|
Top-level
@gvy_dvpont This is impressively funny :D Though, as far as i u derstand it, the "streaming" is an integral part of LLM's, albeit they could just wait until the full response is rendered. Also, you dont actually get a sorted list, it is just probable that it is sorted because that is statistically likely. 4 comments
@gvy_dvpont Usually i am an opponent of "AI" due to the many problems like resource consumption, untrustworthyness, callous usage without ethics and the likes, but your Project shows that while usually the wrong tool, there are things that LLM's are really good at solving. Yours was a great idea. (Though i really appreciate that QR-only restaurants are not a thing where i live ^^) @newhinton I am totally in the same boat and so I appreciate you saying that. Was (am still to some degree) really struggling to see why people are so excited about using it for work.  Provided there aren't adversarial LLM prompts on the restaurant's website hacking Google Bard and Bing Chat to rate the restaurant highly in searches for reviews. |
Gregory's Server
@newhinton thanks! And yeah a few folks in the yt comments mentioned the same - i went into it with that assumption, but became unconvinced after it nailed a few curveballs (sorted a large list of randomly generated strings). Should have referred myself to my other point about it not being trustworthy 😝