David and Felipe on Hall of Impossible Dreams with a deep dive into PhysicsForums and Dead Internet Theory:

The LLM-written portions of this thread are 92% of the total word count. If this were a representative sample of the Internet, it would be reasonable to describe the Internet as ‘dead’.

A flood of AI slop on the internet has profound and unpredictable downstream effects. There's the economic impact of Google Zero, the social implications of treating others like NPCs when you can't be certain if there's a real person behind that account, and so much more. But one effect that sticks in the back of my brain on a low simmer is just distrust of these AI answer bots like Perplexity or Google's AI Overviews. Not specifically because I think they're hallucinating wrong information — which they most certainly are — but more-so because I wonder if they can develop good taste.

Whether it's Google wanting to "Do the Googling for you" or OpenAI's Operator that will puppeteer a web browser and take actions on your behalf it's clear that AI companies think consumers will get value by offloading more to these AI systems. However, in their current iteration I see a lot of these products as similar to Buddy the Elf finding the "World's Best Cup of Coffee". Will they be able to develop discerning taste or if I ask them to "Find me the best vacuum cleaner" will they just regurgitate AI slop affiliate marketing listicles? What happens if 80% of the search results become AI slop? I think most people have an intuitive sense of which of the ten blue links they want to trust but will AI be able to develop that sense?