GENEKI
Start a project
  1. Home
  2. Blog
  3. Superintelligence vs AGI

AI

Superintelligence is not just bigger AGI.

Researchers, AI labs and now governments all say 'superintelligence', and they do not mean the same thing. What the word originally meant, how companies use it today, and a simple test for reading any claim that uses it.

5 min read Geneki team

On this page
  1. What the word first meant
  2. AGI vs superintelligence
  3. How the word is used now
  4. A four-question test
  5. Questions people ask
Dark metal chess pieces lined up on a polished wooden board
Photo: Lou Levit, Wikimedia Commons (CC0)

The short version

  • AGI is about breadth: AI that can do most of what a human mind can do. Superintelligence is about level: far beyond the best humans.
  • The word has moved from philosophy into company slogans and, in the US, government paperwork.
  • Companies use it for very different things, from a bounded helper to an unbounded researcher.
  • Four questions will tell you what a given claim really means.

What the word first meant

The philosopher Nick Bostrom defined it in a 1998 paper called "How long before superintelligence?". He described superintelligence as "an intellect that is much smarter than the best human brains in practically every field". He listed examples such as scientific creativity, general wisdom and social skills. He did not specify how it would be built, and he did not say it had to be conscious.

In his 2014 book Superintelligence: Paths, Dangers, Strategies, Bostrom described several forms: one that thinks much faster than a person, one made of many smaller minds working together, and one that is simply better in kind at intellectual tasks than any human can be. We did not have the book to hand, so we describe these in general terms rather than quoting it.

The key point is the bar. "Smarter than the best humans in practically every field" is a far higher bar than "as capable as a typical person".

Dark metal chess pieces lined up on a polished wooden board
Beating the best player in every game is a different bar from playing like a typical one. Photo: Lou Levit, Wikimedia Commons (CC0)

AGI vs superintelligence

Our guide to what AGI is covers the definitions in detail. Briefly, OpenAI's charter describes AGI as highly autonomous systems that outperform humans at most economically valuable work, and Google DeepMind has described AGI as AI "at least as capable as humans at most cognitive tasks". Both are about covering a wide range of tasks at human level or above.

OpenAI's 2023 post on governance draws the line this way. Its subtitle describes superintelligence as "future AI systems dramatically more capable than even AGI". So in the labs' own usage, superintelligence is the step after AGI.

AI todayAGISuperintelligence
BreadthStrong but unevenMost cognitive tasksPractically every field
LevelOften expert in narrow tasksAt least human levelFar above the best humans
StatusExistsContested, not agreed to existNot claimed to exist in the technical sense

How the word is used now

By 2026 the word had spread well beyond philosophy. Here is how six organizations use it, in their own words.

  • OpenAI. In November 2025 it wrote "we treat the risks of superintelligent systems as potentially catastrophic" and "no one should deploy superintelligent systems without being able to robustly align and control them".
  • Meta. Its July 2025 letter says "Meta's vision is to bring personal superintelligence to everyone" and "Developing superintelligence is now in sight". It frames the choice as a tool for personal empowerment or a force "focused on replacing large swaths of society".
  • Microsoft AI. Mustafa Suleyman's November 2025 essay proposes "humanist superintelligence": "incredibly advanced AI capabilities that always work for, in service of, people and humanity more generally." It says "we are not building a superintelligence at any cost, with no limits" and "not an unbounded and unlimited entity with high degrees of autonomy".
  • Safe Superintelligence Inc. The company says it has "one goal and one product: a safe superintelligence", and that it plans to advance capabilities "as fast as possible while making sure our safety always remains ahead".
  • Anthropic. Dario Amodei's January 2026 essay does not use superintelligence as its own term. It describes "powerful AI": smarter than a Nobel Prize winner across most relevant fields, running as millions of instances at 10 to 100 times human speed, "a country of geniuses in a datacenter".
  • The US government. Executive Order 14434 (29 September 2026) tells federal agencies to say "Super Intelligence" in place of "AI" and defines it as the technologies already covered by the legal definition of AI. See our explainer.

Notice the range. Meta describes a personal, individual tool. Microsoft describes a bounded one. SSI describes a single research goal. Amodei describes an unprecedented concentration of cognitive power. The US order describes today's systems. They all say "superintelligence", and they are not talking about the same thing.

A four-question test

When you see a claim that uses the word, ask:

  1. 1. Which meaning?Is it Bostrom's meaning (far above the best humans in nearly every field), a product slogan or a government label?
  2. 2. Today or tomorrow?Is the claim about a system that exists, or a system the speaker expects?
  3. 3. Measured how?Is there a benchmark, a test or a definition behind it?
  4. 4. Who gains from the claim?Not to dismiss it, but to weigh it. Companies raising money, governments seeking votes and campaigners seeking attention all have reasons to use the word.

For business leaders: you do not need to settle the definition to act sensibly. Plan around what AI can do today and the next two years, and treat superintelligence as a scenario to track, not a date on your calendar. Our guide to what the forecasts really say helps with that.

Related guideEveryone talks about AGI. Nobody agrees what it is.The step before superintelligence, explained for business leaders.

Questions people ask

What is the difference between AGI and superintelligence?

AGI is usually defined by breadth: AI that can do most cognitive tasks a human can. Superintelligence is defined by level: Nick Bostrom described it as an intellect much smarter than the best human brains in practically every field. OpenAI describes superintelligence as future systems dramatically more capable than even AGI.

Does superintelligence exist today?

No source we reviewed says so in the technical sense. Meta's 2025 letter says developing superintelligence is 'now in sight', and Sam Altman wrote that humanity is close to building digital superintelligence, but both describe the future. A US executive order uses the label 'Super Intelligence' for today's technology, while defining it as the same systems the law already calls AI.

Is superintelligence the same as sentient or conscious AI?

No. Bostrom's definition does not say how such an intellect would be built or whether it would be conscious. It is about capability, not experience.

What does 'personal superintelligence' mean?

It is Meta's term. Its 2025 letter describes a vision of bringing personal superintelligence to everyone, as a tool for personal empowerment. It is a product vision, not a technical definition.

Why do different companies use different words?

Partly because the field has no agreed definitions, and partly because the words carry meaning beyond technology. Superintelligence suggests both great promise and great risk, which suits different messages. Anthropic's CEO avoids both terms and says 'powerful AI' instead.

Cutting through the AI hype?

Geneki helps teams separate what AI can do today from what is promised, and build what is useful now.

Talk to us

Sources

This explainer quotes company and government texts as published. Bostrom's 2014 book was not available to us, so its three forms are described in general terms. Definitions in this field are contested and change quickly.