I don’t see why chat gpt always include those black and white squares, besides some confusion with some other strategic games. Shameful
chat gpt only generates the text, it doesn’t draw the picture
dalle 3 do, and its weak old model already now
Not bad for a 5 stones handicap game on 11x11 board!
It must be pair go. The dragon couple appear to both be playing white, so the stalks must be playing black
Not sure I would have wasted 3 stones in the top left corner though
11x12, but I won’t judge them for playing crazy board variants.
Hand is trying to get Go stone from Go bowl. Cat paw very fiercely defends the stone with exposed claws.
Here is the true capacity of the Image3.0 generating Go board position. It’s improving a lot, but had to try quite a few times, and often with artifacts on stones or the board (and very limited and specific art styles repeatedly). And often when it gets the board size right, the stone wrong, or the stone right, but the size wrong. (Ideogram-v2, and recraft-v3 sometimes would get good enough images that players would be able to recognize the “attempt”, but many details are very wrong, not just the positions, but like stones not on the grids, not the right colors, or even strange spacing for the lines and the board itself). And using the keyword weiqi or baduk is very crucial, using the phrase “the game of Go”, the success rate of generating good images would drop significantly.
prompt: A poster with encouraging phrases related to weiqi, showing a diagram of a 9 by 9 board of the game of weiqi in the center with artwork surrounding it
Good one (and peek into the inner workings of the model with some kind of symmetry)
Slightly off, (and quite off for those in the background)
Very wrong
Just weird
Repeated style but wrong in many ways
Relatively good one (can photoshop out the artifcats, but not on 9x9 for some reason)
I’d say 1 out of 3 or 4, I can get a good enough output that can use photoshops to produce a poster that is somewhat usable for outsiders (who don’t know about board positions and realize they are just nonsense)
Replacing the board position with a real one should be easy, if you use Photoshop (or GIMP etc.) anyway.
For some, maybe, but sometimes the “surrounding artworks” would intrude the center, and often with questionable quality that is purely nonsense words or the background includes boards which are also off (and defeat the purpose of photoshopping the centerpiece).
https://openai.com/index/introducing-4o-image-generation/
Do you see 3 star points on the right? Black stones should be placed on them, that is what san ren sei is, they are too close to each other now.
…
fix it already and zoom out so floor around goban is seen
19x19 Go game position on thick vintage Japanese four-legged goban with 19 horizontal lines, 19 vertical lines and 9 hoshi points. Stones are placed on the intersections, not inside squares.
…
…
remove all stones
Did anyone try a go variant with stones played on the lines? I wonder if that’s doable.
Exactly!
So it seems from the picture.
That’s intriguing, because on an empty square you could move on any side, but a move on a side prevents moves on the adjacent sides.
There’s no room for playing on all four sides of a single square.
Make hoshi points slightly more bold.
Place Go bowls and lids on the floor.
Open Go bowls and place lids on the floor.
Place zabutons.
Frankly, in all of the recent pictures that you’ve shared, the hoshi points are unusually large for a traditional Japanese-style goban.
Now, let’s see if the AI is smart enough to properly place the captures in the lids. Maybe another angle to defeat super-human AI at Go??