An Open Letter to President Trump: Do Not Place Blind Trust in ChatGPT
Dear Mr. Trump,
When you first emerged as President, I wrote on this platform and sent out to the world, with the utmost respect, that you were "the greatest trickster in history," in response to your extremely concise and insightful remark: "Why doesn't Japan arm itself with nuclear weapons?"
I felt both anger and contempt when I heard Biden—the epitome of cunning—dismiss your remark by saying, "That guy is completely ignorant; he doesn’t even know that we made sure Japan couldn’t develop nuclear weapons."
In the monthly magazine WiLL, released on April 25, Mr. Mutsuo Mabuchi wrote an article that, in my view, hits the mark precisely.
Initially, I thought about immediately introducing it to the world.
However, after reading an article published in the same day's issue of Hanada magazine—Paul Krugman's "The Cost of Trump Tariffs"—I decided to hold back.
The reason is simple: my gut told me that Krugman's points were correct.
(Introduction omitted)
"You might be surprised," Krugman writes, "but it is said that during the Trump administration, when they didn't know what figures to present about tariffs, they asked ChatGPT: 'Please find the academic paper that supports this.'
ChatGPT then fabricated data by pulling fake figures from real papers.
The original authors declared that the numbers were fabricated.
As you may know, ChatGPT often 'makes up' information.
The numbers appeared plausible at first glance but were, in fact, sloppy.
Some people claim they used formulas, but even if formulas were used, they were simplistic.
Peter Navarro, Trump's senior adviser on trade and manufacturing, said they included non-tariff barriers in the calculation, but how they converted those into a formula remains unclear.
Even if a formula was used, even if ChatGPT provided answers, the data was still unreliable.
—So, does this mean that Trump simply wanted high tariff numbers?
Krugman: Yes, that’s correct.
However, since he is a negotiator, if the U.S. could negotiate to lower tariffs imposed on American imports or eliminate non-tariff barriers, there would still be room for progress.
This is not the end of the story."
(Conclusion omitted)
When I watched Akari Iiyama’s YouTube program, where she said, "ChatGPT is quite competent and useful," I was intrigued and decided to give it a try myself.
I asked it about the results of the LPGA tournament.
However—to my astonishment—ChatGPT lied to me.
Unable to find the actual results I asked about, it fabricated an answer, presenting the result of a tournament won by Nasa Hataoka one year earlier as if it were current.
When I shared this experience with an extremely intelligent acquaintance, who is highly skilled in both IT and analytical thinking, he responded casually:
"AI lies. Search for 'hallucination' and you'll find out."
As a leading scholar in his field, he made it clear that AI is entirely unusable for genuine research or intellectual inquiry—and that he does not even regard AI as something worthy of serious consideration.
The scene where you, Mr. Trump, held a press conference shortly before your inauguration—standing alongside Masayoshi Son, the founder of OpenAI, and other AI industry leaders—must still be vivid in the memory of people around the world.
It was there that massive investments in AI industries within the United States were announced.
From that event, it is evident that you harbor very little skepticism toward AI.
At least, the level of skepticism my acquaintance has toward AI clearly does not exist in your mind.
On the contrary, it seems very likely that you might even place blind trust in it.
More troublingly, I recall the manner in which you once accused Japan of imposing a 700% tariff on U.S. automobile imports—a wildly inaccurate figure—and your mistaken criticism of Japan’s vehicle inspection system.
Reading Krugman's article, I immediately realized he was correct.
Because it mirrored exactly the kind of falsehoods I myself had experienced through ChatGPT.
Mr. Trump,
You must not place blind trust in ChatGPT.
Your keen insight—your historic decision and resolve that China, and China alone, poses the fundamental threat to freedom, democracy, and human intelligence itself—must not be undermined.
We cannot allow China to become any more arrogant.
We cannot allow them to continue stealing intellectual property at will.
Your stance on China remains a monumental act of statesmanship for the world.
However, blind faith in ChatGPT will only damage your reputation.
You must immediately verify the lies generated by ChatGPT.
Even when negotiating tariffs, it is, of course, essential to do so based on accurate numbers.
After reading Krugman's article, I immediately asked ChatGPT whether the story was true.
Here is what ChatGPT itself answered:
"ChatGPT does sometimes fabricate information... this is called hallucination. Therefore, it is recommended to verify information by reading original sources."
In other words, even ChatGPT admits that it should not be trusted 100%.
