Post

dan sinker
@dansinker.com

Guess where they got their weird trade deficit math from?

2.1K reposts7.1K likes

Farseer Del
@farseerdel.bsky.social

People with any significant knowledge of it all, already have been convinced, even the advocates. It's just to the AI Bros, being an abhorrent mess that will cause chaos is an intended feature, not a drawback.

Gilbert
@tinpottop.bsky.social

I did actually debate about including "technocrat". I fear these non-nerds have stolen the word and ruined it for a generation. Luckily, I trust the language art nerds to come up with a good replacement. STEAM > STEM

Jeff R.
@jeffr23.bsky.social

I wonder if the Afghanistan anomaly (doesn't use the formula) is an AI glitch like counting r's in 'strawberry'. The ending of the one China policy probably is.

aPizzaBagel
@apizzabagel.bsky.social

Close, but really the problem is they aren't intelligent or responsible enough to ask the right question. So more like: -Computer, what is the Ultimate Answer to the Economy, the Tariffs, and Everything? -42

Gordon C
@vfxgordon.bsky.social

I screwed up; bsky.app/profile/vfxg... I'd still lay money on them running this crap through an LLM and they certainly did not check their work, but the country/territory breakdown is due to ISO country/territory codes (which also govern internet TLDs).

Gordon C@vfxgordon.bsky.social

Deleting my posts about domains, because it does look like this dataset is the underlying source. It’s obvious that they didn’t bother studying the data in any kind of detail, however.

Gordon C
@vfxgordon.bsky.social

I guess. Still wish I'd made the right connection, though. I'm still confident that they did almost no actual research though and just pushed the us census data on international trade through an LLM. That opinion is only reinforced by their laughable "paper" justifying their methodology.

Sascha_Vykos2
@saschavykos.bsky.social

This is what happens when you let people who never had to earn a gd thing in their lives fail up too far. People. It is your mfing civic duty to get these people fired or arrested or whatever should have happened before before the end up in *any* position of power.

mawus.bsky.social
@mawus.bsky.social

" “Instead, for every country, they just took our trade deficit with that country and divided it by the country's exports to us. What extraordinary nonsense this is.” " - James Surowiecki, former financial columnist for The New Yorker posted on X. www.politico.eu/article/dona...

politico.euTrump’s tariff math is crazy, says ‘Wisdom of Crowds’ authorJames Surowiecki challenges the White House’s rationale for punishing Europe, China, India and Japan with reciprocal tariffs.
Benj
@splatterthought.bsky.social

We've come... some distance?... from the day when Man Lost invented an app that marked the end of potential family feuds over suggesting Man Lost pull over and ask for directions...

Xathian
@xathian.com

You know, maybe there is a razor thin silver lining to this. If this becomes the mainstream story on the tariffs, and these tariffs crash the global economy, it may end up being the thing that finally ends this AI bubble garbage once and for all. Our global anti-tech moment of enlightenment

Computational Cosmetologist
@dferrer.bsky.social

The singularity doomers were worried about “super intelligences” mastering persuasion so well they could make humans do anything, but it turns out a lot of people will do anything an AI says even it says “lol I’m so full of shit” first

OldCodgerOfAppalachia
@softshellthac0.bsky.social

I am not an economic genius. I don’t even day trade. But, to me, that second paragraph that says “Just so you know, this method ignores a whole bunch of shit that might be integral” seems kinda like a thing that would make a normal human go: “Hey. Maybe we should call the stonks people, instead.”

Bell
@smutsia-gigantea.bsky.social

"I think the fact they're using LLMs is good but they def need better prompt engineers I think." Translation: directing world-changing economic policy based on readings of tea leaves is fine, they just need better tea leaf readers.

geofflangdale.bsky.social
@geofflangdale.bsky.social

In the recent paper about AI whiffing Math Olympiad 2025 questions (in a shock development, AI doesn't do as well with questions that aren't in its training corpus) the authors mentioned that over-confidence in the wrong answer was a core AI trait. A core trait shared by AI boosters, too.

Ohio Todd
@lettucewrangler.bsky.social

Iceland is so funny. Where the fuck do you plan to buy aluminum from? Iceland has such a powerful advantage in aluminum smelting that they're basically the only place that it is done. Are we going back to the era of aluminum being a precious metal?

The Ordinary Australian
@guzzisti.bsky.social

clearly no-one put any rational thought into it, just run it through Grok!! Idiots, incompetent. Stupid because it will crash the US economy which is what Musk wants because then the oligarchs will move in & buy up Gov't property in a fire sale and start running the country as aprivate business!

