What are the projects that are winning the robotics race?

The new jewel in the technological crown is startups who specialize in intelligent behavior androids. Olympus is reserved for whoever develops the first generative robot which, in turn, replicates the complexity of the human frame and its functioning.

Applicants must synthesize equal parts hardware and software in the package of a humanoid. The condition will be that you can obey instructions and maintain sensible conversations, through deep machine learning models.

Behind this concept stands one of the chimeras of classic science fiction: having a mechanical assistant suitable for tasks at home, in the office or in the factory.

“There is still time to fulfill the dream of humanoid robots, except Roombas o Peppers in airports. The problem is that they are still expensive complicated to maintain and security measures are lacking,” warns Marcela Riccillo, doctor in computer science, expert in AI and robotics.

These intelligent models descend from those machines that moved clumsily and slowly. The point of evolution came with the robotic vision and the advances brought about by generative artificial intelligence (AI).

Computer vision not only made it easier to navigate complex environments, but it made it possible to climb stairs, deftly avoid obstacles, differentiate faces, and make decisions in real time.

In turn, the battery with greater energy density provided limb movement quickly enough to maintain balance dynamically and avoid stumbling.

Goldman Sachs analysts predict that the humanoid robot market will reach $38 billion in 2035 and by 2030, more than 250,000 units.

On the back of this dark glass, many of the gurus of the tech industry such as Elon Musk, Reid Hoffman, Bill Gates and Eric Schmidt are layered. Along with some leading firms such as Microsoft, NVidia, Samsung, Intel and Amazon.

“Incorporating chatbots into humanoids is not new. In Argentina, IBM adapted Watson to an NAO to improve speech. What could be seen as novel is a revival of Humanoid Robotics (which had its peak from 2012 to 2016) with the advantage of adding ChatGPT to its structure,” explains Riccillo.

The four firms that are at the forefront today are: Figure, 1X Technologies, Tesla and Sanctuary AI. Who are the magnates who support them?

Figure A.I.managed to capture the attention of many investors by receiving $675 million in financing, coming from renowned figures and institutions, such as Jeff Bezos.

Leading this initiative is billionaire Brett Adcock, whose goal is to create, in record time, “the world’s first commercially viable general-purpose humanoid robot.”

The team, made up of 60 robotics experts recruited from Boston Dynamics, Tesla, Google DeepMind and Archer Aviation, last month showed off Figura One, a robot designed to execute tasks with human-like precision.

And while the technical aspects are scarce, in the preview, Figura One is seen in workshops, warehouses and retail outlets, handling tools and performing household tasks, such as making coffee.

This week, OpenAI signed a collaboration agreement to provide Figura with “next-generation AI models for this robot that can help in everyday life.”

This support allowed Figure AI to close its first commercial agreement with BMW Manufacturing, marking its introduction into the luxury vehicle production chain.

1X Technologies is a Norwegian company that, in the last year, received more than $125 million from the Japanese conglomerate SoftBank and OpenAI, to shape NEO.

This project is trained in practical skills thanks to a technique of embedded learning. Thus, instead of programming robots with algorithms, it instructs them through virtual simulations based on everyday environments.

In early March, a preview was given of how NEO behaves in society and some of its parameters were provided. It measures 1.67 meters, weighs 30 kilograms and can carry objects weighing up to 20 kilograms.

Its distinctive feature is that it has a soft body, without pinch points or sharp edges. The intention is that its padded structure does not represent a threat to its owner.

It has a range of up to 4 hours and allows the android to run up to 12 kilometers per hour. When it detects that the power is low, it automatically finds its way to the charging station.

In 2024, Canadian firm Sanctuary AI introduced Phoenix, a humanoid robot equipped with a hydraulic drive system. This technology gives you the ability to manipulate objects with unique precision, speed and strength.

This manual skill was inspired by how babies learn and interact with their environment. As a result, Phoenix is ​​able to replicate 31 of the 33 different types of hand grips and twists.

It is powered by AI Carbon, a software platform that uses sophisticated AI to make the machine think and act like a real factory operator.

This control software reproduces subsystems of the human brain such as memory, vision, hearing and touch. By simply observing a craftsman, Phoenix can imitate any movement, no matter how complex.

The height of the robot is 1.70 cm, it weighs about 70 kilos and can move at a speed of 5 kilometers per hour. It can lift up to a weight of 25 kilos.

This second version was announced by Elon Musk himself and noted that it was profiled to resemble a human being. This is noticeable in his height as well as in his mobility, thanks to the improvements implemented in his hands, feet, neck and body structure.

This Tesla automaton is capable of walking 30% faster and its weight was reduced to a total of 10 kilos, to improve general balance when moving.

Among the highlights of Optimus Gen 2 are hands that can manipulate fragile objects such as an egg, through touch sensors on all fingers that measure the pressure of the elements.

Its ‘brain’ consists of a central board embedded with AI and a navigation system similar to the one Tesla incorporates for its autonomous vehicles.

As a whole, the robot gained more control over its body and balance, which makes it easier to transport heavy loads, something that also helped it maintain stability when moving across uneven surfaces.