AI is inventing materials that don't exist yet | with the Faraday Institution

Overview

This Royal Institution talk explores how AI is revolutionizing materials science for next-generation batteries. Professor Aron Walsh (Casper AI) details AI's role in exploring vast chemical spaces to design novel materials, Dr. Mona Faraji (Alan Turing Fellow) explains AI's use in analyzing complex sensor data to predict battery degradation and lifespan, and Dr. Sam Cooper (Polaron) showcases AI's application in understanding and designing material microstructures. The presentations highlight AI's transition from data analysis to generative design and co-scientist roles, accelerating discovery from atomic to system scales.

Key takeaways

AI models like Chameleon and Crystallize are enabling the generation of novel materials and acting as 'co-scientists' by automating literature review, hypothesis generation, and experimental design.
The chemical space of possible materials is astronomically large (over 10^100 combinations), making AI essential for efficient exploration and discovery of battery components.
Dr. Mona Faraji's work uses AI to analyze complex sensor data and numerous influencing factors (user habits, chemistry, manufacturing) to predict battery degradation and remaining useful life.
Dr. Sam Cooper highlights AI's role in characterizing and generating 3D microstructures from 2D images, enabling the design of optimal electrode architectures for faster charging and longer-lasting batteries.
Physics-informed AI and hybrid modeling approaches are crucial for battery science, merging data-driven insights with fundamental physical principles to improve model accuracy and transferability.
AI is transforming materials science by automating routine tasks, enabling exploration of vast design spaces, and orchestrating complex experimental campaigns, significantly accelerating the development of next-generation batteries.

Chapters

0:00 Introduction: From Davy and Faraday to AI in Battery Science

Dr. James Luhoo introduces the historical context of scientific discovery at the Royal Institution, referencing Humphrey Davy and Michael Faraday.
Highlights the evolution of electricity from a curiosity to a scientific tool, drawing parallels with AI's current role.
Explains the Faraday Institution's focus on energy storage and the importance of batteries in modern infrastructure and the low-carbon transition.

7:16 The Challenge of Battery Degradation and the Role of AI

Details microscopic processes within batteries, such as ion shuttling, swelling, cracking, and film growth, leading to degradation.
Explains how AI can piece together information from disparate sources to understand these complex degradation mechanisms.
AI enables the generation of new material classes by analyzing redox couples and designing novel materials, accelerating discovery.

8:43 AI at the Atomic Scale: Designing New Materials with Professor Aron Walsh

Professor Aron Walsh discusses AI's application in exploring the vast chemical space for battery materials.
Explains how encoding centuries of chemical information into numerical vectors enables machine learning analysis.
Introduces generative AI and 'co-scientists' like Crystallize, which can review literature, propose hypotheses, and interact with robotic labs.

22:17 The Immense Chemical Space and the Quest for Ideal Batteries

Illustrates the astronomical number of possible material combinations (over a googol) compared to atoms in the universe.
Discusses the limitations of traditional simulation and experimentation in exploring this space.
Highlights desired future battery characteristics: sustainability, increased power, solid-state electrolytes, and reduced development time.

27:33 Generative AI and Co-Scientists for Materials Discovery

Explains generative AI's ability to create novel outputs (like images or materials) based on prompts and learned context.
Introduces Chameleon, a text-to-material model, as a proof-of-concept for AI-driven material generation.
Discusses 'co-scientists' that leverage reasoning models and scientific toolkits (literature search, simulation, robotic labs) to act as AI researchers.

36:57 Cusp AI and Mira: A Materials Intelligence Engine

Details the founding of Cusp AI, a startup focused on AI for materials design, and its 'materials intelligence engine'.
Introduces Mira, a co-scientist agent with a text interface for querying material properties.
Addresses the challenges of AI agents exploring new chemical spaces and the need for human oversight and deeper integration with labs.

41:47 AI for Battery Systems: Forensic Investigation of Degradation with Dr. Mona Faraji

Dr. Mona Faraji introduces AI's role in understanding battery performance and degradation at the system scale.
Explains how batteries degrade due to mechanical stress, particle cracking, lithium plating, and dendrite formation.
Identifies numerous factors influencing degradation: user habits, chemistry, cell format, manufacturing processes, and environmental conditions.

52:06 Predicting Battery Lifespan and the Power of Data-Driven AI

Highlights the challenge of estimating remaining useful life (RUL) for batteries due to complex degradation factors.
Emphasizes the vast amount of sensor data (temperature, current, voltage) available for analysis.
Explains how AI, particularly machine learning and deep learning, can identify underlying patterns linking degradation factors to capacity loss.

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, The Royal Institution.