SpacePrez
@spaceprez.bsky.social

Trump and his cohort are all so stupid and know nothing about foreign policy so they literally just asked chatGPT "how much should I tariff other countries" and then posted that as official US policy. Even though said chatGPT told them "please don't trust me with US policy Im just a stupid chatbot"

exceptcody.bsky.social
@exceptcody.bsky.social

Ah yeah that's not great. Do we actually know that they did that or are the responses from AI just """coincidentally"""" identical to theirs? Either way I wouldn't think that the exact response from an AI is the complete right answer. I mean it.might be close but it wouldn't get everything.

Kevin F. Quinn
@kevfquinn.bsky.social

I would assume the AIs are regurgitating/synthesizing from something like an academic textbook on the topic. Unfortunate that these things don't cite their sources! Curious to find the source(s) to see what they concluded would be the impact, or the context of the examples.

Scott
@smosc.bsky.social

They built a flimsy formula to back into a flimsy solution that fits their flimsy narrative why to do something so historically flawed

Josué
@nerdery.bsky.social

Did Grok give any qualifications like the other three. Or would a reasonable assumption based on this data be that Elon basically said "Use my AI?"

Josué
@nerdery.bsky.social

I checked Grok. Asking the same question in those screenshots, it does yield similar warnings. I don't know what outcome I was hoping for. The Elon and Trump went with the "biased" AI that conveniently left out info. Or that regardless of which AI they referenced, they actively ignored warnings.

Lava monster
@lavamonster.bsky.social

These goobers keep going on about how the Singularity will bring about an AI that will solve all our problems, but even if a genius AI provided them with a foolproof roadmap to implementing a Star Trek utopia they'd likely go "nuh uh, that's sounds like commie shit" and burn the world anyway.

Bornach
@bornach.bsky.social

Not that far from the truth. An early version Copilot (called BingChat back then) when asked to write stories about billionaires would more often than not it cast them as villains of the story getting their comeuppance. The AI tech companies have since realigned their LLMs to never do this.

DrRobCz
@drrobcz.bsky.social

None of those explanations even makes any logical sense. Unclear why the deficit proportion would reduce the imbalance. Rather, it collects an equivalent to the imbalance into government coffers. However, it assumes ceteris paribus.

steamfuchs.bsky.social
@steamfuchs.bsky.social

It won't solve the "human problem", but it will stop it from pointing at the internet and saying 'How can we be wrong! The internet chatbots never lie!' And getting a bunch of brownie points from their base in the process.

Iain
@iainfrew2024.bsky.social

His 'huffy bike' should be a pink tricycle with a white basket with flowers on it and rainbow streamers on the handles, I doubt he ever learned to ride a bike and wouldn't have the energy to keep one upright even without the paddle... #TraitorTrump

Funk Dr
@funkdr.bsky.social

This is so stupid. Thanks for doing this. I’m really afraid that megalomaniacs with money are disastrously failing to understand how generative ai works in any way that isn’t a disaster. it basically is very agreeable and will try and make whatever cocksure nonsense work.. they take it as gospel

glass-possum.bsky.social
@glass-possum.bsky.social

I don’t know how quickly AI incorporates training data, but this could be a chicken and egg situation. A lot of articles just came out talking about trump’s trade deficit plan and calculation. It’s possible that the AI models incorporated those articles already, and will give what they read.

Tahoma
@tahoma28.bsky.social

Sad part are the people around him that came up with this and the embarrassing part is a tariff on an island that has zero population and only penguins live there.

curiousgreg.bsky.social
@curiousgreg.bsky.social

Given that you accept that math, you still have to calculate the combined GDP of the countries you’re enacting tariffs on and the economic power behind the alliances you’ll form by threatening their supply and distribution chains en masse. That’s the arsenal you unleash once you start a trade war.

Cody K.
@codykinski.bsky.social

Yikes if this is real, I think they should probably slow down a little before completing the big dumb switch to A.I. as the Big Decision Maker for the whole government. It sure ain't ready for prime time.

Jeff Nicholson
@vitamaltz.bsky.social

Everybody said AI was going to change the world sooner than we expect. I did not expect it to destroy the world economy in early 2025 by misunderstanding tariffs and making a mathematical error.

Lennart Regebro
@regebro.bsky.social

And it's absolutely incorrect. It assumes that the two trading countries make exactly the same products, and the only difference is price. It would work for electricity, but pretty much nothing else.