Linus Torvalds often says that he does not know how to do X (like install a Linux distribution, or other simple stuff). I wager that it's a status thing.
Linus Torvalds often says that he does not know how to do X (like install a Linux distribution, or other simple stuff). I wager that it's a status thing.
If anything, LLM's seem more resistant to propaganda than any other tool created by man so far, except maybe the encylopedia. (Though obviously this depends on training.)
The good news is that LLM's compete commercially with each other, and if any start to intentionally give an ideological or other slant to their output, this will be noticed and reported, and a lot of people may stop using that LLM.
This is why the invention of "objective" newspaper reporting -- with corroborating sources, reporting comments on different sides of an issue, etc. -- was done for commercial reasons, not civic ones. It was a way to sell more papers, as you could trust their reporting more than the reporting from partisan rags.
How would you know? My first thought is that the data on which LLMs are trained is biased, and the commercial LLMs enforce their own "pre-prompts".
If they can get away with a query that takes 2s to return a single row, they will be quite content and will not be bothered to look at the query plan.
It's a shame, because everything could be a little better with hardly any effort.
No, it's not. This goes against the whole thread and the article posted.
Phew! Thanks for clarifying.
Deleted Comment
Deleted Comment
[1] https://www.youtube.com/watch?v=qHGTs1NSB1s