Remix.run Logo
▲ grumbel 3 hours ago

> and a thousand others with no sign of stopping.

StackOverflow used to be one of the big ones in that area and that is as good as dead. I see the same coming to TVTropes, fandom wikis, movie dbs and Co. once AI gets a good at interpreting video and getting access to some sizable movie libraries. When it comes to /r/tipofmytongue/ style questions AI is already much better than large groups of humans together, though only if the necessary information is available on the public Web.

Lack of access to content outside the Web is really what is hampering progress so far, but that could go away very quickly whenever any of the big players decides to let AI run wild on whatever libraries they have.

As for humans, I just don't see how they could compete. Wikipedia took decades to write, millions of volunteers and Grokipedia (despite being a bit of an unmaintained mess) caught up to its article count in a few months. On top of that comes that all those projects aim for a neutral POV, which is very non-human to begin with and only going to benefit AI.

▲cardanome 3 hours ago | parent | next [-]

> On top of that comes that all those projects aim for a neutral POV, which is very non-human to begin with and only going to benefit AI.

LLMs are not more neutral than humans. Not only because their learning material is biased but also because they can develop unexpected biases due to the non-deterministic and emergent nature of generative AI.

The idea that an LLM would be more neutral and less partial than a human is a very dangerous and very wrong one. It misunderstands the very nature of generative AI.

And while some people are able to tolerate AI generated text, a lot of people would never read Grokipedia. I would not. Despite great progress, generative AI is not able to create images that don't look absolutely disgusting in the eyes of actual artists. And the same is true regarding text output for anyone who cares about writing.

▲arandomhuman 2 hours ago | parent | prev [-]

It’s almost like it’s not that difficult to scrape Wikipedia and turn its articles into LLM generated slop. I wouldn’t call musks wiki clone catching up by any means to Wikipedia and definitely not politically neutral.