Hacker News Reader

Comment by croes | Hacker News Reader

Some think AI helped with the formula

https://universeodon.com/@cryptadamist/114271484597938775

The tariffs might be a bad idea, but this accusation is ridiculous. For a simple methodology like this, it's trivial to prompt engineer a LLM to produce the same response. It doesn't mean that that's how the administration came up with the policy, any more than a LLM getting the same (correct) solution as a student on an assignment means the student used a LLM on that assignment.

I think the accusation carries more weight than you're implying because while correct answers are all similar having two students give eerily similar wrong answers is often used as evidence for cheating.

The fact that this announcement is, to put it generously, weird, and not how anyone in the real world actually implements tariffs points to this administration going with the formula from some undergrad econ textbook or econ blog they found and applying it to a spreadsheet. Maybe they got it from an LLM, maybe they didn't, but what matters is the college sophomore level of consideration of the individual tariffs.

>Maybe they got it from an LLM, maybe they didn't, but what matters is the college sophomore level of consideration of the individual tariffs.

Sounds like we're in agreement? The tariffs are badly thought out, but the "lol they got their policy from LLMs" accusation is entirely spurious.

I really don't think this is entirely spurious considering Elon has been using LLMs to fire government employees. It seems like a very plausible explanation for how they came up with this policy.

>Elon has been using LLMs to fire government employees

Source?

from Mike Johnson, about DOGE's use of algorithms: ""Elon has cracked the code. He is now inside the agencies. He’s created these algorithms that are constantly crawling through the data. And as he told me in his office, the data doesn’t lie. We’re going to be able to get the information."

My speculation, but it's very likely that one of those pieces of analyzed information was the 5 bullet points from each federal employee justifying their job.

https://www.nbcnews.com/politics/donald-trump/live-blog/trum...

https://www.nbcnews.com/politics/doge/federal-workers-agenci...

Is Elon really that stupid?

Just because doesn't lie doesn't mean we understand it correctly.

I think so, but I'm not actually sure that any of the realities that would have lead to this outcome is any less embarrassing for the administration. Because there is real information contained in the fact that pretty much every LLM gives roughly the same answer— it doesn't help that the wording is similar either. Does it mean that they used an LLM? No not really, but it does strongly imply that their formula and the source for it are the "StackOverflow answer" for lack of a better term.

And while it's not wrong I guess it's also the kind of thing that you would expect to use for your econ homework than a real application of policy. I don't think it's unreasonable to assume that the person who made this spreadsheet really did do something to the effect of type "how to calculate reciprocal tariffs" into Google. And that's not a bad thing necessarily if you're a rando who's just been tasked with figuring that out but you couldn't have found someone with more experience with actually doing this and modeling the economic effects?

I think that it's a testament to the genuinely breakneck speeds they been trying to get policy out there that it's been so slapdash.

Seems like it's at least a plausible possibility for Paul Krugman

https://paulkrugman.substack.com/p/will-careless-stupidity-k...

The takeaway from the X-poster being "it's good that they're using LLMs" is hilarious. Why praise the method if the end result is so bad?

Perfection is just a bugfix away /s

Reply on news.ycombinator.com