From AI to SI

On 29 September, an executive order told US federal agencies to stop using "artificial intelligence" and "AI" in official communications, and to write "super intelligence" and "SI" instead. It goes further than a style guide. Agencies are instructed not to acknowledge the older term in any applicable setting.

Trump had said it the day before, at the United Nations, and gave the reason plainly: "The use of the word artificial makes intelligence sound fake."

I keep coming back to that sentence. A word became policy in a single afternoon, and almost nobody had to agree to anything.

Why a rename works

Adoption rarely moves through policy. Policy needs budget, enforcement, and someone willing to spend political capital on something unproven. A word needs repetition, and it costs nothing. Rename the thing and you reset the terms of every conversation about it, including the conversations inside the institutions that will decide what to do with it.

"Artificial" carries a suspicion. It means a copy, a simulation, something standing in for the real thing. "Super" carries an arrival. It means above, beyond, an upgrade. The systems themselves did not change between those two words. What changed is what people are allowed to feel about them. One word invites the question of whether we should trust it. The other moves past the question.

This is the macro version of something I have watched happen at much smaller scale. Adoption begins with one sponsor with appetite, and survives only where culture learns to carry it. A rename is a culture move made from the top. It is the cheapest infrastructure an institution has, and it keeps working after the sponsor leaves.

The vocabulary of the warning became the vocabulary of the mandate

The order landed the same day as a voluntary accord signed by leaders from Anthropic, OpenAI, Google, Meta and xAI, pledging safety audits, internal monitoring and external assessment. The accord has no enforcement mechanism. So the day produced two instruments: one binding in language, one voluntary in behaviour. The binding one was the word.

Two days later, the United States and fifteen other countries endorsed the Kyoto Vision for a Golden Age of Science, led by Michael Kratsios ahead of Japan's STS forum. No new funding was attached. The White House framed it in the same vocabulary: "Super intelligence applied to science will be the greatest force in history for democratizing research."

There is a detail here I cannot stop noticing. Superintelligence is the title of Nick Bostrom's 2014 book about existential risk. On 3 August 2014, Elon Musk recommended it: "Worth reading Superintelligence by Bostrom. We need to be super careful with AI. Potentially more dangerous than nukes." Twelve years later, that word names the era, and Musk's own company is following the naming, with SpaceXAI set to become SpaceXSI.

What I cannot resolve yet

The case for it: "artificial" carried a sneer, and a sneer is not an argument. It made a serious technology sound like a costume, and it made serious concern sound like a complaint about special effects. A plainer word might let people think about capability instead of defending a feeling.

The case against: "super" carries no information about the thing, and it closes a question that "artificial" kept open. A name that flatters is easier to adopt and harder to argue with. This particular name also arrives from the same branch of government that is signing the safety accords, so it carries institutional weight and no process behind it.

The shift is happening at two levels at once. At the macro level, governments and ministries now have a new word in their documents. At the micro level, individuals have a new word in their mouths. The micro level is the one that decides whether the technology becomes normal.

What would prove me wrong

I am wrong if, by the end of 2027, the term has stayed inside government documents and ordinary people still say AI. Then the rename was housekeeping, and the culture did not move with the paperwork.

The cheap version: count the word. Listen for "super intelligence" in a sentence that is not quoting the order. The first time you hear it used casually, in a meeting, in a group chat, on a sales call, the frame has moved. If you never hear it, I was wrong, and I will say so in the next quarterly review.

What it changes for the people holding the decision

Leaders inside institutions: your vocabulary will drift before your policy does. Vendor decks, internal comms and job descriptions will follow the federal language within a few quarters, mostly without anyone deciding to. Governance documents that still say AI will read as a year out of date by the middle of 2027.

Anyone selling into government or regulated sectors: the term in your documents becomes a marker of whether you are current. That is a cheap advantage and a cheap embarrassment.

Investors: renames are free and they travel fast, so a vocabulary change is often the earliest visible sign of what a market expects. Watch the words in pitch decks before the numbers.

The bridge, offered with a straight face

One more thing, offered with a straight face and a smile. Silicon is Si. Super Intelligence is SI. The letter between them is O, and that O is oxygen, the element that lets two silicons bond instead of sitting apart as sand. Read the new abbreviation honestly and it says the bridge holds only if there is an oxygenated human in the middle of it. My Tuvaluan name, Siosi, contains that bond. I did not arrange it, and I am not going to pretend I mind it.

What I am watching

I am tracking vocabulary as an adoption signal, the same way I watch sponsor behaviour. My own sentences have already changed, which is the honest reason I am writing this now rather than in six months.

If you have heard "super intelligence" used in an ordinary sentence, I want to know where, and whether it landed as a promise or a threat.