What’s in a name?

A name can influence value before anyone examines the thing itself. It establishes an expectation, suggests a comparison and quietly frames the judgement that follows. In business, this has consequences for what people will buy, what investors will finance and what society will accept. The debate over what to call artificial intelligence belongs in this territory, alongside an apparently unrelated argument about diamonds.

Speaking at the United Nations on 22 September 2026, President Donald Trump announced that United States government documents would use ‘super intelligence’ instead of ‘artificial intelligence’. His explanation was that ‘artificial’ made the technology sound fake. Whatever one thinks of the proposed change, the underlying observation deserves consideration beyond politics: a technically defensible word can convey a misleading impression to the public.

‘Artificial’ describes something produced through human intervention. It need not imply that the result is ineffective or unreal. Yet in ordinary conversation, an artificial smile suggests insincerity and an artificial experience suggests a substitute for something more authentic. Those associations accompany the word into other settings. When attached to intelligence, it can invite the reader to discount the capability before considering the evidence.

The diamond industry offers a useful comparison. A laboratory-grown diamond is a diamond, with essentially the same chemical composition, crystal structure and physical properties as a natural diamond. Its origin differs, and specialist examination can identify that difference. ‘Synthetic’ is a legitimate technical description, but a consumer may hear something closer to ‘imitation’. The scientific category and the emotional response do not necessarily coincide.

‘Lab-grown’ approaches the same material through its production process. It tells the buyer where the diamond came from without suggesting that its brilliance is fraudulent. The word ‘grown’ may also feel more familiar and reassuring than ‘synthetic’. This is why a more descriptive name can make a product more acceptable without changing its substance. The language changes the terms on which the buyer considers it.

The commercial significance is substantial. A buyer may reasonably value a natural diamond’s geological history and rarity, while another prefers the price or characteristics of a laboratory-grown stone. Neither preference can be settled by chemistry alone. People purchase meaning alongside material. A name helps establish that meaning, although production costs, supply and personal attachment continue to influence value. Perception matters deeply without being the only thing that matters.

The comparison also reveals a limitation in the proposed renaming of AI. ‘Lab-grown’ describes origin; ‘super’ asserts a level of capability. Moving from one to the other therefore does more than remove an unfavourable association. It introduces a new claim. Unlike diamond, intelligence has no single chemical structure against which equivalence can be established. A more flattering adjective cannot resolve the question of what a machine understands or how reliably it performs.

For large language models, a more restrained description already exists. LLM stands for ‘large language model’. These are computational models trained on extensive datasets to learn patterns that support the processing and generation of language. People design their architectures, establish training objectives and make choices about the material and feedback used in their development. The term identifies a kind of system without declaring it either counterfeit or superior.

In this respect, ‘large language model’ performs a similar function to ‘lab-grown diamond’. It directs attention towards how something is constituted or produced. It is not a complete account of every capability, and LLMs are only one part of the broader field of AI. Nevertheless, for the systems to which it applies, it is a useful starting point for discussion without requiring agreement on a philosophical hierarchy between people and machines.

Perhaps ‘accumulated intelligence’ offers another way to think about them. Much of their capability draws on the recorded work of people: knowledge developed over generations and expressed through language, mathematics and code. The phrase acknowledges that inheritance. It makes the technology less like an intelligence arriving from elsewhere and more like a new means of drawing upon capacities that humanity has collectively developed.

That description needs its own qualification. A language model is not simply a warehouse of human knowledge. Training changes mathematical parameters, and the resulting system can generate new combinations and useful inferences, as well as confident errors. Its training material can also include machine-generated content. ‘Accumulated intelligence’ is therefore a philosophical interpretation, not a precise technical replacement. It illuminates the debt to human endeavour while leaving the mechanism only partly described.

‘Super’ carries a different ambiguity. Its Latin root includes the senses ‘above’ and ‘beyond’; contemporary usage also suggests something better or superior. Beyond human capacity in a particular task, however, does not establish superiority in every meaningful sense. In an influential technical definition, the philosopher Nick Bostrom describes superintelligence as an intellect that vastly exceeds human cognitive performance in virtually all domains of interest. That is a demanding claim, quite different from a more appealing name for all AI.

Bostrom’s concern extends to how such intelligence might emerge. A system approaching human-level capabilities and able to improve its own design could enter a positive feedback loop of recursive self-improvement: each improvement would strengthen its ability to produce the next. In the rapid-transition scenario he examines, this could trigger an ‘intelligence explosion’, carrying the system from roughly human-level ability to superintelligence within weeks, days or even hours. Such a transition could leave humanity with little time to understand what had happened, let alone respond. This is a possible trajectory, rather than an inevitable sequence or an established timetable.

Superintelligence therefore carries a double significance: exceptional capability and the risks that may accompany it. The capacity to exceed human performance raises the possibility of also exceeding our ability to supervise or control the system. This gives the proposed renaming an unexpected irony. A term chosen to inspire confidence already belongs to a philosophical discussion about human vulnerability. Replacing ‘artificial’ with ‘super’ may remove the suggestion of imitation while introducing the prospect of losing control. The reader ultimately determines which association carries greater weight.

Machines already perform particular tasks at speeds and scales unavailable to an individual person. They also fail in situations that people navigate with apparent ease. As their capabilities expand, the territory we regard as exclusively human may contract, although progress need not be uniform or continuous. None of this makes intelligence a simple ranking. Speed, judgement and understanding are different qualities. Greater competence in one does not automatically confer wisdom, responsibility or authority in another.

Perhaps, then, the deeper difficulty lies in ‘intelligence’ itself. We use the word to connect capabilities that may arise through very different processes. Calling them artificial directs attention to their origin. Calling them super directs attention to their supposed position relative to us. Both can be defensible in an appropriately defined context. Neither provides a complete, value-free description, and neither removes the obligation to demonstrate the claims it implies.

Objectivity remains possible in testing performance or establishing how a system was made. It becomes harder when a label is expected to settle what that system means for humanity. Readers bring their own convictions to the word. Commercial and political interests can reinforce those convictions, but even without them, one person’s promise of progress may be another’s suggestion of displacement. A debate ostensibly about technology can become a debate about the place people believe they should occupy.

For business, the discipline is to keep description close to evidence. A name should help a customer understand what is being offered and what expectations are justified. ‘Large language model’ is useful precisely because it leaves room for evaluation. ‘Accumulated intelligence’ may enrich the philosophical conversation. ‘Superintelligence’ should invite scrutiny of the claimed superiority. Acceptance achieved through clarity has a firmer basis than acceptance dependent on the emotional force of an adjective.

The question will become more difficult if technology eventually produces entities whose behaviour, autonomy or experience makes us reconsider the boundary around the human. We may then debate ‘artificial humans’ and ‘superhumans’. One expression could encourage us to withhold recognition; the other could encourage us to assign a higher status. Neither would answer whether such entities possess consciousness, deserve rights or should exercise power. Those questions would require arguments far more substantial than their names.

A diamond does not change when we rename it. A computational system does not become wiser because we call it super. But people may change what they are willing to pay, trust or recognise. What is in a name is, to a considerable extent, the relationship we are preparing to have with what comes next.