Meta takes a decisive step in the humanoid robot race by acquiring Assured Robot Intelligence, a startup that develops advanced AI systems to give robots a "brain." According to TechCrunch, this move marks Zuckerberg's company’s official entry into a field where Tesla, Figure, and Apptronik have been leading. The purchase of Assured Robot Intelligence, founded by former researchers from Boston Dynamics and Carnegie Mellon, does not involve robot manufacturing but rather the integration of software capable of interpreting and acting in the physical world. Its specialty is "embodied AI": algorithms that allow humanoids to adapt to unpredictable situations, such as reacting to falling objects, interacting with moving people, or navigating uneven surfaces. Meta is thus betting on the development of the "brain" needed to animate the hardware, in line with its history in AI: models like Llama, Segment Anything, and its advancements in multimodality. The acquisition, which TechCrunch reports could exceed one hundred million dollars, strengthens the robotics division led by Marc Whitney, who was hired last year to drive this strategy. The context is favorable: humanoids are no longer just lab prototypes. Tesla plans to commercialize Optimus in 2027, Figure collaborates with OpenAI and BMW, and Apptronik is advancing alongside Google DeepMind. Meta, which has already published its roadmap in robotic AI, seeks to leverage its ability to train large models and its experience in opening technologies to the ecosystem. A potential open-source "Llama-Robot" could revolutionize the sector, allowing manufacturers to integrate advanced AI into their own robots. The story recalls Google's attempt with Boston Dynamics in 2013, which ended in a sale due to a lack of strategic fit. In contrast, Meta is focusing on intelligence rather than hardware, a move more aligned with its strengths and potentially more scalable. The challenge now is execution. While Meta has had successes like Ray-Ban glasses and Quest, other hardware projects like Portal or the metaverse have not taken off. The humanoid market already exists, and competition is real, so the key will be whether Meta can realize its vision and, importantly, if it dares to open its models to the public, which could change the game. There are no official dates, but Meta is expected to present its first AI model for humanoids between late 2026 and early 2027, possibly at its Connect event. If it opts for an open launch, it would be a strategic blow to the competition. For now, the bet is solid and consistent with Meta's trajectory in AI. Source: TechCrunch.