During a recent open house visit to General Fusion in Vancouver, I had the unique opportunity to tour the Nuclear Fusion facility, getting an up-close look at their advanced equipment, research areas, and the passionate team driving their mission. As a tech enthusiast and futurist it was great to be able to see and even touch some of the tech that might bring one of the greatest energy revolutions to the planet. During a Q&A session where the CEO and the founder responded to numerous queries from the 100 or so people in attendance, I asked about the possible role that AI might have in solving some of the intricate physics challenges in fusion. The response was pretty shocking to me. There was simply skepticism regarding AI’s potential in providing value in Fusion beyond helping to parse some data and provide minor efficiency gains. 

But, after interviewing enough tech leaders and experts, and knowing that AI has made significant strides in other complex scientific realms, I wondered if AI actually was a fundamental key that could be used to unlock the next energy frontier even with the expressed skepticism. Artificial Intelligence’s importance in the Fusion Industry was already brought up briefly in my conversation with Andrew Holland, the CEO of the Fusion Industry Association (Podcast #106), but a deeper dive is what we will focus on here. So let’s first look at where AI has already been useful in some other scientific endeavors outside of the fusion industry. 

Examples of AI Making Leaps in Complex Scientific Problems

1. Alphafold and Protein Folding:

The world of biology was recently astounded by the capabilities of DeepMind’s Alphafold in solving protein-folding problems — a conundrum that puzzled scientists for decades. By predicting protein structures with remarkable accuracy, Alphafold has the potential to revolutionize the understanding of biological systems and accelerate drug discovery.

2. Astronomy and AI – Identifying Exoplanets: Using AI, researchers have sifted through data from NASA’s Kepler Space Telescope to identify exoplanets — planets that orbit stars outside our solar system. This method has not only increased the efficiency of discovery but has also found previously overlooked planets in existing datasets.

3. Discovering Novel Materials with AI:
AI has played a crucial role in predicting and discovering new materials. Scientists have been using machine learning models to predict material properties and behaviors, which can significantly cut down the trial-and-error (and often expensive) approach in material science. One such example is the discovery of a novel metallic glass, which would have been a herculean task using traditional methods.

4. Geological Exploration and Natural Resources: AI is playing a pivotal role in the identification and extraction of natural resources. By analyzing complex geological data, AI assists in identifying potential oil reservoirs, mineral deposits, and ensuring the efficient extraction of these resources with minimal environmental impact.

5. Archaeological Discoveries:
AI, specifically machine learning, has been used in archaeological endeavors to discover ancient lost civilizations. In a recent project, researchers utilized AI to analyze satellite images and identify buried structures and settlements. For instance, AI helped in uncovering over 3,000 ancient human settlements in the Midwest U.S.

6. AI-Powered Robotic Laboratory for Drug Design:
The “robot scientist” named Eve was designed to speed up the drug discovery process. It can automate the early stages of drug design by hypothesizing to explain observations, devising experiments to test these hypotheses, conducting the experiments using laboratory robotics, and then interpreting the results.

7. Lawrence Livermore Lab and Fusion Breakthrough:

CogSim, a Cognitive Simulation tool developed at Lawrence Livermore National Laboratory (LLNL), was instrumental in advancing inertial confinement fusion (ICF) research by providing highly accurate predictions for fusion experiment outcomes. By harnessing advanced AI and machine-learning techniques, CogSim analyzed extensive data from previous fusion experiments and seamlessly integrated this information with hydrodynamics simulation codes, notably HYDRA. This enabled researchers to gain a more quantitative and nuanced understanding of the myriad complex interactions occurring during fusion. The capabilities of CogSim culminated in its accurate prediction of an ICF experiment in December 2022, where a groundbreaking net energy gain of about 150 percent was achieved, underscoring a monumental step towards the broader goal of controlled fusion ignition and the potential for fusion as a viable energy source.

AI Offers Unique Perspectives to Scientific Problems

How is all of this possible?

