President Lee Jae Myung recently took to social media to rebut critics of a new practice: having the country's top investigators and prosecutors sit in on his Cabinet meetings. His evidence was an exchange with Gemini, Google's generative artificial intelligence.
According to the post, the chatbot labeled the objection as sophistry that “flatly denies” the president's constitutional authority to supervise the executive branch. It went further, warning that barring investigators from the room would leave them unaccountable. As far as I know, no head of state has ever cited a chatbot to defend a government decision.
Anyone who uses these tools knows how much a language model's answer depends on how a question is framed. Change the premise, switch the model or ask again at a different hour, and you can draw the opposite conclusion.
Regular users understand this and rarely treat a reply as settled truth. That is worth remembering, because Korea already has ministries that interpret statutes and a sizable academic field of constitutional scholars. Yet the way the president framed the exchange suggests that neither appears to have been consulted first, though I doubt this reflects how the decision was actually reached.
The president later explained that he had posted the exchange to show the country how logically and factually shaky the opposing case was.
Perhaps he meant it. But when constitutional interpretation and political prudence are braided together, brandishing a chatbot's answer to end a debate is a misstep. It makes the work of governing look trivial and invites ridicule. And ridicule is precisely what followed.
The opposition launched an “I asked AI, too” challenge, and the president's method became the joke of the town rather than the policy under debate. Calling his critics' views sophistry did not help either. A serious argument deserves a serious reply, built on statutory interpretation, verified facts and a fair rendering of the other side's strongest case.
None of this makes AI the culprit. It is genuinely useful for summarizing piles of documents and locating relevant clauses in seconds. Yet a language model reflects the assumptions embedded in a prompt, and because these systems are tuned to satisfy the person typing, they tend to agree unless explicitly pushed to object.
A plausible answer is not a justified one. It cannot confer legitimacy on a political decision or assume the accountability that decision carries. In a democracy, legitimacy should come from reasons the public can examine and from officials who can be voted out.
Whether investigative chiefs may attend Cabinet meetings is one thing. Whether they should is another.
Their presence alongside the president will inevitably cast doubt on the impartiality of the cases their offices pursue. Yet something can be lawful and still be injudicious. With approval ratings drifting downward, the episode reads to many voters as impatient sarcasm rather than thoughtful argument.
Anyone who speaks in public can learn from the president's mistake. While we are at it, here is my own list of things thoughtful people should stop saying in the age of AI.
“I asked AI, and it says so.” People say this to borrow authority they have not earned and to hand off responsibility for their own judgment. The result is a claim nobody has verified and nobody is willing to stand behind. The lesson: A source you cannot cross-examine is not evidence.
“Why should I bother? AI does it all.” Here, the speaker wants to skip the reading, the writing and the arguing. Over time, the ability and agency to judge for yourself wither from disuse, much like a language you no longer speak.
“AI wrote it, but these are basically my thoughts.” This one is more nuanced than it sounds. The temptation is to dress up half-digested ideas in fluent prose and pass them off as reflection. Eventually, somebody asks a follow-up question, and the person cannot defend their own thinking. The problem is not simply an inability to explain the words, but a failure to take intellectual ownership of the ideas.
“AI made a mistake, so it is useless.” Whoever says this throws out the whole tool over a single error and never asks why they failed to catch it themselves. Checking for errors is your job.
In a similar vein, you should not say, “Who cares how, as long as I get the answer?” This treats results as the only thing that counts. But questioning, testing and failing are how your mind develops, and a student who skips that work never learns where an answer comes from.
“If AI does it better, that job should disappear.” I hear this often now from my students. It overlooks the fact that human experts built these systems in the first place and still produce the knowledge they run on. A field emptied of people will eventually fail to generate new ideas and begin repeating itself. Humanity's intellectual life will wither if people stop creating knowledge.
Read together, the list says less about machines than about us. Each phrase hands something over to AI that belongs to people: agency, dignity, effort, honesty and accountability. A society that develops this habit will eventually find itself unable to argue, invent or govern on its own.
Institutions have a part to play as well. Schools, newsrooms, courts and legislatures should be explicit about where machine output stops and human judgment begins, not out of technophobia, but out of basic professional responsibility. A judge who cites a statute should know the text. A reporter who quotes a summary should have read the document.
Scientific progress, new knowledge and the long, untidy advance of human culture have always rested on people who read closely, ask uncomfortable questions, weigh evidence and accept responsibility for their imaginations, hypotheses and conclusions.
The engineers who will build tomorrow's technologies need those same old-fashioned habits, and they need them more, not less, as the tools grow stronger and smarter. AI should extend those habits, not replace them.
The next time a hard question comes up, the president included, “AI says …” is the last thing any of us should start with.
- - -
Lim Woong
Lim Woong is a professor at the Graduate School of Education at Yonsei University in Seoul. The views expressed here are the writer’s own. — Ed.
khnews@heraldcorp.com
