AI vs SI.
What’s in a name?

THE MOMENT

Why are these names being debated?

On September 19, 2026, Donald Trump proposed an online naming poll with three alternatives to artificial intelligence: Superior Intelligence, Extreme Intelligence and Supreme Intelligence. Anadolu reported the original proposal. On September 20, a follow-up narrowed the choices to Superior and Extreme, according to Just the News’s report.

That is the immediate context for this site. A popular name can change how people talk about a technology. It does not, by itself, change how a system works, establish a scientific category or demonstrate a new capability.

Superintelligence belongs in the comparison for a different reason. It is an existing idea in research about machines that could greatly exceed human cognitive abilities. It was not one of the three alternatives listed in that original poll.

We can take the naming question seriously while asking more of it. What does each word actually claim? What evidence would make the claim credible? And whose experience gets better if the technology succeeds?

01 / THE ESTABLISHED TERM

Artificial intelligence: the broad category

Artificial intelligence, or AI, is the established umbrella term. It covers different approaches and applications rather than one product, one model or one fixed level of ability.

The OECD’s definition describes systems that use inputs to infer outputs in pursuit of objectives. Those outputs may include forecasts, generated material, recommendations or decisions. The systems can differ in how independently they operate and whether they adapt after deployment. Its explanation includes both machine learning and approaches based on represented knowledge. Read the OECD.AI explanation.

In everyday language, “artificial” can sound like a criticism: an artificial smile, for example, suggests insincerity. In the name of this field, the word does not settle whether an output is useful, accurate or valuable. Those are separate questions about a particular system and its use.

What the label tells you

It places the technology in a broad family. It does not promise that the system is smarter than every person, understands every situation or deserves authority over a decision. “AI-powered” on a product page is the beginning of an explanation, not the result of an evaluation.

Ask: What task does this system perform, and how reliably does it perform it?

02 / OUR CHOSEN NAME

Superior intelligence: a comparison to earn

Superior means better in a comparison. That makes it an attractive name, but also an incomplete claim. Better than what? On which task? Under what conditions? At what cost?

In the naming debate, Superior Intelligence is a proposed replacement label. On this website, it is also our editorial standard: people and machines should earn the word through outcomes we can recognise. Neither use supplies a technical certification or an agreed threshold for intelligence.

A result can be superior without being universal

Imagine a tool that reduces the time needed to find a passage in a large archive. It could be better at that job than a particular manual process, while still being poor at interpreting an unfamiliar document. The comparison should stay attached to the task that was measured.

We would also distinguish the system’s performance from the service around it. A tool that completes an internal step faster may not shorten a resident’s total wait. A saving for an organisation may not become a saving for its customers. “Superior” becomes meaningful when we say who benefits and how.

Superior intelligence can be human. Our domain leaves room for a person’s judgment, a machine’s capability or what they achieve together. Human worth is not a benchmark; the comparison is about a specific task and result.

Ask: Better than which alternative, and better for whom?

03 / A NAME ABOUT DEGREE

Extreme intelligence: intensity needs a direction

Extreme suggests something pushed far beyond an ordinary level. In this debate, it is a proposed name rather than a technical stage with an agreed test. The spelling “Xtreme” gives the word a different visual style; it does not establish a separate kind of intelligence.

The word can communicate ambition, speed or unusual power. Its weakness is that it leaves the direction of improvement open. Extreme volume is not necessarily useful volume. Extreme speed is not necessarily careful work. A system can be impressive while solving a problem that is not your problem.

More is a measurement; better is a judgement

Consider a hypothetical service that generates a thousand responses in the time another produces ten. That difference matters only after you understand the task. If most responses need correcting, volume may increase the work people must do. If they are dependable and useful, scale may be a real benefit.

This is why we prefer the comparison invited by “superior.” It encourages a follow-up about the outcome. That is our interpretation of the names, not a scientific rule assigning one word to humans and another to machines. People and machines can both produce extreme results; either can perform better or worse on a particular task.

Ask: Extreme in which dimension, and does that help anyone?

04 / A NAME ABOUT RANK

Supreme intelligence: the strongest claim

Supreme suggests the highest position or authority. As a proposed technology label, it makes a broader rhetorical claim than superiority in one comparison. It sounds as though the ranking has already been settled.

But a product name does not define the field over which that ranking applies. Best at translating which language? Best at planning under which constraints? Best according to whose priorities? Without boundaries and evidence, the word gives the reader very little to examine.

Capability and authority are different questions

A system might contribute an excellent forecast without being entitled to choose the policy that follows. People can disagree about a goal even when they agree about the evidence. Deciding how to distribute a benefit, accommodate an exception or accept a trade-off requires more than announcing which tool has the highest score.

For this site, that makes “supreme” a less useful invitation to a public conversation. We want readers to question a claim, compare results and ask for corrections. A name that sounds final can make those questions feel settled before the discussion begins.

Ask: What justifies this ranking, and who remains answerable for the decision?

05 / THE RESEARCH CONCEPT

Superintelligence: a much broader capability claim

Superintelligence is usually written as one word. When the discussion concerns an artificial system, you may see “artificial superintelligence,” abbreviated ASI. This guide uses ASI where an abbreviation helps, rather than making “SI” carry several different meanings.

In Nick Bostrom’s account, the concept concerns an intellect that dramatically exceeds the strongest human abilities across a very wide range of fields. That breadth matters: excellence in one bounded task is not the same claim. Read Bostrom’s discussion of general superintelligence.

Used in this sense, superintelligence is not a more flattering spelling of “superior.” One term can describe an advantage in a specified comparison. The other raises a much larger question about the reach and level of cognitive capability.

