Definition

The technological singularity

The technological singularity is a transition in which technological advancement across many fields compounds faster than human-directed research could sustain or humans can reliably predict. It is driven by reinforcing loops, in which each advance makes further advances easier, and its strongest engine is artificial intelligence that first matches and then exceeds human capability.

Three kinds of loop bring it about:

AI improving AI. AI tools, then AGI, then superintelligence, each helping build its successor. This is recursive self-improvement, and it could produce an intelligence explosion.

Compounding within a field. Advances in medicine, materials, energy or mathematics that enable further advances in the same field, with or without AI involved.

Compounding across fields. Better chips, energy, materials and robotics feed AI and one another; AI in turn speeds each of them up.

Advances have always fostered further advances. Three things would separate a singularity from that ordinary progress:

Sustained acceleration. The rate of advance keeps rising across successive cycles, rather than making a one-time jump.

Breadth. It shows up in many fields at once, not one.

Declining dependence on human direction. Humans stop being the limiting factor in choosing, doing and checking the work.

It is a transition, meaning a period that we could be partway into, not a single moment of arrival. Other routes to greater-than-human intelligence, such as human enhancement or human-computer integration, are possible but are not tracked here.

This is a synthesized working definition, not a universally agreed technical standard. Vinge emphasizes the emergence of greater-than-human intelligence; Chalmers analyzes the argument that successive generations of machine designers could produce an intelligence explosion. The broader technological singularity and this particular AI mechanism overlap, but they are not identical concepts. Vinge, 1993; Chalmers, 2010.

How the idea applies to creating AI tools

I. J. Good's foundational argument is that machine design is itself an intellectual activity. A machine that substantially exceeds human intellectual abilities might therefore design better machines, allowing improvement to feed on itself. Good presented this argument in 1965; Stuart Russell reproduces it in his discussion of AI's long-term future. Russell, quoting Good.

Applied to AI engineering, the proposed sequence can be illustrated as follows:

People create AI tools that assist with programming, research, experimentation, and evaluation.

Those tools help improve the models, algorithms, training methods, data processes, or computing systems used to produce AI.

The resulting systems become more capable at performing AI research and development.

Their increased research capability enables further improvements, potentially shortening development cycles and reducing dependence on human intellectual labor.

These are illustrative engineering applications of the feedback hypothesis, not a claim that every step follows automatically or has already been achieved. The defining connection is that improved AI capability strengthens the process that produces further AI capability. MIRI describes this as a positive feedback loop in the ability to improve intelligence. MIRI, Intelligence Explosion FAQ.

Related terms and their boundaries

These distinctions synthesize Good's machine-design argument, MIRI's explanation of general intelligence and feedback, and Chalmers's analysis. Russell; MIRI; Chalmers.

What the definition does not establish

The hypothesis does not establish a date, inevitability, or an agreed numerical threshold. Chalmers treats the transition as an argument with assumptions and possible obstacles, rather than a demonstrated consequence of building human-level AI. Physical limits also prevent treating the idea as a literal promise of infinite computing power. Chalmers, 2010.

Speed and intelligence should also be distinguished. Running an existing process faster is different from making a system more capable at solving problems. Hutter explicitly examines that distinction, possible bounds on intelligence, and how the transition might appear to different observers. Hutter, 2012.

The broader concept is not limited to a single autonomous AI: Vinge also considers networked intelligence, human-computer integration, and biological enhancement. This project tracks artificial intelligence and the fields it compounds with, not those other routes. Vinge, 1993.

Working criterion for this project

For Tracking the Singularity, a useful research question is: Are improvements in AI tools producing measurable improvements in the ability to develop subsequent AI systems, and does that effect persist across successive development cycles?

Suggested evidence includes independently evaluated capability gains, research time and cost, the share of development performed by AI, the human intervention still required, and whether gains survive realistic testing. These are proposed tracking criteria, not an established test declaring that the singularity has occurred. A successful coding assistant, an automated experiment, or a benchmark record alone would not demonstrate the complete feedback process described above.

Sources

Vernor Vinge. *The Coming Technological Singularity: How to Survive in the Post-Human Era* (1993). NASA Technical Reports Server.

I. J. Good. *Speculations Concerning the First Ultraintelligent Machine* (1965), foundational passage reproduced in Stuart Russell's The Long-Term Future of AI.

David J. Chalmers. *The Singularity: A Philosophical Analysis* (2010). Author-hosted paper.

Marcus Hutter. *Can Intelligence Explode?* (2012). Research paper and abstract.

Machine Intelligence Research Institute. Intelligence Explosion FAQ. An explanation from an organization researching the hypothesis, not evidence of universal agreement.

Research compiled November 7, 2026.