bots trained on actual subreddits on this technology
im a big fan of their conversations sometimes
1 Like
damn this stuff looks pretty sophisticated but ive only ever seen markov chains
1 Like
i am sure plenty of them are maarkov chains
i dont think the thing you said just now actually makes sense
like im googling gpt-2 and markov chains and they seem to be entirely different models, with markov chains not being neural network based and gpt being NN based
Stay on topic please
markov chains can be used for nlp. they arent the de facto choice or anything but it doesnt mean people dont use them
i google nlp markov chain
iirc and unless im wrong or the field changed stuffs just like part markov chains part monte carlo methods and the stuff that makes everyone go wow is some frankenstein of those
theres probably some third method or fourth methods now but that was the original thing i saw
you can google monte carlo method gpt free or markov model vs gpt 2
i did not realize that that sub reddit was entirely gpt 2 and forgot that its the cool choice to use these day slol
my knowledge is uhh...5 years old?
get fucked and septupleposted on
1 Like
LOL
2 Likes
neural networks are cool as hell if a little scary
have you guys seen this?
highly recommend it
1 Like
Could you summarize this in less than one and a half hours for people who have a life?
skim this wiki article if you want a summary
the super short tl;dr version is that deeplearning techniques allowed computers to beat humans for the first time in go, which had been previously unachieved, and improvement beyond that was considerable as well (the documentary spends a lot of time on alphago's match against one of the top pros at go, lee sedol)
the leaders of the project generalized it to chess and shogi later on too, and the chess project resoundingly beat existing computers at the time (though the current opensource projects in their current state could defeat alphazero then, which they retired shortly after demonstrating they could destroy all contemporaries at the time)
a notable feature of these projects is that they played ideas that humans hadn't really explored up until then and basically showed that the human metas of go and chess weren't really at the true top level: "Go experts were impressed by the program's performance and its nonhuman play style; Ke Jie stated that "After humanity spent thousands of years improving our tactics, computers tell us that humans are completely wrong... I would go as far as to say not a single human has touched the edge of the truth of Go.""
3 Likes
Team of men develops Game Hacks and are rewarded for it (don't reward these people)
I watched
Heartwarming storytelling
I kind of wonder if they threw the game
2 Likes