I saw that tweet last night and thought that if GPT, Gemini, Grok, Claude all gave the same answer it must be because of the training data. So I asked both GPT and Claude to give citations for the formula without using web search. Both cited Peter Navarro's work. So it's more likely the Trump team just poisoned the training data than that they got the formula from the LLM.
The US based LLMs are filled with garbage data since our tech-company geniuses decided to let their LLMs scoop up everything they find on the Internet. This is why the Chinese AIs are more efficient and faster. They compartmentalized their data for specific subject matter.
This would imply that if the only "answer" to this problem that the LLMs had access to comes from Navarro's work, his is also the only one who is looking into the problem.
This is "unique genius" territory. Navarro is the Terrence Howard of economics.
LLMs are neither smart nor dumb, they are tools, like hammers, that are either well or poorly designed, suited or unsuited to a task, and used by skilled or unskilled individuals. And, as a NYT columnist pointed out, they are so far a mid technology. It the hands of low skilled, sloppy people, the mid can help crash the global economy.
LLM’s aren’t dumb, they’re just lazy. The ultimate stopping point is determined by the user, they’re simply crafting complementary content aligned to that users point of view. I’m pretty sure there’s a feedback mechanism in many of the Close system LLMs that interpret the end of a conversation as a successful application of the technology. That feedback then seems to be fed back into the overall stream of data that the user is working with. The quality of the content has nothing to do with the accuracy but rather the user preference and engagement.
I agree. I refuse to believe anyone is this stupid. It has to be the work of some incomplete text output from one of those incompetent LLM implementations.
Although unconfirmed, LLMs were suspected of eventually causing significant issues due to their inherent flaws but they quickly went straight for humanity's current dominant empire's throat by merely existing. Both funny & horrifying at the same time. A few ignoramuses messed up a lot.
You missed the two other things that make it seem like an LLM:
- The country list includes questionable places (like islands with only penguins or only US bases) but all are country internet domains. Thats the sort of context miss that an LLM would make.
- The maths includes defining epsilon as below zero and then setting it to four. The sort of mathematical logic error that LLMs make.
"Mr. Kokotajlo, ....he said, sometime in early 2027, if current trends hold, A.I. will be a superhuman coder. Then, by mid-2027, it will be a superhuman A.I. researcher — an autonomous agent that can oversee teams of A.I. coders and make new discoveries. Then, in late 2027 or early 2028, it will become a superintelligent A.I. researcher — a machine intelligence that knows more than we do about building advanced A.I., and can automate its own research and development, essentially building smarter versions of itself. From there, he said, it’s a short hop to artificial superintelligence, or A.S.I., at which point all bets are off."
Additionally, can also address this: "But I think this kind of forecasting is worth doing, even if I disagree with some of the specific predictions. If powerful A.I. is really around the corner, we’re all going to need to start imagining some very strange futures."
It read like a paid PR promotion, with the occasional caveat that is phrased in such a way that it will be completely overlooked. I won't mind getting three emails from you today.
They are like doomsday preachers who predict the end of times (or the coming of the Mahdi, as someone wrote here). The possiblity that their prophecy could fail does not embarrass them, since they can always make another. And just like the doomsday preachers, one of them might be right at some point.
How him and Casey Newton got to run the NYT tech podcast is really something. They're as credulous as a YouTuber fanboy who puts "THIS IS HUGE!!!" on the thumbnail of every video.
MSNBC (I think it was Rachel) ran a story last night about how even sparsely-inhabted volcanic islands were included in the tariff list, suggesting that a bot was involved.
The fact that people generally, and politicians specifically, do not understand that tariffs are simply a tax, and one that broadly impacts consumers, and by that I mean citizens, is embarrassing to the country. The fact that Trump would say yesterday that the ‘people don’t pay the tariffs, but rather governments do’ shows a level of economic ignorance that I find infuriating!
I still wonder if Trump himself really believes that foreign governments will pay the tariffs, and not Americans. I am inclined to think, based on his known criminality, that he is lying and counting on the fact that his supporters are just too ignorant to realise this.
"Well, Mr. Trump, first you must crash the economy. Make millions of people unemployed."
"How do I do that?"
"Create a global trade war. Put arbitrary tariffs on every country's exports to the US. The more arbitrary, the better."
"That might crash the economy. But wouldn't the people blame it on me?"
"Not at all. Just say that all countries were taking advantage of the US and that some pain is to be expected."
"Right. Then what?"
"OK, so now the economy is in ruins. Millions are desperate. Next you claim you know a way to turn it all around, how to heal the economy. But you will need special executive powers to do so. An enabling act is needed. Once that has been passed, you are president for life. El Presidente."
"Sounds great. But will it work?"
"It worked for the other guy with the funny moustache."
At least the US has a surplus on Dumb. The biggest surplus.
We need to turn that into a trade surplus on dumb.
Like trade them all to El Salvador in exchange for Kilmar Abrego Garcia.
Sorry for the slightly OT.
I saw that tweet last night and thought that if GPT, Gemini, Grok, Claude all gave the same answer it must be because of the training data. So I asked both GPT and Claude to give citations for the formula without using web search. Both cited Peter Navarro's work. So it's more likely the Trump team just poisoned the training data than that they got the formula from the LLM.
The US based LLMs are filled with garbage data since our tech-company geniuses decided to let their LLMs scoop up everything they find on the Internet. This is why the Chinese AIs are more efficient and faster. They compartmentalized their data for specific subject matter.
This would imply that if the only "answer" to this problem that the LLMs had access to comes from Navarro's work, his is also the only one who is looking into the problem.
