Google Overtook ChatGPT with Gemini; Is Superconductor Next? [Tech Talk]

Google DeepMind Founder Shows Interest in Nuclear Fusion Research
Plans to Automate Superconductor Research with AI
Robotic Lab to Conduct Hundreds of Experiments Daily

Demis Hassabis, CEO of Google DeepMind and a central figure in Google's artificial intelligence (AI) strategy, has drawn attention by naming "superconductor" development as his next goal. Success in developing superconductors is expected to make research into future energy technologies, including nuclear fusion, much easier, thereby securing sufficient energy supply. While Google possesses both AI hardware and software technologies, the issue of energy supply for its AI data centers remains an unresolved challenge.

Google DeepMind CEO: "The AI Problem Ultimately Becomes an Energy Problem"

Demis Hassabis, Founder and CEO of Google DeepMind. Photo by CNBC

In an interview with CNBC on January 16 (local time), CEO Hassabis asserted, "Google is well positioned to benefit in the AI market." As the founder and CEO of Google DeepMind, an AI research subsidiary based in the UK, Hassabis currently leads AI development projects including Gemini. He is credited with turning around Google's AI business, which had been considered to lag behind OpenAI's ChatGPT until just a few years ago.

While expressing confidence that Google is "the company with the long-term vision to achieve (human-level) AI the fastest and safest," Hassabis also acknowledged that power supply bottlenecks are a serious constraint. He explained, "Ultimately, the AI problem boils down to an energy problem," and added, "What is interesting is that AI can also be used to develop future energy technologies, including superconductors."

Superconductor Research in an AI Lab Conducting Hundreds of Experiments Daily

In fact, Google DeepMind is actively investing in next-generation energy technologies. Last October, Google DeepMind formed a partnership with the US nuclear fusion research startup Commonwealth Fusion. Commonwealth Fusion is developing a compact nuclear fusion power device called SPARC.

Google DeepMind is envisioning an automated scientific research facility connected to the artificial intelligence Gemini. Google DeepMind website

In the process of developing nuclear fusion reactors, Google DeepMind is focusing on the field of room-temperature superconductor research. Superconductors are both a core component and a major technological challenge for nuclear fusion reactors. A nuclear fusion reactor generates energy by producing extremely hot plasma at around 100 million degrees Celsius through fusion reactions, and superconductors are needed to generate strong magnetic fields to stabilize this plasma.

However, currently developed superconductors only operate in extremely low-temperature environments of about minus 280 to minus 200 degrees Celsius, which makes them impractical. If a superconductor that functions at room temperature (15 to 25 degrees Celsius) could be developed, it would be a major step toward the realization of nuclear fusion reactors, often called the "dream energy source."

In the interview, CEO Hassabis emphasized, "Developing a room-temperature superconductor is also my personal pet project," and added, "In fact, I believe superconductors are the most useful application case for AI."

Commonwealth Fusion, a US nuclear fusion reactor startup collaborating with Google DeepMind. Google DeepMind homepage

Google DeepMind's superconductor research will be conducted at an "automated scientific research facility" scheduled to be established in the UK. This facility will integrate robotic equipment needed for new material experiments with Gemini, automating the entire research process, including material synthesis, experimentation, observation, and data analysis. Google DeepMind explains, "The lab's AI will control robots to synthesize and analyze the properties of hundreds of materials in a single day," and "this will dramatically shorten the time required to discover innovative new materials."

AI to Discover Control Methods for Nuclear Fusion Reactors

Additionally, Google DeepMind has agreed to provide its proprietary AI to Commonwealth Fusion. This AI, called "TORAX," simulates the plasma formation process in nuclear fusion reactors. Until now, scientists have had to operate nuclear fusion reactors directly to study how plasma forms, which is both time-consuming and extremely costly. According to Google DeepMind, TORAX "generates fast and accurate plasma simulations to discover real-time nuclear fusion reactor control strategies," and "will be used to find the most efficient and safest methods to maximize nuclear fusion power generation in the future."