Nobody agrees on what it means. Everyone agrees it will change everything. The funding continues.
Literal meaning: Artificial General Intelligence (AGI) describes a hypothetical AI system capable of performing any intellectual task that a human can perform — not specialised for one domain but generally intelligent across domains. AGI is distinguished from current AI systems, which are capable in specific domains (language generation, image recognition, game playing) but not generally intelligent in the human sense.
Origin: The concept has been discussed in AI research since the field’s founding. The current use of AGI as a near-term target emerged from OpenAI’s founding mission (“to ensure that artificial general intelligence benefits all of humanity”) and similar framing by Anthropic, DeepMind, and other frontier AI labs. The definition is contested: different organisations use different benchmarks, and some researchers argue that current large language models already meet or approach various definitions of AGI. OpenAI CEO Sam Altman claimed in 2025 that AGI had been achieved. The claim was disputed.
A threshold concept whose definition is contested, whose arrival is claimed and disputed, and whose anticipation shapes policy and investment regardless of whether it is real.
The Appeal: AGI provides a goal state that gives the work of AI development cosmological significance. If you are building AGI, you are not building a chatbot — you are working on the most consequential technology in history. The concept aligns the interests of researchers, investors, and policy makers around a shared vocabulary.
The Friction: The threshold is definitionally unstable. Each time current AI capabilities are described as AGI, the definition shifts: AGI becomes what AI cannot yet do. This moving target serves commercial interests — the closer you claim to be to AGI, the more fundable your work is. The Singularity depends on AGI as a precursor. Mythos Moment is the specific threshold where a model is judged too capable or dangerous to release — a governance concept that presupposes an AGI-like capability threshold. Longtermism uses AGI as the event whose scale justifies present sacrifice. The contested definition means the concept simultaneously mobilises urgency and resists evaluation.
Why This Matters: AGI names a threshold that everyone agrees matters and nobody agrees how to measure. Once you see that ambiguity as functional — not accidental — the race toward AGI is legible as a race to claim a prize whose rules have not been written.
Related terms: The Singularity · Digital Frankenstein · Mythos Moment · Longtermism · Doomerism · AI Literacy · AI as a Utility · Red Teamer
Read more:
- Position: Stop treating ‘AGI’ as the north-star goal of AI research — Blili-Hamelin, B., Graziul, C., Hancox-Li, L., Hazan, H., El-Mhamdi, E., Ghosh, A., Heller, K.A., Metcalf, J., Murai, F., Salvaggio, E., Smart, A.J., Snider, T., Tighanimine, M., Ringer, T., Mitchell, M. & Dori-Hacohen, S. (2025). Proceedings of Machine Learning Research (ICML 2025)
- Levels of AGI for Operationalizing Progress on the Path to AGI — Morris, M.R., Sohl-Dickstein, J., Fiedel, N., Warkentin, T., Dafoe, A., Faust, A., Farabet, C. & Legg, S. (2023). arXiv
- Artificial General Intelligence: Concept, State of the Art, and Future Prospects — Goertzel, B. (2014). Journal of Artificial General Intelligence
- AI researchers are debating what counts as AGI — Hao, K. (2024). MIT Technology Review
- Three Observations — Altman, S. (2025). Sam Altman Blog
- Position: Stop treating ‘AGI’ as the north-star goal of AI research — Blili-Hamelin, B. et al. (2025). Proceedings of Machine Learning Research (ICML 2025)
- What is AI? We drew you a flowchart to work it out — Karen Hao (2018). MIT Technology Review