Foundation model
A versatile pre-trained AI model that can be adapted to a wide range of tasks or robot morphologies.
A large AI model pre-trained on massive, heterogeneous datasets and designed to be adapted (fine-tuned) to multiple tasks without full retraining. In robotics, the term refers to a model capable of controlling robots of differing morphologies (arm, humanoid, quadruped) from a shared foundation. EN: foundation model.
Articles about “Foundation model”
- Unitree before the stock market: 6,099 billion raised, but the robot is no longer the rare product
- Early 2027, LG to unveil a two-legged humanoid: what if it one day walked into your home?
- FCC: The 200 kbit/s threshold that traps foreign mobile robots
- A robot performs a gallbladder operation, but a surgeon remains in control
- Munich, Paris, Chicago: the robot has a body, but its brain is rented from two addresses
- It doesn’t do the housework, but it has surpassed 2,100 reservations: the robot that will change your evenings
- The robotic hand that cracks the egg: the West changes course
- Robots: three strategies to capture the missing data
- Five announcements, one word: autonomy does not carry the same weight depending on who claims it
Related terms
Join the discussion Soon
A question, a disagreement, hands-on feedback on this term? The Botoide community forum is coming soon.