There was a meeting between the top labs and the current administration, and what came out of it is a preference for calling the technology "Super Intelligence" rather than "Artificial Intelligence." SpaceX said it would rename its AI unit to SpaceXSI, and Musk's summary was "No more AI. SI is better." Some states, California among them, have said they plan to keep saying AI. So there is a desire from the top to move to SI, and the rest of the ecosystem is figuring out whether it follows.

This post is about the name itself. I am not going to get into anybody's motives or into who wants which label and why. I am interested in whether SI is a better word than AI, and in one thing the episode makes me think about, which is how much I value being able to pick my own words.

My high level take on the swap: it will not change much. SI, AI, ASI, AGI, these are all made up labels for a moving target. Today the label of the moment is superintelligence. Tomorrow there will be another level and another label for it. The technology does not change because the letters changed.

The Reason I Can Write This At All

Worth saying out loud: I get to sit here and have an opinion about what to call this thing because freedom of speech is a load bearing part of US law. That is not a small thing, and it is the reason a post like this exists at all.

It is also where my one hesitation comes from. I like persuasion. I am much less comfortable with anything that shades toward compelled speech. Asking me to say "super intelligence" where I would have said "artificial intelligence" is a small ask, and small asks are where that slope starts. Encourage the word, make the case for the word, ship products that use the word, those all work on me. "You must use this word" is a different mechanism, and I would rather we not practice it on vocabulary nobody has settled yet.

That is a point about process, not about the word. On the word itself I am mostly positive.

I Actually Like the SI Framing

"AI" has picked up a lot of weight that has nothing to do with the technology. It is the thing that takes jobs, the thing that generates slop, the thing people are told to be worried about. A large share of the public conversation is still stuck on whether any of this should be happening.

It is happening. It is here, it is staying, it is already useful, and it is going to get much better. At some point the productive move is to stop arguing about whether and start working on how. If a different word helps people get there faster, I am not going to object to the word. Changing the rhetoric around this technology strikes me as a reasonable thing to want.

We Are Not At Superintelligence Yet

The place the new name overshoots is accuracy. What we have right now is not superintelligence. It is artificial intelligence, and it is very good at a growing number of things. Coding in particular, when you pair the right model with the right framework and the right harness, is genuinely excellent and improving fast.

Will we get to something that earns the word "super"? Probably. Maybe in a month, maybe in a couple of months, maybe in a couple of years. I would not bet confidently on any of those. But we are not there today, and using the top level word for the current level is a hard sell to people who work with these systems daily and can see exactly where they still fall over.

AI Is a Term of Art, and Those Are Hard to Move

AI is a term of art. The whole point of a term of art is that everybody uses the same lingo to refer to the same thing, and the value of it is almost entirely in the fact that it is shared. It has been in use since the 1960s. Decades of papers, courses, job titles, org charts, product names, laws, and conference proceedings are built on those two letters. That is a lot of inertia, and inertia here is not stubbornness, it is just what shared vocabulary feels like from the inside.

That said, I do agree with one half of the complaint. AI is extremely general right now. In practice it means "something the computer does by itself," which covers a thermostat schedule, a recommendation ranker, a spam filter, and an agent that opens pull requests. When a word covers that much range it stops carrying information, and every conversation starts with ten minutes of figuring out which AI anybody is talking about.

So there is a real problem. I just do not think "call all of it SI" fully solves it, because SI is also one label for the same enormous range. You trade a vague word for a more flattering one.

Name the Levels, Not the Field

The useful distinction is not between AI and SI. It is between the levels, and those already have names that mostly work:

• Narrow AI. Good at one bounded task. Almost everything shipping today.
• AGI, artificial general intelligence. Competent across domains at roughly human level, including domains it was not specifically built for.
• ASI, artificial superintelligence. Beyond the best humans across essentially everything.

If I were naming these from scratch I would probably not have used the word "artificial" as the stem at all. "General AI" and "Super AI," or just "general intelligence" and "super intelligence" as levels, would have been clearer than smuggling the level into a three letter acronym and hoping people know which one you mean. Notice that the new term collapses this: it takes the label for the top level and applies it to the whole field, including the spam filter. That is the opposite of the precision that motivated the complaint in the first place.

The levels are also going to keep multiplying. Whatever gets built after ASI will need a name, and that name will be made up too. The lesson is that the field level label is the least interesting one. If you want to communicate, say what the system can actually do.

The One Good Argument for SI Is Linguistic

Here is the place where I think SI genuinely wins, and it has nothing to do with vibes.

"Super" is close to a universal morpheme. It is the same word, or nearly the same, across the Latin-derived languages and well beyond them. "Intelligence" is also Latin-rooted and survives translation with its shape intact. So "super intelligence" transposes unusually well. Not to every language, but to a lot of them.

Compare that with "artificial intelligence." In Spanish it is inteligencia artificial, so the acronym is IA, not AI. The adjective goes after the noun, and the initials flip. Every Spanish language document, every bilingual conversation, every search query has to carry both forms. But "super" always goes in front, so in Spanish it is superinteligencia, and the acronym stays SI. The word order problem disappears because the modifier is a prefix rather than a trailing adjective.

That is a small, real, practical advantage, and it is the argument I would make if I had to defend the change.

The Association I Cannot Unhear

"Super" also carries baggage, and not the serious kind. The first thing it reminded me of is Dragon Ball Super, where "Super" is literally the name of the season that comes after the original run. The second is "supers," the way superhuman beings get referred to collectively in comics.

The comic book association might be closer to accurate than I would like. We are heading into a period where people with these tools can do things that look superhuman relative to what the same person could do five years ago. Not because the person changed, but because the leverage did. If the word pulls in that direction, toward augmented humans rather than replacement machines, that is a more honest picture of what is happening right now than either "artificial" or "superintelligence" suggests on its own.

My Actual Concern Is Not the Name

If the superhuman framing is right, then the thing worth worrying about is not which two letters we use. It is distribution.

The whole value of this technology, to me, is that it raises the floor for what one person can do. That only matters at scale if the capability actually reaches the whole population instead of concentrating. If super intelligence ends up being something the top one percent has and everybody else reads about, we will have built the most powerful leverage in history and handed it to the people who already had leverage. If it gets disseminated broadly, a lot of people get to do things they previously could not, and humanity gets the benefit rather than a slice of humanity.

That is the part I care about. The name is a detail by comparison.

Where I Land

I like the SI framing more than I expected to. It is friendlier, it travels better across languages, and it points at a future worth being excited about. It is less precise than AI in the one way that matters, since it hands the top level label to the entire field while we are still clearly below that level. And I would rather the shift happen by persuasion than by requirement, because picking my own words is something I do not want to give up cheaply.

I will probably keep saying AI for now, mostly because everyone I work with says AI and the entire value of a term of art is that it is shared. I will not argue with anyone who says SI. And I will keep trying to be specific about the level, because that is the part that carries information, and keep paying attention to who actually gets to use the thing, because that is the part that matters.