The intrinsic value of Artificial Intelligence (AI) lies in its ability to provide unique perspectives to tackle complex problems across various scientific and industrial domains. Unlike human cognition, which is often tethered by inherent biases and a linear approach, AI, with its computational prowess, unveils multidimensional solutions, accelerating problem-solving and innovation. The examples cited earlier – from Alphafold’s breakthrough in protein folding, astronomical discoveries, to enhancing geological and archaeological explorations – showcase AI’s capability in data analysis, pattern recognition, and predictive modeling, which often surpasses human capability. The amalgamation of AI with traditional scientific methods has not only expedited the pace of discoveries but also unearthed new avenues and hypotheses, pushing the boundaries of human knowledge. Its effectiveness in tackling intricate problems underpins a promising future, where AI, fused with human intelligence, can drive an era of unprecedented technological advancement and understanding. The scientific papers and reports described earlier further support the transformative potential of AI in offering novel, invaluable perspectives in addressing complex, multi-faceted problems. Here are some examples of AI providing unique insights in scientific perspectives:

• Data-Driven Insights:
  • “Big Data: Astronomical or Genomical?” by Z. D. Stephens et al. discusses the application of big data analytics in genomics to drive insights that were previously unattainable.
  • Link to paper
• Automated Hypothesis Generation:
  • “Automated hypothesis generation based on mining scientific literature” by Don R. Swanson demonstrates how AI can help in generating hypotheses by analyzing scientific literature.
  • Link to paper
• Predictive Modeling:
  • “Predictive modeling of the hospital readmission risk from patients’ claims data using machine learning: a case study on COPD” by Somashekhar et al. discusses the use of predictive modeling in healthcare.
  • Link to paper
• Optimization:
  • “A review of machine learning and optimisation methods in clean energy generation and management systems” by Yusuf Sanni et al. explores how AI optimization is aiding in clean energy generation.
  • Link to paper
• Simulation:
  • “Artificial intelligence in the simulation of human behavior” by M. Arulampalam et al. examines how AI simulations help in understanding complex human behavior.
  • Link to paper

• Pattern Recognition:
  • “The modern discovery of ancient civilizations in archaeological remote sensing” by Giardino and Haley talks about the use of pattern recognition in archaeology.
  • Link to paper
• Generative Models:
  • “Generative Adversarial Nets” by Ian Goodfellow et al. delves into generative models and their potential in various fields.
  • Link to paper
• Cross-Disciplinary Insights:
  • “Cross-disciplinary data science: The critical role of a data scientist in modern research” by Jacobs and Wiraatmadja discusses how data science, driven by AI, is facilitating cross-disciplinary research.
  • Link to paper

These papers provide a glimpse into the diverse ways AI is offering a different perspective and new approaches to various problems.

The Future Role of AI in Fusion Energy

Despite the skepticism I encountered at General Fusion, the undeniable advancements AI has brought to various scientific domains raise an essential question, not if, but how will AI shape the future of fusion energy?

Recent reports suggest that AI has already had a profound impact in fusion, namely in optimizing plasma control, diagnostics, and materials science related to fusion. And referring back to my podcast episode, the CEO of the Fusion Industry Association, AI had already played an indispensable role in one of the fusion member companies’ experiments and advanced their timeframes and enabled them to do more experiments. And even some of the most powerful computers in the USA are already being used to model fusion dynamics, so a different perspective states that AI has been and will continue to be very important to realize fusion energy.

Given these insights and developments, one might wonder if dismissing AI’s potential in fusion might be premature. Perhaps it’s time to re-evaluate and embrace the symbiotic relationship between fusion energy and AI, harnessing their combined potential to reshape our energy future and unlock fusion energy once and for all.

Conclusion

In light of the advances demonstrated by AI across various scientific fields, it’s compelling to consider its potential in catalyzing fusion energy research. The recent success story from Lawrence Livermore National Lab, where CogSim’s AI-driven insights significantly bolstered inertial confinement fusion, serves as a robust indicator of this potential. Despite some skepticism, the unfolding narrative suggests that a synergistic integration of AI could indeed be a pivotal factor in overcoming complex challenges inherent to fusion energy. As AI continues to evolve, exploring its application in fusion could not only accelerate progress towards viable fusion energy solutions but also open new horizons for innovative, interdisciplinary research, marrying the cutting-edge domains of artificial intelligence and nuclear fusion. Such advancements could inevitably play a crucial role in shaping a sustainable energy landscape for the future, thus warranting a deeper investigation and greater discourse within the fusion and AI communities alike.