What a convincing claim would need

We would want evidence across substantially different kinds of work, with clear comparisons, disclosed conditions and serious attempts to find weaknesses. A striking demonstration can motivate further investigation. It cannot answer every question about how the system behaves in unfamiliar settings.

The definition also does not, by itself, establish consciousness, kindness, infallibility or a right to govern. Bostrom explicitly discusses the possibility of motivations unlike human ones. Strong capability and beneficial use need to be examined separately.

Predictions about when such capabilities might emerge should identify their assumptions and uncertainties. This guide does not assign an arrival date or treat a renaming poll as evidence that a research milestone has been reached.

Ask: How broad is the claimed advantage, and what independent evidence supports it?

A RELATED TERM

Where does AGI fit?

Artificial general intelligence, or AGI, concerns general capability across tasks rather than strength in a single speciality. Definitions and thresholds vary, so a useful discussion begins by stating which definition is being used.

The research paper Levels of AGI proposes distinguishing performance from generality, and treating autonomy as an additional deployment consideration. It offers a framework for comparison, not a universally binding finish line. Read the research paper.

These distinctions help explain why the five names on this page should not be drawn as an upgrade sequence. A proposed name tells you how someone wants to present a technology. A capability claim needs its own definition and evidence. Permission to act is another design decision again.

A hypothetical assistant could be broadly useful but require approval before doing anything consequential. Another could act independently within a very small task. The second system’s freedom to act would not automatically make it more generally intelligent.

HUMAN VS MACHINE

Could the superior intelligence be us?

Yes—our view is that “superior” can describe human judgment. It could be the person who catches a false claim, understands the context or decides that an answer needs checking. Choosing this name leaves that possibility open.

We ask the same question of a person, a machine or a team using one: what did they do better? A particular result can favour the human without settling every other comparison.

Consider a hypothetical application-writing tool. One evaluation could ask whether its draft is grammatically polished. Another could ask whether the information is correct. A third could ask whether a person can understand, amend and stand behind what is submitted. Those evaluations concern different parts of a successful outcome.

We would also compare a person using the tool with a person working without it. The most useful question for a workplace or public service may be how people and systems perform together. That comparison leaves room for assistance, supervision and choices about where automation belongs.

What if the superior intelligence is us? Explore four documented cases of machine mistakes and the human actions that exposed or addressed them.

Responsibility still needs a home. If a resident is sent to the wrong service, “the AI said so” does not tell them how to get help. A practical improvement includes an understandable route to review and correction.

THE CANADIAN TEST

What would make it better here?

Canada’s June 2026 strategy announcement describes ambitions for health care, businesses and public services, alongside trust, opportunity and sovereignty. These are stated policy aims, not evidence that every proposed benefit has been delivered. Read the Government of Canada announcement.

For this site, a Canadian conversation should bring those ambitions down to the experience of using a service. The following are questions we would ask, not claims that a particular system has already solved them.

Those questions give our preference for “superior” a direction. We are interested in improvements that can be felt and checked, including by the people most likely to be overlooked.

A PRACTICAL READING GUIDE

Six questions behind any impressive name

Our editorial checklist below is a way to read a claim more carefully. It is not a certification scheme. Formal resources such as NIST’s AI Risk Management Framework also treat trustworthiness and risk as matters to address throughout a system’s development and use.

1. Name the task. “Better at intelligence” is hard to test. “Helps people locate the correct application requirement” gives you something to examine.
2. Name the comparison. Ask which existing service, workflow or group is being compared, and whether it had a fair chance under the same conditions.
3. Read beyond the average. Look for the difficult cases, the different groups affected and what happens when an answer is confidently wrong.
4. Count the whole cost. Consider checking, correction, training and waiting time as well as the apparent speed of the first output.
5. Find the person responsible. An organisation should be able to explain who can investigate a problem and how an affected person reaches them.
6. Check the benefit later. A launch demonstration and everyday use answer different questions. Ask whether the improvement lasts once the novelty wears off.

The strongest answer may be a modest, well-supported improvement. A smaller claim that survives examination is more useful to a person making a decision than a grand name with no clear way to test it.

WHY THIS DOMAIN

Superior is our choice. Evidence is the condition.

We chose superior-intelligence.ca because the word invites a comparison we think is worth having. We want the conversation to stay close to people’s expectations: better service, more opportunity, more time and a meaningful say in decisions that affect them.

That is an editorial preference for a name, not a claim that “AI” has been scientifically replaced. Artificial intelligence remains useful vocabulary for discussing the field. Superintelligence remains a separate capability concept. Extreme and supreme remain names readers can examine and debate.

Our choice becomes useful when it leads to a specific question: if this is supposed to be superior, what should change in your life? The answer may be different for a nurse, a student, a small-business owner or someone trying to reach a public service.

COMMON QUESTIONS

The short answers

READ FURTHER

Sources and editorial approach

Technical definitions are grounded in the primary resources below. The interpretation of the proposed names, the practical examples and our preference for “superior” are this site’s editorial perspective. Examples are hypothetical unless identified otherwise.

1. OECD.AI · What is an AI system?Definition and explanation, March 2024.
2. Anadolu · The original naming pollReporting, September 19, 2026.
3. Just the News · The follow-up naming pollReporting, September 20, 2026.
4. Nick Bostrom · Ethical Issues in Advanced Artificial IntelligenceResearch discussion of general superintelligence.
5. Morris and colleagues · Levels of AGIResearch framework; paper revised September 2025.
6. Government of Canada · National AI strategy announcementPolicy aims, June 5, 2026.
7. NIST · AI Risk Management FrameworkPrimary resource on AI risk and trustworthiness.

Reviewed September 21, 2026. This is an independent publication, not a government or research-institution statement. Suggest a correction or another perspective.