coeneedell

There’s no reason to believe that companies running open weights models would use consumer hardware. I can speak on behalf of my employer and say that we’d vastly prefer to have enterprise grade servers running inference on premises.

Not really. He had a really good PR person that made people think this, but most technologists are focused on, y’know, actual technology

This is basically unreadable. The llm generated text feels like it’s dancing around an idea because it doesn’t actually have one.

They should get paid. The press should pay them. The college pays the press. The students pay the college. The press releases information for free.

I’m gonna need a more detailed argument than that.

I have the right not to sell poison to someone who I have reason to believe will use it to kill a third party. The idea of simply trusting the patron to be responsible makes sense when the patron is anonymous or a new contact. It’s generally good to assume good intentions in the absence of evidence, I think. If the government is not anonymous enough to get this treatment.

Violating the laws of our country while being in a position of public trust reaches a higher standard than “politicians I don’t like” IMHO.

I don’t see how you can make this claim when the budget balance at universities is extremely tilted towards stem. The budget for the kind of programs you seem to be complaining about come out to percents. Hard to say universities are “more interested” in very very small parts of their budget. What is happening now though is the federal government holds back billions of dollars in medical research grants, punishing, in some cases even killing patients, over “ideological” issues.

You made a throwaway account to say this?

Yeah I see now I was mixing techniques in two sections of that document

I haven’t really had this issue. UV’s recommendation is to mount the uv.lock and install those manages package versions to the container’s global pip environment. We haven’t had much issue at my work, where we use this to auto-manage python developer’s execution environments at scale.

“Agents that extract the data” Are we really reinventing data frame readers to have an LLM in the critical path?

Do you measure your results in a repeatable way? In a way where your hypotheses about accuracy are falsifiable? Or do they just “feel” right?