Google is set to develop artificial intelligence to manage the process of nuclear fusion. To achieve this, the company has partnered with Commonwealth Fusion Systems (CFS).
Inexhaustible Energy
The joint project aims to create AI agents that will control the plasma in the SPARC reactor and optimize its operation. The goal is to move closer to creating a virtually inexhaustible energy source.
"Nuclear fusion -- the process that powers the Sun -- promises humanity clean and virtually inexhaustible energy without long-lived radioactive waste," according to a Google press release.
The parties have focused on three key research areas. In the first, Google and CFS have developed an accurate open-source simulator, TORAX, which allows for virtual experiments with various operational modes of the installation.
The system is built on the JAX platform and can operate on both CPU and GPU, integrating AI models to enhance performance.
"[TORAX] has saved us countless hours in setting up and running models for the SPARC project," noted CFS Senior Manager of Physics Operations Devon Battaglia.
The second area involves using reinforcement learning to find optimal parameters for the nuclear fusion reactor's operation. AI agents analyze millions of possible configurations of magnetic coils, fuel delivery systems, and heating power to achieve a positive energy balance.
The third focuses on developing real-time systems to manage thermal loads. SPARC will generate intense heat concentrated in small areas, and AI must learn to distribute these loads dynamically.
Google has also become an investor in CFS, supporting the commercialization of the technology. The SPARC reactor is expected to become the first magnetic nuclear fusion installation to surpass the breakeven point, where the generated power exceeds the energy costs of maintaining the reaction.
Earlier in October, the company, in collaboration with Yale University, developed an AI model for cancer treatment.