Data Crisis and GPT-2 Era Debates in Physical AI and Robotics
While physical AI companies try to overcome the training data crisis, the future of the industry and the processes of moving beyond the GPT-2 era are being debated.
While the physical AI sector is receiving billions of dollars in investments, the lack of high-quality training data required for robots to perform reliable commercial tasks creates a significant crisis. Industry representatives compare this situation to the GPT-2 era.
Data Crisis in Physical AI
Companies operating in the field of physical AI are facing a major data crisis due to the lack of high-quality data to train models. While generalized robots remain distant, end-to-end learning processes have yet to achieve commercial success.
Unitree's Loss of Value
China's leading robot manufacturer, Unitree, lost a significant portion of its market value shortly after its initial public offering. Analysts point to the lack of ability of robots to perform value-added tasks, even though their physical capabilities are increasing, as the primary reason for this decline.
Autonomous Vehicles and Robotics Research
The autonomous vehicle sector is pioneering model development in robotics thanks to the advantage of collecting data through human driving. While companies like Wayve and Uber are establishing research labs focusing on humanoid robot form factors, Tesla continues its Optimus project.
Hardware and AI Integration
Genesis AI CEO Théophile Gervet argues that hardware and artificial intelligence must be co-designed, whereas Wayve CEO Alex Kendall supports a hardware-agnostic general model approach.
Vertically Focused Solutions and Future Expectations
Companies such as Gritt, Agility, and Bedrock operate in the field by focusing on specific vertical domains, while general-purpose humanoid robots have not yet fully emerged from the laboratory phase. Industry leaders are debating when a true inflection point for physical AI will arrive.