French robotics-focused startup Genesis AI has taken a significant step forward in the field of embodied artificial intelligence with the release of its first foundational model, GENE-26.5, and a human-like robotic hand built entirely in-house. The hand, which closely resembles a human appendage, is capable of executing complex and delicate tasks such as cooking a full meal, playing the piano, or solving a Rubik's Cube. This development marks a pivotal moment for the company, which completed a $105 million (approximately £77.2 million) seed funding round in July of last year—one of the largest seed rounds in European robotics history.
The company, co-founded by chief executive Zhou Xian and president Theophile Gervet, began with a focus on developing an AI model specifically designed to power real-world robotic systems. Rather than relying on third-party hardware, Genesis AI made the strategic decision to build its own robotic hand, believing that vertical integration would allow for more precise control over the full technology stack. This approach is uncommon among startups, which often outsource hardware development to focus solely on software or AI. By controlling both the model and the physical platform, Genesis AI aims to achieve tighter feedback loops between simulation and reality, accelerating the training and deployment of its robots.
Redefining Robotic Dexterity
The robotic hand is particularly noteworthy for its anthropomorphic design. Unlike many robotic end-effectors developed by other companies, which typically feature two or three fingers optimized for specific industrial tasks, Genesis AI's hand boasts five fingers arranged similarly to a human hand. This design choice is intentional: it allows data collected from human demonstrations—whether from gloves worn by workers or video recordings—to be more directly transferable to the robot. The company argues that the closer the hardware matches human anatomy, the easier it is to leverage existing human-centric data sources, such as YouTube tutorials or skilled labor recordings, for training. This reduces the need for thousands of hours of bespoke robot data collection.
In a demonstration video released alongside the announcement, a pair of Genesis AI's hands is shown performing a variety of tasks with impressive fluidity and precision. In one scene, the hands prepare a smoothie by picking up a container, adding ingredients, and operating a blender. In another, the hands play a recognizable tune on a small keyboard. The movements are not jerky or robotic but appear almost natural, with fingers curling and grasping objects in a manner reminiscent of human motion. These demonstrations highlight the potential for such robots to operate in environments designed for humans, from kitchens to factories, without requiring significant environmental modifications.
The GENE-26.5 Foundational Model
At the core of the robotic hand's capabilities is GENE-26.5, Genesis AI's first foundational model. While the company has not released extensive technical details, the model is designed to translate high-level task instructions into precise motor commands for the hand. Foundational models in robotics represent a shift from narrowly trained systems, which can only perform one task (e.g., picking up a specific object), to more general-purpose systems that can handle novel tasks and environments. GENE-26.5 is trained on a combination of simulated data, real-world demonstrations, and human video data, allowing it to generalize across tasks. The model's name, referencing the 26.5 million parameters, suggests a relatively compact size compared to large language models, but it is optimized for real-time control and low latency—critical for safe and efficient robotics.
This approach aligns with broader trends in AI and robotics, where large pretrained models are increasingly used to provide a foundation of common sense and motor skills, which can then be fine-tuned for specific applications. For robotics, this means a single model could potentially control a hand to perform hundreds of different actions, from turning a doorknob to assembling a circuit board. The potential for such general-purpose dexterity has attracted significant investment from major venture capital firms. The seed round was co-led by Eclipse and Khosla Ventures, with additional backing from France's Bpifrance, HSG, former Google CEO Eric Schmidt, telecoms entrepreneur Xavier Niel, MIT CSAIL director Daniela Rus, and Apple distinguished scientist Vladlen Koltun. This diverse group of investors signals confidence in both the technical vision and the market opportunity.
Sensor Gloves and Data Collection
In addition to the hand and model, Genesis AI has developed lightweight sensor gloves that can be worn by human workers in fields such as pharmaceuticals, manufacturing, and logistics. These gloves capture detailed motion data, including finger positions, grip forces, and contact points, which can then be used to train the AI model. The company claims that data from the gloves can be combined with information extracted from videos of workers carrying out tasks, providing a multi-modal training dataset that is both rich and scalable. However, this raises an important ethical question: would workers be willing to help train a potential robotic replacement for their own jobs? While Genesis AI frames the gloves as a tool for improving efficiency and safety, the long-term implications for employment and labor rights remain a concern. The company has not publicly addressed these issues in detail but has stated that its goal is to augment human capabilities, not replace them entirely. Nonetheless, the robotics industry as a whole must navigate the delicate balance between automation benefits and societal impact.
Co-founder Theophile Gervet previously worked at Mistral, a prominent French AI startup focused on large language models. His experience in generative AI likely influenced Genesis AI's approach to using large-scale pretraining. The company retains substantial operations in Europe, which it attributes to the continent's large talent base and potential market of industrial customers. European manufacturing, particularly in Germany and France, represents a significant opportunity for robotics adoption, as companies seek to automate repetitive tasks while dealing with labor shortages. Genesis AI's team of 60 employees is split roughly evenly between the United States and Europe, with a slightly larger group based in San Carlos, California. The company also has offices in Paris and London, allowing it to tap into diverse engineering talent and maintain close relationships with European industrial partners.
Future Plans and Industry Context
Looking ahead, Genesis AI is working to supplement its mechanical hand with a full-body, general-purpose robot. This is an ambitious goal that many robotics companies have pursued, with varying degrees of success. Competitors such as Tesla (with Optimus), Figure AI, and Agility Robotics have also announced plans for humanoid robots, but the field is still in its infancy, with few commercially viable products. Genesis AI's strategy of starting with a highly dexterous hand and then building out the rest of the body makes a certain amount of technical sense: manipulation is often the hardest problem in robotics, and solving it first could provide a strong foundation. The company has not revealed a timeline for the full-body robot, but the seed funding should provide sufficient runway to develop prototypes over the next few years.
From a broader perspective, the release of GENE-26.5 and the robotic hand signals that the convergence of AI and robotics is accelerating. Foundational models are no longer limited to text and images; they are expanding into the physical world. As these models improve, we can expect robots to become much better at understanding and interacting with unstructured environments. This has implications far beyond manufacturing, including healthcare (e.g., assisting surgeons), elder care, household chores, and disaster response. However, challenges remain. Safety is paramount—a robot with a human-like hand has the potential to cause harm if not properly controlled. The cost of production is also likely to be high initially, limiting deployment to high-value applications. Regulatory frameworks for autonomous robots are still evolving, and public acceptance will be critical.
Genesis AI's approach of combining a strong AI model with purpose-built hardware positions it well in the competitive landscape. The company's dual presence in the US and Europe allows it to access top talent and capital while staying close to industrial customers. The involvement of high-profile backers like Eric Schmidt and Xavier Niel adds credibility and opens doors. If the company can successfully scale its technology and deliver a full-body robot, it could become a major player in the emerging humanoid robotics market. For now, the demonstration of a dexterous hand capable of cooking and playing the piano serves as a compelling proof of concept—and a glimpse into a future where robots work alongside humans in increasingly sophisticated ways.
Source: Silicon UK News