The Machine Word Problem

AI, super intelligence, artificial—none of the labels fit. The naming war over what we call these tools is also a fight over what we expect from them.

By The ai-compute Desk

A few weeks ago, President Trump decreed by executive order that the technology is called Super Intelligence, while also correcting reporters who ask him about AI, telling them it's now SI. [2] The announcement landed with the familiar eye-rolls that follow most of his rebranding efforts. But buried beneath the spectacle is a question that matters: what do we actually call these tools, and does it change how we use them?

The case against 'artificial' is straightforward. [2] Apply that logic consistently and, under duress, you might buy a car described as artificial transportation—but you would not feel good about it. The label poisons the well before the product reaches the tap.

But the proposed replacement is worse. 'The idea of "super intelligence" floating around in various clouds,' Sharp writes, 'is actually less comforting to me than a version of knowledge output that's branded as synthetic.' [2] The word confers authority these systems have not yet earned. [2]

Names are never neutral. Every label carries a theory of value, a set of implied promises, and a political economy of expectation. Choose the wrong one and you don't just mislead the public—you distort the incentives of every institution that builds around the thing being named.

The technical reality is more prosaic than either label suggests. Models are trained to be evidenced-based and risk averse, and they will rigorously reflect contemporary facts, concerns and values as elite cultural institutions understand them today, not as society may understand them tomorrow. [2] That conservatism is a design feature, not a bug—but it is also a ceiling. Reproducing familiar arguments with lots of evidentiary support is a reliable path to getting a good grade, whereas offering an original insight that doesn't align with our intellectual consensus is likely to be graded as a mistake. [2]

The practical upshot: ask AI to explain most areas of modern life, and what you'll receive is closer to midwit intelligence. [2] Its answers will sound smart to dumb people, and sound like derivative and boring pablum to any subject matter expert who happens to have mastered the substance at hand. [2] That is a damning summary—and an accurate one. It is also, notably, not a summary of something that deserves the prefix 'super.'

None of this is to say the technology is without consequence. The argument is about precision—about matching the words we use to the capabilities we actually have, so the decisions we make on top of them are not built on borrowed confidence.

What these systems genuinely offer is something more specific: stamina. [2] LLMs can instantly perform a week's worth of shockingly comprehensive work that might otherwise require multiple research assistants. [2] That is real, and it is valuable. It is not omniscience.

Intellectual stamina is of course an important component of intelligence, but it's not synonymous with the sort of omniscience and omnipotence that's implied by super intelligence. [2] Many of the most consequential figures in history had specific experiences, talents, obsessions, and theories that developed for decades, at odds with consensus, and they arrived at the right time to reshape their fields, societies, and worlds. [2]

Sharp's proposed alternative is 'machine intelligence.' The logic is tidy: we already have machine learning—that's what training is—and machine intelligence is the output of that process. [2] Machines can be controlled, machines can be refined, and when machines malfunction and do lasting harm, machine makers can be held liable. They are tools with a scope.

The business stakes attached to naming are not abstract. [1] If agents are built on a substrate the public trusts, adoption is fast. If they are built on a substrate the public fears or dismisses, adoption stalls—and the entire value chain downstream stalls with it. [1]

[1] Confusing branding compounds confusing naming. When neither the product category nor the specific product is legible to a general audience, the default reaction is distrust, not adoption. As Thompson noted, there is more vision here than it might seem. [1]

There is a pattern here that recurs whenever a genuinely new capability class arrives in the market. The technology moves faster than the vocabulary does, and the gap gets filled by whoever shouts loudest—usually the people with the most to gain from a particular framing. The rest of us pay for that in confusion, in misallocated investment, and in policy built on metaphors rather than mechanics.

[2] Boring, though, is not disqualifying. In any event, our choice of language will matter. [2] Spend the next 20 years selling intelligence that's 'artificial' and these products could have a difficult time building trust. Call it super intelligence and we are conferring too much authority on these tools, and likely scaring the crap out of everyone along the way. [2]

The naming fight over AI is not a semantic parlor game. It is a slow-motion negotiation over what these systems are for, who controls them, and who bears the cost when they fail. 'Machine intelligence' may be a boring label, but boring and accurate is a defensible place to stand—especially when the alternative is a brandscape built on science fiction and trillion-dollar self-interest. [2]

Its gift is stamina, though—not brilliance.Andrew Sharp