <rangerscience>  not a lot, nit
[04:14]  <rangerscience>  *but
[04:14]  <rangerscience>  some
[04:14]  <hexwab>  what's ABC?
[04:14]  <rangerscience>  and the two-hour documentary we're in is on right now
[04:14]  <rangerscience>  err, American Broadcasting Channel, I think?
[04:14]  <rangerscience>  Something like that
[04:14]  <rangerscience>  non-cable TV
[04:15]  <rangerscience>  *non-cable US station
[04:15]  <hexwab>  oh right, misunderstood, thought it was some other place where you were hanging out
[04:15]  <rangerscience>  nah, we're in LA
[04:16]  <hexwab>  I'm not in the US but I'll try and locate a copy if you tell me what to look for
[04:16]  <rangerscience>  ABC "Screen Time by 20/20" or somesuch
[04:16]  <hexwab>  (you guys are researchers? journalists?)
[04:16]  <rangerscience>  we're only in... six minutes? or something
[04:16]  <rangerscience>  nah we do stuff around app addiction as SaaS products