What Is Super Intelligence? SI vs AI, and Why Trump Renamed AI

Superintelligence is the top rung of a 3-step ladder that AI researchers have used for decades. Narrow AI does 1 kind of task well. General AI, or AGI, would match a capable adult across most intellectual work. Superintelligence, often abbreviated ASI, would leave every human expert behind in nearly every field at once.

The standard definition comes from the philosopher Nick Bostrom, whose 2014 book Superintelligence: Paths, Dangers, Strategies defines it as “any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest.” 2 words in that sentence carry the weight: greatly and all. A system that beats grandmasters at chess is narrow AI. A system that beats every grandmaster, every surgeon, every lawyer and every engineer, and then redesigns itself to do it faster, is the thing the word was coined for.

The word is older than the technology. Merriam-Webster dates the first known use of superintelligence to around 1822 and defines it as “an entity that surpasses humans in overall intelligence or in some particular measure of intelligence.” The dictionary updated the entry on September 16, 2026, 6 days before the UN speech, with example sentences about a congressional bill to ban the thing. That is the state of the vocabulary this week: 1 word, 200 years old, claimed by a president as a synonym for AI and by lawmakers as a technology to prohibit.

AI is the category. Superintelligence is a capability level inside it that has not been reached. So under the technical meaning, SI versus AI is not a contest between 2 technologies; it is the difference between a family and its most powerful hypothetical member. Every assistant you can open today, ChatGPT, Gemini, Claude, Copilot, is narrow AI with a wide vocabulary: impressive within a conversation, unable to set its own goals, and still dependent on a human to judge whether the answer is right.

The rename collapses that distinction. When a government document says super intelligence, it will mean the spam filter, the chatbot, the fraud model and the hypothetical future system in the same breath. That is the cost the research community objects to: the field loses its word for the one thing it worries about most. It is also why you will keep seeing both usages side by side, and why this page keeps them apart.

A practical test cuts through the confusion. Ask whether the system can take a goal it was never trained on, learn what it needs, and outperform the best human at it. If the answer is no, you are looking at AI, whatever the document calls it. If the answer were ever yes across the board, you would not need an announcement to notice.

The rename was not a single line in a speech. It was a 10-day arc that began as a dismissal of AI safety warnings and ended at the General Assembly podium, with a poll in between.

The argument behind the new name is consistent across the posts and the speech: artificial undersells the technology and makes it sound less than it is, so super is the more accurate word. The policy behind it is consistent too. Trump has said the United States will not slow the industry, that the Department of Justice and other law enforcement bodies can “rein things in if we have to,” and that no international body should control AI development. The AI czar seat has been empty since David Sacks stepped down in March 2026, according to TechCrunch; the AI Force is modeled on the Space Force he created in his first term, per CBS News.

A president can change the words in executive-branch documents on his own authority. He cannot change them everywhere else, and everywhere else is where the term lives.

• The law says artificial intelligence. Federal statute defines the term (the National Artificial Intelligence Initiative Act of 2020), and so do state AI laws, the EU AI Act and the OECD principles. Rewriting statutes takes legislatures, not a speech.
• The products say AI. Google AI Overviews, Google AI Mode, Apple Intelligence, Microsoft Copilot, Meta AI. Product names are trademarks and habits, and none of the companies has signalled a change.
• The industry keeps superintelligence for the future. OpenAI calls itself a superintelligence research company, Meta named a division after it, and Safe Superintelligence Inc. put it in the company name. All 3 mean the stage they have not reached, and they are unlikely to give up the distinction.
• The public says AI. The search data is not close.

Source: Google Ads keyword data via DataForSEO, 12-month averages to August 2026. The 9,900 for super intelligence is itself inflated by 1 month: in October 2025, when the Future of Life Institute published its statement calling for a prohibition on superintelligence, searches for the word jumped to 49,500 and fell back under 6,000 within 2 months. Announcements move search interest for weeks. Habits move it for years.

The likeliest outcome: super intelligence trends for a few days, appears in federal documents for as long as this administration keeps it there, and AI remains the word people type, buy and legislate under. The name to watch is the other one. If superintelligence in the research sense ever stops being hypothetical, no rename will be needed to make the news.

Nobody knows, and the honest range runs from a few years to never. The optimistic end is loud. OpenAI’s Sam Altman wrote in June 2025 that “humanity is close to building digital superintelligence” and that “we are past the event horizon; the takeoff has started” (The Gentle Singularity). Meta created Meta Superintelligence Labs the same month and put it under Alexandr Wang. Ilya Sutskever, OpenAI’s former chief scientist, left in 2024 to found Safe Superintelligence Inc., a company with 1 goal in its name, valued above $30 billion within a year.

The cautious end is louder this month than it has ever been. In October 2025 the Future of Life Institute published a 1-sentence statement calling for a prohibition on developing superintelligence until there is broad scientific consensus it can be done safely and strong public buy-in; it has passed 75,000 signatures. On September 3, 2026, Senator Bernie Sanders and Representative Greg Casar announced the Ban Artificial Superintelligence Act, which would prohibit developing or deploying superintelligent AI and pause advanced AI development until a federal regulator sets safety rules. This week, 22 heads of government signed a declaration that frontier models need mandatory safety testing and that the UN should consider a body to convene states when capability thresholds are crossed. And on the sidelines of the same Assembly, the United States is hosting its own event to argue the opposite.

What happens if it arrives is the question underneath all of this, and the 2 camps answer it in the same breath. Proponents describe cures, cheap energy and a compressed century of science. Critics describe a system whose goals humans can no longer check, correct or switch off. Both describe a capability that, as of today, exists only in plans. That is the most useful fact for reading the news this week: the rename is a change of words, and the argument the world is having is about a change of kind.

Very little, plus 1 thing that matters. The tools are unchanged, the prices are unchanged, and the models you use tomorrow are the models you used yesterday. The short list:

1. 1Keep saying AI on your site. Your buyers search for artificial intelligence 30 times more often than for super intelligence. Write the words they use; the federal style guide is not your audience.
2. 2Do not rebrand a product around the term. Superintelligence has a fixed technical meaning that your product does not meet, and the claim invites the wrong comparison.
3. 3Expect the phrase in government paperwork. If you sell to federal agencies, mirror their wording in proposals and compliance documents, and keep AI everywhere else.
4. 4Watch the policy, not the vocabulary. An AI czar, an AI Force and a formal US rejection of international oversight are the substantive announcements of the week. Those shape procurement, data-center approvals and export rules. The rename shapes nothing.
5. 5Check whether the assistants recommend you. Whatever they are called, ChatGPT, Gemini, Perplexity and Google’s AI answers are already fielding buyer questions in your category. That is the AI development with a direct line to your revenue this quarter.

The last point is measurable. A Cruelx AI Visibility Scan asks 6 AI engines the questions your buyers ask and reports, engine by engine, whether you are recommended, mentioned or missing. If you want the concepts first, our guides to LLM visibility and getting recommended by ChatGPT cover the metrics and the playbook. The vocabulary will settle on its own schedule. The answers your buyers see today do not wait for it.