LLMs have legitimate uses today
No they don’t. The only thing they can be somewhat reliable for is autocomplete, and the slight improvement in quality doesn’t compensate the massive increase in costs.
In the future they will have more legitimate and illegitimate uses
No. Thanks to LLM peddlers being excessively greedy and saturating the internet with LLM generated garbage newly trained models will be poisoned and only get worse with every iteration.
The capabilities of current LLMs are often oversold
LLMs have only one capability: to produce the most statistically likely token after a given chain of tokens, according to their model.
Future LLMs will still only have this capability, but since their models will have been trained on LLM generated garbage their results will quickly diverge from anything even remotely intelligible.
The problem is that people selling LLMs keep calling them AI, and people keep buying their bullshit.
AI isn’t necessarily bad. LLMs are.
there’s nothing you have to remember which makes it more convenient for you to use
Unlike my devices, I always have my brain on me. Devices are much more easily lost or stolen than memories. I often might want to access sites using my account from third party devices which I don’t want to be able to use my accounts when I’m not using them.
I just can’t understand how using passkeys (or password managers, for that matter, massive single points of failure that they are) is supposed to be in any way shape or form more convenient than simply remembering a passphrase (which can easily be customisable for each site using some simple formula so that no two sites will share the same but it’ll still be trivial to remember).
Both password managers and passkeys seem like colossal inconveniences and security risks to me when compared to passphrases, frankly. And if you want extra security there’s always two factor authentication (with multiple alternatives in case you don’t have access to one of them, of course; otherwise you might as well delete your account).