There isn't, some subreddits just use CSS to hide the downvote button, which is stupid because it advantages the people who "break the rules" by turning off custom CSS so that they can downvote people.

Reddit, like all popular social media, sucks for real discussion. It strongly promotes a groupthink/bubble mentality, and furthermore people/organizations are constantly manipulating the system in order to fabricate a "groupthink". But people find it easy and satisfying to use, so they use it.


Current machine learning tech is OK at answering questions like "does this contain ". So it's OK at detecting clear-cut things like spam and obscenity (though dealing with an intelligent adversary makes the task much more difficult), but it's nowhere near being able to answer more abstract questions like "is this post off-topic?". This is why Twitter & friends keep failing to reliably detect things like "hate speech", since that's too abstract a concept for current ML tech.

I don't think that the hardware costs would be particularly expensive for any plausible bitcointalk.org usage of ML, but I don't think that it's worthwhile currently. Maybe if there was a massive influx of blatant spam. When you hear about big companies pouring tons of money into machine learning, I think that's almost entirely them trying to make the tech less terrible by hiring either ML experts or people who help produce training data.