CJ Lemire: "Sped up this lengthy process, still needs to be checked by a human" is the actual use case for generative AI. But that's not what the hype would have you believe.

"Sped up this lengthy process, still needs to be checked by a human" is the actual use case for generative AI. But that's not what the hype would have you believe.

One of the most fascinating things about generative AI is the businesses and states that are using it are largely doing “we’ve sped up this lengthy process”, whereas the tech companies’ largely seem to be going “we made this thing you like worse”.

And if that's what it was sold as, massively less people would be bothered by it in my experience.

I was part of a meeting recently where a group that did evidence synthesis of clinical trials was talking about how they used generative AI. And I could feel the tension leave me when they got to "and then we get a clinician to check the results"

100%. It should be sold as a hammer, not a magic wand. But then they couldn't charge so damn much for it, I suppose.

It's because they spent billions of dollars to make a new tool they didn't exactly understand, and it turned out to just be a hammer.

I mean it's not really a new tool, it's high dimensional algebra. They knew what they were doing. Then the marketing department heard about it.

The people making it knew, but none of those people have been the ones selling it or sinking billions into funding it, lol.

Yeah. In other words, capitalism continued its hot streak of ruining everything.

Something that frustrates me immensely is the lack of understanding that for a *lot* of knowledge work the human will still have to do most of the analysis to accurately check the output (i.e. it doesn’t actually save all that much time)

Post