Generative artificial intelligence has profoundly transformed software by relying on models capable of generalizing cognitive tasks from massive volumes of data. Robotics is now attempting to make a comparable transition in the physical world. It is on this ambition that Genesis AI, a robotics company founded between Paris and California, is positioning itself, which is unveiling GENE-26.5, a foundation model designed to provide robots with physical manipulation capabilities close to those of humans.
Genesis AI seeks to build a physical foundation model capable of learning complex manual tasks, transferring them between different environments, and generalizing motor skills at scale.
The main obstacle to advanced robotics today remains data. Unlike language models, trained on gigantic corpora accessible on the internet, robotic systems have little physical data that can be used on a large scale. Human gestures, tactile interactions, object manipulation or coordinated movements are difficult to capture, standardize and reproduce.
Genesis AI claims to have built an architecture to circumvent this limitation. The company is developing a robotic hand reproducing human morphology as well as a glove equipped with tactile sensors making it possible to establish a direct correspondence between the human hand, the glove and the robotic hand. This structure aims to transform human gestures into training data directly usable by AI models.
This system would greatly reduce the morphological gap that historically limits robotic learning from human data. Genesis AI also says its collection glove would cost a hundred times less than existing solutions while significantly improving the quality and speed of data collection.
The approach chosen is based on a logic that is now classic in artificial intelligence: accumulating massive volumes of proprietary data in order to train models capable of generalizing their skills. Genesis AI plans to deploy its collection devices directly to industrial partners in order to continuously recover data from real tasks carried out by human operators. At the same time, the company uses subjective point of view videos as well as video content from the internet in order to power its foundation models.
GENE-26.5 was presented through a series of demonstrations highlighting particularly complex tasks for robotic systems: meal preparation, coordinated bimanual manipulation, laboratory experiments, solving a Rubik’s Cube or even musical interpretation on the piano.
These demonstrations precisely target areas where robotics still remains limited: fine manipulation, continuous coordination, force control, dynamic adaptation and interactions in semi-structured environments.
The Genesis AI project is part of a broader evolution in the sector. After language models and multimodal systems, several players are now trying to develop “Physical Foundation Models”, that is to say models capable not only of understanding the world, but also of acting physically in it, like AMI LABS
This transition could profoundly modify the industrial balance of robotics. Until now, most industrial robots have relied on specialized systems, designed for specific tasks in controlled environments. Robotic foundation models, on the contrary, seek to produce more general systems, capable of quickly adapting to new uses.
To achieve this, Genesis AI adopts a full-stack strategy integrating hardware, AI models, data collection systems and simulation. The company is notably developing a simulation platform intended to reduce the “sim-to-real gap”, that is to say the gap between the performances observed in a virtual environment and those obtained in the real world. With more realistic physics and visual engines, Genesis AI seeks to accelerate the training and evaluation cycles of its models without relying exclusively on costly physical testing.
Genesis AI plans to soon unveil its first general purpose robot based on the technologies presented with GENE-26.5. Behind this announcement lies an increasingly intense industrial competition around the next generation of artificial intelligence capable of interacting directly with the physical world.
Co-founded by Zhou Xian And Théophile GervetGenesis AI raised $105 million in seed last year, notably from Eclipse, Khosla Ventures, Bpifrance and HSG, with the support of investors like Eric Schmidt and Xavier Niel. This level of funding places Genesis AI among the most highly capitalized projects in the robotics sector from their initial phase.