This is "unique genius" territory. Navarro is the Terrence Howard of economics.
Steps in shit, immediately walks on carpet.
The use of the offensive term "retarded" is unfortunate and detrimental to the larger importance of the point being made. Jus' sayin'.
1,000% agree. Can't give much credence to anyone not smart enough to see how offensive that is. Thank you for pointing that out.
Did the LLM say what happens to an economy that uses its formula?
The problem is, I refuse to believe an LLM is that stupid. This has to be the work of one of the incompetent DOGGIES Musk has working for him
LLMs are neither smart nor dumb, they are tools, like hammers, that are either well or poorly designed, suited or unsuited to a task, and used by skilled or unskilled individuals. And, as a NYT columnist pointed out, they are so far a mid technology. It the hands of low skilled, sloppy people, the mid can help crash the global economy.
LLM’s aren’t dumb, they’re just lazy. The ultimate stopping point is determined by the user, they’re simply crafting complementary content aligned to that users point of view. I’m pretty sure there’s a feedback mechanism in many of the Close system LLMs that interpret the end of a conversation as a successful application of the technology. That feedback then seems to be fed back into the overall stream of data that the user is working with. The quality of the content has nothing to do with the accuracy but rather the user preference and engagement.
LLMs are sycophants. If you say, "America needs tariffs, give me a tariff plan", they won't push back on the starting proposition, but roll with it.
I agree. I refuse to believe anyone is this stupid. It has to be the work of some incomplete text output from one of those incompetent LLM implementations.
Although unconfirmed, LLMs were suspected of eventually causing significant issues due to their inherent flaws but they quickly went straight for humanity's current dominant empire's throat by merely existing. Both funny & horrifying at the same time. A few ignoramuses messed up a lot.
God help us all.
You missed the two other things that make it seem like an LLM:
- The country list includes questionable places (like islands with only penguins or only US bases) but all are country internet domains. Thats the sort of context miss that an LLM would make.
- The maths includes defining epsilon as below zero and then setting it to four. The sort of mathematical logic error that LLMs make.
It’s a RELU function. His team knows linear algebra.
Please, please, please, Gary, give us your take on “AI 2027,” which Kevin Roose has written up in close to glowing terms in the NYT.
https://www.nytimes.com/2025/04/03/technology/ai-futures-project-ai-2027.html (Sorry, it's behind a paywall.)
"Mr. Kokotajlo, ....he said, sometime in early 2027, if current trends hold, A.I. will be a superhuman coder. Then, by mid-2027, it will be a superhuman A.I. researcher — an autonomous agent that can oversee teams of A.I. coders and make new discoveries. Then, in late 2027 or early 2028, it will become a superintelligent A.I. researcher — a machine intelligence that knows more than we do about building advanced A.I., and can automate its own research and development, essentially building smarter versions of itself. From there, he said, it’s a short hop to artificial superintelligence, or A.S.I., at which point all bets are off."
Additionally, can also address this: "But I think this kind of forecasting is worth doing, even if I disagree with some of the specific predictions. If powerful A.I. is really around the corner, we’re all going to need to start imagining some very strange futures."
It read like a paid PR promotion, with the occasional caveat that is phrased in such a way that it will be completely overlooked. I won't mind getting three emails from you today.
god i don’t think i can respond to his PR again
I understand the feeling. However, today's was worse than ever! I did leave a comment indicating my displeasure with the article.
Has nobody noticed that the time horizon keeps creeping further out?
This is as silly as Trump's tariff formula.
They are like doomsday preachers who predict the end of times (or the coming of the Mahdi, as someone wrote here). The possiblity that their prophecy could fail does not embarrass them, since they can always make another. And just like the doomsday preachers, one of them might be right at some point.
And they are making money while doing it.
How him and Casey Newton got to run the NYT tech podcast is really something. They're as credulous as a YouTuber fanboy who puts "THIS IS HUGE!!!" on the thumbnail of every video.
I doubt it. It's trivial to show any modern LLM is a lot smarter than Mr. Trump.
HOLY SHIT BALLS
MSNBC (I think it was Rachel) ran a story last night about how even sparsely-inhabted volcanic islands were included in the tariff list, suggesting that a bot was involved.
It's possible they included those places so that some shifty country wouldn't move their headquarters there to evade the tariff on their home.
Good point but I doubt they're smart enough for that kind of a chess move.
The fact that people generally, and politicians specifically, do not understand that tariffs are simply a tax, and one that broadly impacts consumers, and by that I mean citizens, is embarrassing to the country. The fact that Trump would say yesterday that the ‘people don’t pay the tariffs, but rather governments do’ shows a level of economic ignorance that I find infuriating!
I still wonder if Trump himself really believes that foreign governments will pay the tariffs, and not Americans. I am inclined to think, based on his known criminality, that he is lying and counting on the fact that his supporters are just too ignorant to realise this.
Lmao a country run by incompetent nepobabies who failed upwards
"ChatGPT, how do I become dictator for life?"
"Well, Mr. Trump, first you must crash the economy. Make millions of people unemployed."
"How do I do that?"
"Create a global trade war. Put arbitrary tariffs on every country's exports to the US. The more arbitrary, the better."
"That might crash the economy. But wouldn't the people blame it on me?"
"Not at all. Just say that all countries were taking advantage of the US and that some pain is to be expected."
"Right. Then what?"
"OK, so now the economy is in ruins. Millions are desperate. Next you claim you know a way to turn it all around, how to heal the economy. But you will need special executive powers to do so. An enabling act is needed. Once that has been passed, you are president for life. El Presidente."
"Sounds great. But will it work?"
"It worked for the other guy with the funny moustache."