"Why spend so much money and energy on a technology that gets 1 out of every 5 answers wrong?"
Because we spent decades building, by-hand, software that still can't get any answers right. If #AI can get 4 out of 5 with minimal prior configuration, we'll fucking take it.
@LouisIngenthron
The things LLMs can do (but poorly) are things that can be scripted.
And they were scripted.
And LLMs can do them (poorly) because they stole those scripts for the training material.
Anyone willingly adopting this crap is an incompetent.
@androcat That's not how it works... If all it did was copy, it would have no utility. That's like saying that the stunt videos players create with GTA were scripted by the developers.
Emergent functionality has been an established phenomenon for a long time. AI just cranks it up to 11.
