Go memes!


#569

About that ^^^^ :


#570

Sentemas? That’s the best he could come up with?

I mean at least pick the obvious Sente Claus, or perhaps sing meijinglebells (alternatively jigobells), or… wish everyone a feliz nadare. :disappointed:


#571

It was actually the inspiration. Well done. :joy:


#572


#573

Mindblowing AI
(Source: Nick Sibicky on Youtube)


#574


Etiquette protocol regarding undo in correspondence tournament
#575

— But they are false eyes.
— It’s just a flesh wound!


#576


#577


#579

2qju4o


#580


#581

2s40ih


#582

2s4nh1

And another on the same topic


#583

DHaFD2g


#584


#585


#586


#587

[no words needed]


#588

I’m just a measly kyu player. And while I’m computer-savvy—calling myself a power user, having created print media since 1981—I understand little of all this AI stuff. Now they’re talking about “weights” for Leela, etc., and one of the files seems to “weigh” 27 MB, but that’s not what “weight” is about, that much I know. It’s also not about the weight of the ones and zeroes on our devices. Oh well, maybe it actually is, in a metaphorical way. Anyway, this just came up in my mind when I read a LZ thread in that FB group.


#589

AI are trained using evaluations. For example you could train AlphaGo to be more aggressive and give it more rewards for attacking than for defensive play. Taking stones would then be “weighted” more than saving stones.

The results of tweaking these kinds of settings (aka weights) are the filters that this style of AI uses to decide on its next move. Changing the weights will result in a different set of evaluations, and thus different moves in specific situations.

I hope that clears up your questions?