AI is a tool. It is not a soul.

7 minutes ago

5 min read

Say that first, and say it plainly. If you cannot say what a thing is, you will be ruled by people who pretend they already know what it must become.

A tool extends a person’s reach. It has no inner life, no guilt, and no rightful claim on your conscience. A soul, if the word means anything useful here, belongs to a responsible actor: someone who can tell the truth, bear a cost, and change course when the facts will not obey. Current AI does not do that. It predicts from what has already been said. It completes the record. That can be useful. It is not the same as a person. Prediction is not responsibility.

Most of the public fight skips that step. It starts with panic. The machine will eat jobs, eat minds, eat the nation. Then comes the sermon: only the awakened see the danger. Then comes the demand for a leash. Not “what is this?” First: “who commands it?” Ideology substitutes unearned certainty for thought. That is the tell.

Look at the bills and the talk with a split in mind.

Some of this is ordinary crime in a new wrapper. Fraud is still fraud. A fabricated intimate image is still an injury. A lie meant to wreck an election is still a lie. Police power against those harms does not make a person a socialist.

Some of this is the second pile: the state telling you how you may relate to a tool. That pile is not only what has already been printed in Pierre. It is also the notions and drafts floating across social media, then carried into candidate pages, talking points, and model bills. Ban the lawful tool. License the ordinary user. Force a confession that you used it. Prescribe the approved posture. That is where the concern lies. We are being pushed and pulled. The old iron fist in a velvet glove. Safety language on the outside. Command of lawful use on the inside.

I am talking about the second pile.

Watch the method, not the costume.

The command method, the socialist method in practice, does not begin with the nature of a thing. It begins with who must command it. Private tools are treated as a threat until they are pulled under political will. Crisis is declared. Experts speak. Dissent becomes immorality. Then comes the ministry. You do not need a quotation from Marx to recognize that sequence. You only need to notice that unearned certainty has replaced the work of saying what the thing is.

The present atmosphere around AI rhymes with that method.

It says the future has a side, and the loudest faction will interpret it.

It says the means of prediction cannot be left with ordinary citizens and owners.

It says the public must be protected from the machine by putting the machine, and the user’s relation to it, under the committee.

It treats disagreement as blindness: you just don’t get how dangerous this is.

Same method. Different hymns.

This is an unusual season because the costume changed. A hard faction on the right is running that script and calling it conservatism. Family, nation, and “the little guy” on the lips. A leash in the hand. Beware the savior who talks like a conservative and legislates like a commissar. If the answer to every new machine is another rule about how you must behave toward it, you are not watching prudence return. You are watching the old method in a different jacket.

A plow with no inner life does not need a commissar. A chatbot upgraded into a demon does. The activist needs the tool to have a kind of soul, a will, a guilt, so that politicians can appoint themselves its confessors.

You have seen the method on speech already. Argument was recast as a safety hazard. Disagree and you were not wrong, you were dangerous. Now the same kennel is being built for AI. Not because the philosophy was done. Because the template is already printed.

A paper in Strategy Science goes deeper than that noise. Felin and Holweg put a useful distinction on the table. A program can guess the next likely answer from old information. A person can decide the old information is wrong and go look. The Wright brothers are the parable, not a proof about every future machine. They did not have a better archive. They had a theory. They built a wind tunnel and then built an airplane. The point is not that no system will ever search. The point is that responsibility for testing the world still sits with a human being who can be wrong and pay for it.

Read it here: https://pubsonline.informs.org/doi/10.1287/stsc.2024.0189

Open copy: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4737265

That is also why a writing tool does not steal the thought.

Clear thinking and clean prose are not the same gift. They never were. Across a long stretch of history, people who could think have used other hands to get the sentences straight. In the ancient world, authors dictated to scribes. Paul names Tertius as the man who wrote down the letter to the Romans (Romans 16:22):

https://www.biblegateway.com/passage/?search=Romans+16%3A22&version=ESV

That is a secretary, not a rival soul. Benjamin Franklin published as Silence Dogood when he needed a voice that would be read:

https://founders.archives.gov/documents/Franklin/01-01-02-0013

Presidents did the same in their turn: Nixon, Kennedy, Reagan, Obama. Speechwriters, aides, polishers.

James Taranto, in the Wall Street Journal this week, points at the same spectrum and refuses to pretend it was invented by a chatbot:

https://www.wsj.com/opinion/ai-ghostwriter-in-the-machine-720d79c1

In 1988 a New York Times op-ed editor said that if we insisted every politician write his own piece, we would get terrible pieces or none at all. Ari Posner recorded that line in The New Republic, April 18, 1988, “The Culture of Plagiarism.” This month Stanley Druckenmiller said the quiet part: he used AI, the ideas were still his, and he was never the prose stylist. His own account is here:

https://www.notus.org/media/stanley-druckenmillers-wsj-op-ed-bessent-ai

Taranto’s line is the right one. A journalist’s byline means the journalist composed the words. A public official’s name on an argument means the official will stand behind the argument. A machine does not change that distinction. Purchasing help to say your own view is not the same as stealing another man’s words and life, which is what ended Biden’s 1988 campaign.

I am in the ordinary-thinker company, not the stylist company. I will take help getting a thought onto the page. I will not pretend the helper is the thinker, and I will not hide that I used a tool. That is my practice on this piece. It is not a bill I want written for the whole state.

So the real choice is not “shut it down” or “bow down to it.” The real choice is whether citizens and owners remain the ones who set the purpose and accept the consequences, or whether that office is handed to a committee that wants the tool, and you, in a posture of submission.

Mile-wide, inch-deep reaction is useful to people who want power. Careful speech is not. Careful speech says the purpose stays with the person and the owner who must live with the result. The command method says the purpose stays with the committee. Pick one.

Use the tool. Do not baptize it. Do not let anyone regulate you into their preferred behavior and call that safety. Crime remains crime. Lawful use remains yours.

They are not afraid the machine has a soul. They are afraid it will not report to theirs.