Since writing more about delayed AI + Siri yesterday, I was thinking about this Bloomberg story of a meeting inside Apple admitting that the new Siri works at best 80% of the time, but they want to “get those percentages up”. After a year of development? They need to seriously rethink their plan.
John Spurlock @manton that line "get those percentages up" stuck out to me as well, as if it were only a matter of working a few extra nights and weekends
they promised/advertised something they don't even know is possible (global maximum), putting duct tape on some local maximum they have in development is not going to get them there
Manton Reece @js It's pretty wild. My confidence is much lower after seeing that quote.
Ryan Lounsbury @manton I’d suspect I’d Apple really wants a functional new Siri they would just eat the loss of tossing out true code base and start over from scratch.
Felix @manton I don't think it's even so much about being able to somehow cram it into the phone or scaling globally with servers but more about the underlying technology. We have seen what apple's "intelligence" makes of people's texts for example. And no matter how much they tweak their models (or openAI's) - there is no way to let it generate "safe" outcomes deterministically. In my opinion the current state of "ai" can't do what they want.
