Physical Superintelligence (PSI), a startup focused on AI-enhanced scientific discovery, cracked the Aaronson-Ambainis conjecture using Anthropic’s Fable 5.2 model. This breakthrough advances understanding in quantum computing and exemplifies AI’s growing role in transforming theoretical research into practical technologies.
- Unreleased Anthropic AI model helps solve long-standing quantum conjecture
- PSI focuses on bridging theoretical physics with practical technology
- AI’s role in science evolving alongside validation and experimental methods
What happened
PSI employed an unreleased version of Anthropic’s Fable 5.2 AI to solve the Aaronson-Ambainis conjecture, a significant problem in quantum computing that had remained open for nearly two decades. The solution was verified and formalized through advanced mathematical coding using Lean, earning credibility from the original problem’s researchers and academic collaborators. This development marks a critical milestone in using specialized AI models to tackle complex scientific challenges.
The breakthrough not only addressed a key theoretical question about quantum query complexity, but it also closed a potential pathway for quantum computers to outperform classical counterparts on certain problems. PSI’s approach relies on integrating multiple AI tools and memory systems tailored to explore scientific problems from various angles, showing how AI can augment research depth and breadth in ways traditional methods cannot.
Why it matters
Solving the Aaronson-Ambainis conjecture with AI highlights a shift in scientific discovery processes, where machine intelligence can contribute to fundamental theoretical advances. This demonstrates the potential for AI to drastically accelerate problem-solving in fields historically reliant on human intuition and iterative experimentation. PSI’s success emphasizes how AI can become a core asset in scientific research, potentially rendering traditional paper publishing and manual proofs less central.
Moreover, PSI’s business model combines this research capability with immediate industrial applications, such as optimizing data centers by simulating physical variables like heat and power consumption. This dual approach—leveraging AI to generate theoretical insights and apply them practically—signals a new paradigm for startups aiming to commercialize cutting-edge science efficiently while continuing to push technological frontiers.
What to watch next
PSI aims to further integrate AI-driven theoretical physics research with experimental validation to translate scientific insights into usable technologies. Their efforts include building iterative feedback loops where AI-generated hypotheses are tested and refined to exploit physical realities, potentially enabling new device creation or groundbreaking materials. Tracking how PSI evolves its platform and partnerships will be key to understanding AI’s impact on real-world scientific innovation.
The wider scientific community and AI startups like Discovered Materials are also advancing challenges and frameworks to address complex real-world problems, such as superconductivity. How AI models adapt from abstract mathematical domains to nuanced physical sciences will significantly influence the future landscape of automated discovery. PSI’s progress will likely set benchmarks and standards as ‘neolabs for science’ become more prevalent, shaping norms around transparency, validation, and collaboration in AI-powered research.