Actually, that’s the high-value model. Imagine you have a bunch of LLMs tuned to different sensibilities that match great jurists, Oliver Wendell Holmes Jr., Learned Hand, maybe Aristotle to mix things up, maybe a real jurist. And your attorney tunes their arguments to be persuasive to whatever model they believe is dominant.

It’s a short leap to comparing model scores to determine a quick and dirty settlement “winner” which really isn’t that far from manual processes.

Lawyering will look different, but there definitely will be lawyers. Judging on the other hand…. Judging is the one I wonder about.

I imagine it would involve 1000s of LLMs outputting a judgement and then if there were significant disparities it would get flagged in some manner.

That's actually the plot of Minority Report, a lot of people think it is about "what if computers could predict crime" but it is really about "What do you do when your 'omniscient' machines disagree with each other".

Either way the idea of getting sent to prison and having 0 human interaction is terrifying.

That is something I hadn’t even considered. That is super scary; Part of me thinks it’s inevitable. People famously lack any sort of empathy for the falsely accused until it happens to them, so why wouldn’t they vote for a “save the children: use AI judges!” bill in 10 years?

Isn't there a Chris Pratt movie about this coming out in January?