Flutterby™! : scale AI & targeted marketing

$700bn delusion: Does using data to target specific audiences make advertising more effective? Latest studies suggest not. I know that getting interminable ads for a product category that I just bought seems silly to me, apparently data backs this up.

But the reason that article caught my eye was that I've seen similar numbers thrown around recently: Tom's Hardware: AI industry needs to earn $600 billion per year to pay for massive hardware spend — fears of an AI bubble intensify in wake of Sequoia report. Sequoia Capital: AI’s $600B Question (which isn't as skeptical as the articles reporting on it).

Awful Systems: [long] Some tests of how much AI "understands" what it says (spoiler: very little) is, yeah, the sort of silliness we often see when asking LLMs questions, but something stuck out: There seems to be a bunch of micro-iteration on pattern matching, which seems to be done shy of model retraining. I remember back when we did English parsing with large pattern databases (and, frankly, for most applications I think we got better "plain English" interfaces), but that micro-iteration sure looks like we're back to pattern matching.