
An LLM is a statistical model that predicts what makes sense to go next. If its training data usually indicate that tokens like that should come next, it sends them.
Reddit -> Beehaw until I decided I didn’t like older versions of Lemmy (though it seems most things I didn’t like are better now) -> kbin.social (died) -> kbin.run (died) -> fedia.
Japan-based backend software dev and small-scale farmer.

An LLM is a statistical model that predicts what makes sense to go next. If its training data usually indicate that tokens like that should come next, it sends them.

I live in Japan and see most of those every day. Passing things between chopsticks (and, as I found out the hard way, by extension two sets of chopsticks used on the same dish), and stabbing them vertically into something are the two to care about.
Jellyfish tastes OK. It’s more a texture than a flavor. I have no idea which species I’ve been served. Why, yes, I do live in eastern Asia where something trying to be inedible to humans is taken as a personal affront.
Hadrosaurus in a Hurricane? T-rex in a Typhoon? Centrosaurus in a cyclone?