If I have to fill a form for anything that matters, I'm doing it by hand. I don't even use the existing historical auto-complete stuff. It can fill stuff incorrectly. LLMs regularly get factual stuff wrong in mysterious ways when I engage with them as chat bots. It might be less effort to verify correctness than type in all the fields, but IMO there's less risk of missing or forgetting to check one of the fields.

▲

dawnerd 6 days ago | parent [-]

I've had on so many cases autocomplete forms puts something in a field it shouldn't and messes up a submission. I've had it happen on travel documents that caused headaches later at the airport - especially if it fills in a hidden field because some bad web dev implemented it poorly.

▲

charcircuit 6 days ago | parent [-]

It gets it wrong because the current "AI" for filling out forms is extremely weak and brittle compared to the general language models we have now.

▲

recursive 5 days ago | parent | next [-]

Language models seem pretty weak and brittle in my interactions with them too.

▲

oneeyedpigeon 5 days ago | parent | prev [-]

Do you have an example form field that a general language model could fill out better than a human + highly focussed deterministic algorithm?

▲

charcircuit 5 days ago | parent [-]

Ecommerce checkout. Filling out my address, billing adress, and credit card information. Things like drop downs or different formatting can mess up the current basic ones, but it really shouldn't be that hard for AI to figure out how to fill out such information it knows about me into the form.

	▲	oneeyedpigeon 5 days ago \| parent [-]
		I think I've found those unreliable in the past, but much more reliable as time goes on. I can't really remember the last time an address or credit card info was mishandled by autofill. I get that addresses can be poorly defined, but for one you've entered yourself, that you just want to be re-entered, I don't see why we can't solve that problem without AI.