I just made a AGI Spesific Domain: Chess
it have a bit same gameplay like stockfish on 28 games trained
it was hard to work alone
btw it have a low loss: 0.021719 and on my metadata it just do "capture": 2325,
"sacrifice": 616,
"check": 1190,
"promotion": 314 for new born ai
well it train using selfplay method so a large draw is normal right?
"outcomes": {
"white": 14,
"black": 4,
"draw": 72
},
Top comments (0)