Life In 19x19

Test carried out by Wing on the Go AI discord:
Managed to get hold of a 3080ti + 4090 PC plus another 3090 + 4090 PC, ran 3 batches of tests last night.
Networks used: kata1-b18c384nbt-s5832081920-d3223508649.bin.gz (for simplicity I'll refer to it as 583.gz) and b18c384nbt-optimisticv13-s5971M ...

Translation by GPT-4 which should be better than Google Translate:
Interview with Jan Šimara, the first Czech professional

The Czech Republic has its first professional player. The decision was made a week ago at the 7th European Professional Qualification, in which Jan Šimara surprisingly ...

Well, to be precise, it wasn't a human who found the loophole but an adversarial AI trained against a frozen target (KataGo in this case). It by no means takes away from the AI's strength in a normal setting. Leela for example was known to be weak against ladders but that doesn't mean it would still ...

Try using the new b18c384nbt-uec.bin.gz net, it's 95MB in size and currently the strongest available net. As of now it's only available through KataGo's github page, but in a couple days/weeks it should replace 60b as the main training net. You'll need KataGo 1.12 to run it https://github.com ...

This new exploit has already been noted by lightvector and KataGo now implements similar positions to these (only the cyclical ones, not the passing ones) into a small percentage of its training games (roughly 0.08% of games). An example of a training game with a cyclic topology group position ...

Seems to work now, try this link https://drive.google.com/file/d/1VlRD3i ... sp=sharing. I had compiled the katago.exe correctly but I forgot to replace the old one in the folder

Tried it but it only seems to work up to 37x37, anything above it and KataGo just ignores the extra space
Here's the katago.exe anyways

https://drive.google.com/file/d/1H3xT3q ... sp=sharing

You should be using the zlib from this link https://github.com/kiyolee/zlib-win-build . Also in ZLIB_LIBRARY you're supposed to point to it the path to the libz.lib file, which you get by running the visual studio solution inside the build-VS2022 folder which is inside the zlib-win-build folder ...

Turns out it does work, it just took 5 min+ for it to start evaluating for some reason. I had assumed it wasn't working but when you said it worked on your end I tried leaving it running on mine to see if it would sort itself and it did.

Link to the katago.exe:

https://drive.google.com/file/d ...

Result of a 100 games match between the latest 18b vs the strongest 20b, 400 visits/move, Chinese Rules and komi is set to 7.0 which allows for draws. A draw is counted as +0.5 points for both players:
18b vs 20b:

92.5 : 7.5

From 18b's perspective:
Wins: 92
Draws: 1
Losses: 7

Elo Difference ...

It doesn't seem to work sadly. I tried compiling it with the cmakelists.txt you provided but it won't display the evaluation with anything larger than 29x29

Thorough test done by @Medwin on the Go AI Discord, using 400 visits/move with a total of 2800 games played.

image.png

As expected, the new b18-uec is slightly stronger than 60b at equal visits (much stronger with equal time) and is expected to grow even stronger as its training run progresses ...

katago.exe bs29 OpenCL - Compatible with the new 18b networks. Drop it in the 1.11 bs29 folder and have it replace the old katago.exe

https://drive.google.com/file/d/1Da18Hb ... sp=sharing

Result of a 100 games match between the latest 18b vs the strongest 40b (S1184), 1s/move (averaging 200-500 visits/move), Chinese Rules and komi is set to 7.0 which allows for draws. A draw is counted as +0.5 points for both players:
18b vs 40b:

65 : 35

From 18b's perspective:
Wins: 58
Draws: 14 ...

New KataGo 18b out, supposed to as strong as 60b at equal visits while also being faster than 40b, which means it should be stronger than both in an actual game.

https://cdn.discordapp.com/attachments/583775968804732928/1046527064695648306/b18c384nbt-uec-20221121b.bin.gz

Also test game: KataGo 1 ...