yathern

Oh - and colors are grey-blue to red, with red being pareto optimal

Is 'rats' a juggling jargon I'm unfamiliar with? Or do rats stare into the distance often?

The only downside (and this applies for SQLite as well) is that it runs too well that you can get some bad habits - or at least follow patterns that don't support horizontal scaling which you would want to do in production. A number of problems across different projects have bit me because I relied too long on SQLite (or PGLite) when moving from local dev to setting up cloud infra. This includes things like connection pooling, read replicas, consistency issues with sharding. Maybe all those people who productionize *Lite have a point!

> So there should be a prompt that can cause that sequence to be output

Is where I think I might disagree. For example, the odds of predicting verbatim the next sentence in, say, Harry Potter should be astronomically low for a large majority of it. If it wasn't, it'd be a pretty boring book. The fact that it can do this with relative ease means it has been trained on the material.

The issue at hand is about copyright and Intellectual Property - if the goal of copyright is to protect the IP of the author, then LLMs can sort of act like an IP money laundering scheme - where the black box has consumed and can emit this IP. The whole concept of IP is a little philosophical and muddy, with lots of grey area for fair use, parody, inspiration, and adaptation. But this gets very odd when we consider it in light of these models which can adapt and use IP at a massive massive scale.

I absolutely think this is the right approach. The glasses which do 'blue light filtering' which barely change your perception are clearly placebo, but a very strong redshift I think is obviously a different creature.