Did ChatGPT come up with Trump’s tariff rate formula? AI chatbots ChatGPT, Gemini, Claude and Grok all return the same formula for reciprocal tariff calculations, several X users claim.
-
Thanks, much appreciated.
-
Artificial Generalized Incompetence
Probably one of Musk's little goons was given the task, and they immediately went to ChatGPT.
-
And that's not lettuce, it's horseshit.
-
The numbers come from an overly simple way to level out trade deficits.
So if I sell you $100 in goods and you sell me $120 dollars in goods, I'm "losing" money, therefore 20% tariff (tax to sell me something). In reality, you're going to increase your prices and sell me $140 worth of the same stuff.
All the AIs did was expand this to a global scale, what's insane to me is that the math adds up. It doesn't take an AI to do this though, some economics undergrad could come up with the same thing. Understanding the underlying methodology shows how it completely lacks nuance or understanding of how the world really works.
what’s insane to me is that the math adds up
Too bad it's based on wrong assumptions. It's not the arithmetic that's the issue, it's the model.
some economics undergrad could come up with the same thing
And if they did it on a test, they'd flunk it.
Understanding the underlying methodology shows how it completely lacks nuance or understanding of how the world really works.
Yeah, it fails to understand the rationale for comparative advantage (there's a reason Ecuador exports more bananas than Norway does), and it also fails to consider the balance-of-payments effect of things like foreign direct investment (which looks zero-sum when it first takes place but means the profits are outflows from that point on, unless the foreign investors choose to reinvest them).
Also I don't think the idiots who came up with that table know the difference between a current account balance and balance of trade.
-
Isn't his weird formula the trade defecit percentage + Tariffs from that country divided by two?
Yes, it really is that stupid.
-
what if they all come up with that because it has been publicised
Then I'd ask who published and where they got their analysis from. Very possible that we've got an AI that's built up a backlog of Harvard Business Studies and CalTech economics models to reach the ideal hypothetical tariff regime. But it's just as likely they're ingesting 4chan reposts of Ron Paul Newsletters and Michael Savage radio transcripts to build up its economic background.
That's sort of the problem with AI. There's no specialist-driven guidance on what data is valuable and what data is crap. No litmus test to separate fact from fiction or serious discussion versus trolling. And these western developed models, in particular, are very bad about including the origins of their graphed logical output (because that would make the process of hashing and graphing more expensive, in a system that's already inelegant and resource intensive).
I just glanced at it and wouldnt know how something like that is even supposed to be, so I dont really know how unhinged the tariff rate thing is.
The problem is less that we don't know how bad the tariff rate is and more that the people designing the policies don't know either. They're fishing for answers in the answer pond, and they don't even know if they've got a fish or a boot at the end of the line.
Very possible that we’ve got an AI that’s built up a backlog of Harvard Business Studies and CalTech economics models to reach the ideal hypothetical tariff regime.
Possible but vastly improbable.
-
Yea but we can all agree on sky color but the numbers Trump posted are questionable at best
the numbers Trump posted are questionable at best
I'm less diplomatic: the numbers that Trump posted are flagrant bullshit.
-
They are. They record the data, stealing it. They search it, and reprint it (in whole or in part) upon request.
They search the data-space or what they're trained on (our content, the content of human beings), and reproduce statistically defined elements of it.
They're search engines that have stolen what they're trained on, and reproduce it as "results".
Searching and reproducing content they've already recorded, is absolutely part of what they are.
They are.
Their input sides are based on crawling, just as search is.
-
They are.
Their input sides are based on crawling, just as search is.
Yeah, and then they convert that to weighted probabilities or a "data space" they then search during content generation.
-
Very possible that we’ve got an AI that’s built up a backlog of Harvard Business Studies and CalTech economics models to reach the ideal hypothetical tariff regime.
Possible but vastly improbable.
And since when has CalTech been into econometric modeling?
Ranked 14th highest econ department in the country. Math modeling and economics are joined at the hip.
-
System shared this topic
