[caturopath]

I'm getting a lot of refusals these days from multiple LLMs on multiple fronts for silly stuff, a lot more than I had for a while. If this is where things are really going, I think open weight models have a big future.

[w29UiIm2Xz]

I have moderate success telling the model we are securely doing the thing I want to do, and then it seems on board because I'm saying the right words and seemingly am aligned with the guardrails. Basically, you can't express ideological opposition to the idea of security, but then proscribe what you want and anoint it as secure, and then it's happy. Then it focuses less on litigating whether or not it is secure.

[jojobas]

"H.R. 148867 makes all large language models subject to safety certification, introduces penalties for unlicensed training and use of uncertified models"

[bluescrn]

And China wins the AI race

[jojobas]

Who cares of the races when CHILDREN are at stake!

[verdverm]

The politicians still care about their races?

[parineum]

So many ridiculous bills get introduced in the house you can find one that says anything.

[jojobas]

You didn't think this one was real, or did you?

[parineum]

I don't doubt that it is but I didn't check. The point is that you can probably find a HR to propose allowing aliens to land in the grand canyon.

The existence of a house resolution mean that one representative wrote a thing, not that it's on the precipice of becoming law.