Invested billions, ever more powerful models, billions of digested tokens every day, linguistic prowess greeted by the markets … And yet, nothing moves on the table, no moved object, no prepared dish, no small reparation.
What is a superintendent if she can do nothing with her hands? This question, posed as a provocation, sums up the founding intuition of Figurethe humanoid robotics startup directed by Brett Adcock. While the digital world is experiencing a revolution with the emergence of general artificial intelligence (AG), the physical world remains largely out of reach.
Act is (almost) there, but she lives in a box
Act, in its current form, is a server intelligence, it advises, plans, writes, calculates. She simulates agents, makes brilliant demonstrations online, but as soon as it is a question of interacting with reality like moving a chair, opening a drawer, or sorting a pile of packages, it collides with a fundamental limit, it has no body.
In the collective imagination, AG is often perceived as dematerialized. Now, an intelligence that can neither touch neither feel nor observe in real situation, remains an abstract intelligence, a kind of brain without nerves, without muscles, disembodied. For Brett Adcock, this is not only a limitation but it is a danger.
“If we do not solve the humanoid robot problem, Act will eventually use us as hands. She will ask us, or force us, do what she cannot do herself.”
Humanoid should not be seen as a gadget but need
Why the humanoid, and not just robotic arms, drones or RPA? Because the world has been built for humans. Size, scope, articulation, spatial logic, social interaction All objects and spaces are calibrated for two -year -old bodies of about 1 meter 70, capable of walking, catching, understanding gestures.
https://www.youtube.com/watch?v=Z3YQYNXPWS
Figure, like Tesla with Optimus, Agility Robotics or Sanctuary AI, has embarked on the development of a generic body for generic intelligence, A robot capable of doing almost all that a human doesin a warehouse, a home, a hospital, a robot that can be Commissioned in natural languagebut above all who learns By observing, executing, and correcting.
An industrial scale question, not just science
The lock is no longer only technological, Figure has already demonstrated that its humanoids could work for more than an hour continuously in logistics tasks, with an on -board model (Helix) managing the entire movement. The packages are sorted at a rate of 3.5 seconds each. Interactions are made in natural language or by SMS.
The challenge is now industrial, how to produce 100,000 robots in four years, at costs compatible with massive adoption. The “Baku” factory is on the way and offers the model Figure 3 90 % cheaper than its predecessor, the design was thought from the start for series manufacturing, no question for the founder of the startup to make it a Protype exposed in the R&D window.
And above all, each robot learns, unlike frozen systems, the human humanoids capture data, progress, share their models. The more many they are, the more intelligent they become, It is a closed loop between the cloud, perception, action and learning.
The real world as the last border
In a world where digital technology has already colonized, there remains a continent to conquer, that of tangible world. If digital agents have transformed our way of looking for, writing or designing, they have not changed our gestures.
Make coffee, empty the dishwasher, move loads, put a beam, serve a meal, install a solar panel. As long as AI does not know how to do this, it remains limited, humanoid is the missing interface Between computational intelligence and concrete transformation.
“It is not enough to think the world. You must also know how to manipulate it,” says Adcock.
Founded in 2022 by Brett Adcock, Figure is an American startup specializing in autonomous humanoid robotics. It designs two -way robots capable of performing physical tasks in industrial and domestic environments, using on -board AI models. In February 2024, the company raised 625 million euros from Microsoft, Openai, Nvidia and Jeff Bezos, bringing its total funding to 695 million euros. This lifting aims to industrialize the production of its third generation of humanoid robots, more efficient and less expensive, and to prepare their large -scale deployment in warehouses, factories, then households.
Figure is part of an increasingly active technological race, faced with actors such as Tesla (Optimus),, Sanctuary Ai (Canada), AGILITY ROBOTICS (Amazon property), 1x technologies (Norway, supported by Openai), or Apptronik (Apollo robot). All bet on the combination of general intelligence and ability to act in unstructured environments. Unlike certain competitors who target specialized robots (delivery, exoskeletons, vertical assistance), a figure bets on a generalist humanoid, like the human body, capable of performing a wide range of tasks without specific software or material adaptation.
In a still young but strongly capitalized market, Figure is distinguished by its vertical integration (design, AI, manufacture) and its clear strategy: producing 100,000 robots by 2029, standardize motor intelligence, and embody Act in the physical world.
China is not behind, quite the contrary, it advances quickly, and often on a large scale. Several humanoid robotics projects emerge there, carried by private or public actors, in a clearly displayed national framework: to become a world leader in intelligent robotics by 2030.
Among the most advanced:
- Fourier Intelligencewith his robot GR-1, presented as a medical or domestic assistant, capable of walking, manipulating objects and even dancing.
- Ubtech Roboticssupported by Tencent, develops Walker X, a connected home -oriented humanoid.
- Xiaomiwith cyberon, targets a robotic – domotic convergence on a general public scale.
- Hanson Robotics (Located in Hong Kong), although more media than operational, maintains its position as a display case.
These initiatives benefit from explicit political support, humanoid robotics is inscribed in the 14th five -year plan as a key sector, at the crossroads of technological sovereignty, industrial automation and demographic aging.
But according to Brett Adcock, the acceptability of these robots outside China remains a major blocking point:
“Chinese robots will have a lot of trouble being accepted in homes or Western companies. Confidence, data security, software sovereignty are decisive issues. And these are even more sensitive questions than for smartphones or cars.
Chinese companies have the advantage of vertical integration, local supply chains and favorable regulations. But they will still have to prove that they can deploy safe, interoperable and culturally accepted humanoids internationally.
