Google’s AI overviews are pretty hilariously not ready. Maybe they never will be. Generative AI is a powerful, transformative technology, but that doesn’t mean it should be used everywhere! Sometimes dumb code is better.

@manton I'm not sure what they even think the value of these "AI summaries" is. Their LLMs are no different than anyone else's. They have no way to self-assess correctness. So instead of just presenting search results, they add a fabricated cherry on top that will fool many people. How is that a good idea?

@manton I'm not sure what they even think the value of these "AI summaries" is. Their LLMs are no different than anyone else's. They have no way to self-assess correctness. So instead of just presenting search results, they add a fabricated cherry on top that will fool many people. How is that a good idea?

@manton I think it's actually a pretty scathing review of Google's search results. The AI isn't randomly recalling that pizza is made of glue. It's reading that from the search results.

@manton they're blowing it. it's like when the web came out and bill gates made every product do something with the web, and there was a lot of bullshit. they eventually sorted it out and had one great product, the mac version of their browser. no matter how much google was caught flat-footed with AI, they can afford to think very long term and they should.

@manton also, they did what every deeply entrenched dominator does over a decade or more of dominance. they destroy the thing that made them so dominant. it happens over and over.

@manton AI should only be used in low-stakes situations imo. I saw it used in YouTube the other day as a way to summarize what YouTube chat was talking about when I joined a stream. that's a dope use case! but it's low stakes.

@tylercheek.com Low stakes is a good way to think about it. It clearly has value but not everywhere.