Hacker News Reader

Comment by bluecalm | Hacker News Reader

Why not use an LLM to filter research they don't want to fund? Keywords list seem like the dumbest way to go about it.

There is a possibility that some just did not care about the whole financing endavour.

If you don't particularly care for the life of cows, "looked a bit sad" and other false positive triggers become a sufficient marker for suspected "mad cow disease".

You answered your own question.

They did though, an llm filter is just as dumb. Don”t they have those interviews with the edgelord doge employees trying to justify the stuff they killed?

Adding to this, i think llms are just part of their game, they know well enough they aren’t accurate, but it plays into comments like this, where we’ll think an llm would make a smarter decision, but it’s just a smokescreen, they’re already looking at gaming the bots, musk trying to get grok to abide by his petty worldview is the most obvious ploy. But they’re already obfuscating sources of information as a matter of course, soon, we won’t know anything anymore, we’ll have to trust their machines.

I think their fixation on the machine is part accountability-laundering, part obfuscation as you say. They want a monopoly on not just information, but analysis, and breaking all information infrastructure that does not flow through their pet ideology robot is part of that.

I think they don't want to cut funding to math projects or engineering projects. They want to cut funding to research that goes against their politics. Cutting math research because "inequality" keyword is a bad look for them and just gives free ammo to their opposition. If they employed an LLM that wouldn't happen and they could still pursue their political goals.

"Dumbest way to go about it" is SOP for this administration.

Reply on news.ycombinator.com