Artificial general intelligence
The full name for AGI: an AI system with human-level or superhuman general reasoning, not limited to single domains.
Think of it like
The opposite of a calculator: a system that doesn't just add numbers, but can learn to do anything a human mind can do.
Example
A system that learns to code from a textbook, then uses code to solve biology problems, then designs experiments to test them — true transfer and reasoning.
How it actually works
Artificial general intelligence is the inverse of narrow AI. Narrow AI (like current LLMs or game-playing agents) is superhuman in one domain but helpless in another. AGI would transfer learning across domains, learn from minimal examples, adapt to novel problems, and reason with common sense. The central tension: nobody knows if AGI is an emergent property of scale (bigger models -> AGI), requires new architectures, or is even achievable at all. Academic research focuses on benchmarks and partial steps; industry hype often conflates impressive narrow systems with progress toward AGI.
For product teams
A distant goal that justifies long-term AI investment and shapes strategy; timing is highly uncertain.
For engineers
A theoretical AI system with robust transfer learning, meta-learning, and reasoning across all task domains.
Related
- AGI — Often abbreviated as AGI.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome