I don’t have the thousand upon thousands of dollars to upgrade my still working iPhone 11, iPad 9, 2020 MacBook Air, Apple Watch 6, Apple TV 4K and Apple TV HD. Digging the Fitbit Coach, however…
I can just picture this either changing the passwords and not recording the new ones or thinking it has changed the passwords, recording the new ones, when it hasn’t.
The ‘agentic’ password change thingy seemed the most questionable to me. In order for it to work effectively, there needs to be a trust layer on both sides. I suspect that many websites aren’t setup for this trust layer yet, especially financial institutions.
The tech universe is in a very different place than when Apple made AI promises (unfulfilled) over the past several years. I don’t need a robot making appointments for me, or writing shortcuts, or most any of the things announced this week – they all feel like parlor tricks with no substance. I prefer having OS / hardware built to make using the frontier models safe and secure. (Because i don’t trust the model makers to do that effectively on their own.)
Updating the passwords without having to do the work (I’m sure it’s not actually AI but normal code) sounds pretty awesome.
(Edit: does sounds like it’s literally a headless browser tapping into the websites and updating the passwords. That is wild. I had imagined it was Apple working with a selected amount of providers with a standard for updating the passwords)
This has me curious: What do you mean? Local hardware can’t run frontier models and I doubt that there’s anything truly meaningful that an OS could do to make the use of could based models secure (especially since LLMs have effectively removed the barrier between the control plane and data plane).