What Is Super Intelligence (SI)? The Complete Guide

Super intelligence is a form of intelligence that greatly exceeds the cognitive abilities of the brightest humans in practically every domain, including science, engineering, strategy, social skills and creativity. You will also see it written as superintelligence, SI, or ASI (artificial super intelligence). It does not exist yet. But in 2026 it became the stated goal of major technology companies and, in name at least, official US government vocabulary.

This guide explains what super intelligence means, how it differs from today’s AI, how it could be built, who is trying to build it, when it might arrive, and why so many people are worried about it.

Super intelligence: a simple definition

The most widely cited definition comes from Oxford philosopher Nick Bostrom, who describes superintelligence as:

“Any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest.”

Two parts of that definition matter:

• “Greatly exceeds.” Not slightly smarter, or as smart as a genius. Superintelligence would be to Einstein roughly what Einstein is to a child, or more.
• “Virtually all domains.” A calculator beats every human at arithmetic, and chess engines beat every grandmaster. Neither is superintelligent, because each is good at only one thing.

So a super intelligent system would not just answer questions well. It could do original science, invent technology, run organizations, out-negotiate people and improve its own design.

Where the idea came from

In 1965, British mathematician I. J. Good, who had worked with Alan Turing at Bletchley Park, published Speculations Concerning the First Ultraintelligent Machine. His key argument is still the core of the debate:

Since designing machines is an intellectual activity, an ultraintelligent machine could design even better machines. There would then be an “intelligence explosion,” and human intelligence would be left far behind.

Good concluded that the first ultraintelligent machine could be “the last invention that man need ever make,” provided it remained under control. Later writers, including Vernor Vinge and Ray Kurzweil, called the moment of this explosion the technological singularity. Bostrom’s 2014 book Superintelligence: Paths, Dangers, Strategies turned it into a mainstream concern.

AI vs AGI vs super intelligence

The easiest way to understand super intelligence is to see where it sits on the ladder of AI capability:

Today’s chatbots and AI assistants are powerful and increasingly general, but they still make basic mistakes, struggle with long-term planning and depend on human direction. For a deeper comparison, read Super intelligence vs AI: what’s the difference?

The three types of superintelligence

Bostrom describes three forms a superintelligence could take:

1. Speed superintelligence. A mind that works like a human mind but much faster. A system thinking a million times faster than us could do a year of human thinking in about 30 seconds.
2. Collective superintelligence. Many smaller minds working together so effectively that the whole far outperforms any human group, company or nation.
3. Quality superintelligence. A mind that is not just faster but smarter in kind, able to grasp ideas humans simply cannot, in the way humans grasp ideas no other animal can.

In practice a real superintelligence might combine all three.

How could super intelligence be built?

Researchers have proposed several possible paths:

• Artificial intelligence. The leading candidate: scaling up and improving machine learning systems until they reach and then surpass human ability.
• Recursive self-improvement. An AI good enough at AI research could improve its own design, triggering Good’s intelligence explosion.
• Whole brain emulation. Scanning and simulating a human brain in a computer, then running it faster or copying it many times.
• Biological enhancement. Improving human intelligence through genetics or medicine. Most experts think this would be far slower.
• Brain–computer interfaces. Linking human brains directly to machines.

Nearly all current investment goes into the first two, which is why “super intelligence” and “AI” are now so closely linked.

Who is building it?

Super intelligence moved from philosophy to business plan in the mid-2020s:

• Safe Superintelligence Inc. (SSI) was founded in 2024 by Ilya Sutskever, OpenAI’s former chief scientist, as a “straight-shot” superintelligence lab.
• Meta created Meta Superintelligence Labs in 2025, with Mark Zuckerberg promising “personal superintelligence for everyone.”
• OpenAI leaders have publicly said their goal extends beyond AGI to superintelligence.
• SpaceX said in October 2026 that its AI unit, SpaceXAI, will be renamed SpaceXSI. (Read the story.)

Meanwhile, in September 2026, the US government ordered federal agencies to use “Super Intelligence (SI)” instead of “artificial intelligence.” (Here’s what that order actually changes.)

When will super intelligence arrive?

There is no agreement, and every date is a forecast:

• The AI 2027 scenario, written by former OpenAI researcher Daniel Kokotajlo and colleagues, imagines artificial superintelligence by late 2027 in its fastest storyline. (Our AI 2027 explainer.)
• Elon Musk has predicted AI smarter than any human by 2026–2027, and smarter than all humans combined around 2030.
• Ray Kurzweil forecasts human-level AI by 2029 and a singularity around 2045.
• Many academic researchers are more cautious, and some doubt superintelligence is achievable at all.

We track these predictions on our superintelligence timeline.

Is super intelligence dangerous?

This is the central question. A super intelligent system could do enormous good, from designing medicines and clean energy to accelerating science. But the same capability makes it hard to control. Researchers highlight a few key problems:

• The alignment problem. How do we make sure an AI’s goals actually match human values, especially when it is smarter than the people checking?
• Instrumental goals. Almost any goal is easier to achieve with more resources and without being switched off. A superintelligence might pursue those sub-goals in ways we didn’t intend.
• Speed and irreversibility. If an intelligence explosion happens quickly, there may be little time to correct mistakes.

These concerns led to the 2025 Statement on Superintelligence, signed by AI pioneers Geoffrey Hinton and Yoshua Bengio and thousands of others, which calls for a prohibition on building superintelligence until it can be done safely and with public support.

Examples of super intelligence

Since no real superintelligence exists, examples come from thought experiments and fiction:

• Scientific superintelligence: a system that solves problems like protein design or fusion energy in days instead of decades.
• Strategic superintelligence: a system that out-plans every government and company at once.
• Fictional examples: from HAL 9000 in 2001: A Space Odyssey to Samantha in Her, fiction has explored minds that outgrow their creators, though most are written as human-like characters rather than truly alien intelligences.

The bottom line

Super intelligence means intelligence far beyond the best human minds in almost every field. It was first imagined in 1965, defined by Nick Bostrom in 2014, and is now the explicit goal of some of the world’s richest companies. It does not exist yet, and nobody knows when or whether it will. But the decisions being made now, in labs and governments, will shape what happens if it does. That is what this site is here to follow.

Keep reading: Super intelligence vs AI · AI 2027 explained · The superintelligence statement · Latest SI news