3 ms·
Current "AI" are statistical models, not AGI - that's why you're wrong there. To use them well you still need to know everything - whenever you prompt lazily y
by ffsm8 10mo ago
Current "AI" are statistical models, not AGI - that's why you're wrong there.
To use them well you still need to know everything - whenever you prompt lazily you're opening yourself up to a fuckton of technical debt.
That might be acceptable to some, but is generally a bad idea
If it was AGI, you'd be right though...
- palmotea 10mo ago> To use them well you still need to know everything - whenever you prompt lazily you're opening yourself up to a fuckton of technical debt. > That might be acceptable to some, but is generally a bad idea That's my point: to use them effectively you need to know everything, but to use them heavily puts you in a situation where that knowledge atrophies (e.g. the OP's statement "every time I use AI, I feel like I'm getting a little dumber"). The bosses want the results now, and don't mind if in a few years you're much less capable (and maybe not capable of getting effective results anymore). If AGI comes soon enough, their bet will have paid off.