French artificial intelligence startup AMI Labs (standing for Advanced Machine Intelligence) raised $1 billion three weeks ago to develop AI systems designed to understand the physical world "in the way animals and humans do", unlike the language-based models behind chatbots such as ChatGPT.
AMI Labs expects to produce "fairly universal intelligent systems" within five years, a timeline that investors described as "ambitious but compatible with current funding conditions."
Chauffeur
In order to become profitable within twelve months, the first product milestone, code-named Chauffeur, is a humanoid private driver designed to make autonomous driving a reality much faster than most car manufacturers have ever envisioned.
Early profitability was a condition imposed by investors, among them Toyota Ventures, which is pushing for solutions compatible with the existing global fleet of cars, estimated at more than 1.4 billion vehicles.
"Make the driver intelligent, not the car."
Yann LeCun
Chauffeur sits in the driver's seat, fastens the seatbelt, adjusts the mirrors, and performs the usual pre-driving ritual familiar to human drivers.
It manipulates the pedals for the accelerator, brakes, and clutch with its feet, inserts the key, changes gears, turns the steering wheel, and operates dashboard controls with its hands.
The system can refuel at gas stations, pay for parking, and in the event of a flat tire, replace the wheel using tools stored in the trunk.
Chauffeur is powered directly by the vehicle, drawing electricity from the cigarette lighter.
To reduce costs, the most basic version will consist of a lightweight skeletal frame with minimal cosmetic elements. An additional benefit is that it can fold itself away like a Transformer or a Brompton bicycle and be stored in the trunk when five seats are required.
This version targets the European and US mass markets, as well as the Global South, where many decommissioned cars enjoy a long second life.
A premium model will target the luxury segment, with the explicit requirement that the Chauffeur be indistinguishable from a human driver. Engineers describe the design goal as "Rheya in Tarkovsky's Solaris"
Yann Le Cun ouvre une nouvelle page de l’intelligence artificielle. C’est la France des chercheurs, des bâtisseurs et des audacieux. Bravo ! https://t.co/hVxMMFoabJ
— Emmanuel Macron (@EmmanuelMacron) March 10, 2026
A Flying Start
Because AMI Labs focuses primarily on software, the company has partnered with the Polish humanoid robotics manufacurer Clone Robotics, the first company producing a biomorphic hand which is in form and function fully equivalent to the human hand.
On the software side, AMI Labs is developing world models that learn abstract representations of real-world sensor data, ignoring unpredictable details, and making predictions in representation space.
Representation space means geometric entities, means triangle meshes, means CGAL, as well as other Open Source geometry libraries such as Geogram and OpenCascade, but also simulation frameworks such as Sofa and Salome, and, last but not least, Rocq, a trustworthy, industrial-strength interactive theorem prover and dependently typed programming language for mechanised reasoning.
For that reason, AMI Labs will actively apply a trickle-down funding approach and, over the next three years, donate 0.1% of the $1 billion raised to these Open Source projects.
"Using Open Source software developed mainly in France, we stand on the shoulders of giants."
Alexandre Lebrun, CEO of AMI Labs
What exactly will be used from CGAL is not yet fully clear, but point-cloud processing, shape detection, and surface reconstruction are among the leading candidates for the interior perception system.
Even when pre-trained on typical car interiors, Chauffeur must remain capable of identifying objects placed on the dashboard and deciding whether to ignore them, move them aside, or comment on them politely.
