What Is Superintelligence?

Superintelligence is a hypothetical tier of artificial intelligence (AI) that outperforms the best human minds at virtually every cognitive task, and no system today meets that standard.

At Layer3Labs, we build AI workflow automations for small and mid-sized businesses using today's AI, and IBM calls superintelligence "still a hypothetical future state."

Since September 2026, US federal agencies also use "Super Intelligence" (SI) as their word for today's AI, which is a different thing from ASI.

What Superintelligence Means

Superintelligence refers to a hypothetical software system with cognitive abilities that dramatically exceed the intellectual performance of any human being across every domain. In the 2014 book Superintelligence, Nick Bostrom defined the concept as any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest, as documented on Wikipedia. According to IBM, artificial superintelligence (ASI) is a hypothetical software-based AI system with an intellectual scope beyond human intelligence.

Bostrom, IBM and the Levels of AGI paper all treat ASI as hypothetical. As of October 2026, no software system achieves general intellectual superiority over human beings, and current LLMs are not considered AGI by IBM.

The Cognitive Ladder from Narrow AI to ASI

The Levels of AGI paper gives the clearest ladder from today's AI to ASI. Its authors, Google DeepMind researchers including Shane Legg, posted it on arXiv in November 2023 and revised it in September 2025. Its matrix rates software on performance and generality across Levels 0 to 5, from basic computation to superhuman capability.

The DeepMind framework separates systems into narrow AI, which performs specific tasks, and general AI, which transfers competence across diverse domains. The paper rated some narrow systems, such as Stockfish and AlphaFold, as superhuman. It rated the general systems it reviewed (ChatGPT, Bard, Llama 2 and Gemini) only Level 1 Emerging AGI, equal to or somewhat better than an unskilled human.

The paper places artificial superintelligence exclusively at Level 5 of general intelligence, defining it as a system that outperforms 100 percent of human beings across every cognitive domain. Level 2 Competent AGI means at least the 50th percentile of skilled adults, and the paper says it "best corresponds to many prior conceptions of AGI." The paper says no public system had reached it at the time of writing. Our AGI versus ASI comparison sets the top two rungs side by side.

• Level 0 No AI: Narrow systems include basic calculators and compilers, while general implementations rely on human-in-the-loop computing like Mechanical Turk.
• Level 1 Emerging: Narrow systems include Good Old-Fashioned Artificial Intelligence (GOFAI) and SHRDLU, while the paper's general examples at the time of writing were ChatGPT, Bard, Llama 2 and Gemini, at about the level of an unskilled human.
• Level 2 Competent: Narrow systems include Siri, Alexa, Watson, and Jigsaw, while the paper said no public system had reached Competent general AGI, matching the 50th percentile of skilled adults, at the time of writing.
• Level 3 Expert: Narrow systems include Grammarly, Imagen, and DALL-E 2, while the paper listed Expert general AGI, matching the 90th percentile of skilled adults, as not achieved at the time of writing.
• Level 4 Exceptional: Narrow systems include Deep Blue and AlphaGo matching the 99th percentile of skilled adults, while the paper listed Exceptional general AGI as not achieved at the time of writing.
• Level 5 Superhuman: Narrow systems include AlphaFold, AlphaZero, and Stockfish outperforming all humans, while general Artificial Superintelligence has not been achieved.

Origins of the Intelligence Explosion Concept

In 1965, I.J. Good introduced the concept of an intelligence explosion, as documented on Wikipedia. Good wrote that an ultraintelligent machine could design even better machines, and there would then unquestionably be an 'intelligence explosion'.

The term artificial general intelligence was used by Mark Gubrud in 1997 and popularized in the 2000s by Shane Legg and Ben Goertzel, according to Wikipedia. While AGI describes software matching or exceeding human cognitive abilities across any task, superintelligence represents an intellect that greatly exceeds human cognitive performance in virtually all domains.

How AI Companies Use the Term Superintelligence

On May 22, 2023, OpenAI leaders Sam Altman, Greg Brockman, and Ilya Sutskever published a blog post titled "Governance of superintelligence" on OpenAI. The authors described superintelligence as future AI systems "dramatically more capable than even AGI." They wrote that "it's conceivable that within the next ten years, AI systems will exceed expert skill level in most domains."

Two organizations use the word in their names: Meta Superintelligence Labs, Meta's AI division, and Safe Superintelligence Inc. (SSI).

Global Governance and the Superintelligence Statement

On October 20, 2025, the Future of Life Institute (FLI) published the Statement on Superintelligence.

The declaration states: "We call for a prohibition on the development of superintelligence, not lifted before there is broad scientific consensus that it will be done safely and controllably, and strong public buy-in." As of October 2026, the document has gathered more than 77,000 signatures.

Differences Between Federal SI and Technical ASI

The federal government's Super Intelligence label refers to today's AI as defined in federal law. It does not refer to artificial superintelligence. On September 29, 2026, the White House issued Executive Order 14434. The executive order directs executive-branch agencies to use the term "Super Intelligence" or "SI" instead of "AI" in their own non-statutory documents.

The legal text of Executive Order 14434 anchors the definition of SI directly to existing statutory law under 15 U.S.C. 9401(3). Under that statute, artificial intelligence is defined as a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.

The presidential directive does not reference artificial general intelligence or artificial superintelligence. When a federal agency document says SI, it means the AI systems the statute already covers. Our AI vs SI comparison and our breakdown of the Super Intelligence executive order cover the label in detail.