Glossary
The terms you will keep running into, explained briefly.
- AGI (artificial general intelligence)
- An AI that matches human ability across most cognitive tasks, rather than excelling at one.
- Alignment
- The work of making AI systems reliably pursue the goals their designers and users actually intend, including in situations nobody anticipated.
- ASI (artificial superintelligence)
- An AI that greatly exceeds the best human performance in virtually every domain. See What superintelligence is.
- Embodied AI
- AI that perceives and acts through a physical body, such as a robot, rather than only through text or software.
- Foundation model
- A large model trained on broad data that can be adapted to many tasks. GPT, Gemini and Claude are examples.
- Intelligence explosion
- I. J. Good's 1965 idea that AI capable of improving AI could trigger rapid, compounding gains in capability.
- Moravec's paradox
- The observation that reasoning is relatively easy for computers while perception and movement, effortless for people, are hard.
- RaaS (robots as a service)
- Leasing robots by the hour or month instead of buying them; common for warehouse humanoids.
- Sim-to-real
- Training a robot in simulation and transferring the learned skill to a physical machine.
- Teleoperation
- A human controlling a robot remotely, often to collect training data or to handle tasks the robot can't yet do alone.
- VLA (vision-language-action) model
- A model that takes in camera images and an instruction in plain language and outputs robot movements.