Import AI 480: Self-sustaining space economies; AI for science progress, and Meta builds perfect hands for a superintelligence

What happened

Physical Superintelligence, an AI science startup, used an Anthropic AI model to prove the Aaronson-Ambainis conjecture in quantum computing. This conjecture, open since 2008, concerns the conditions under which quantum computers can outperform classical ones. The AI system produced a proof within three days, which was verified by the Lean proof assistant. Meanwhile, Meta Reality Labs introduced Tactus, a tactile sensing glove designed to collect detailed hand manipulation data for future AI applications.

Key details

• The Aaronson-Ambainis conjecture was first proposed in 2008.
• Physical Superintelligence used a non-public Anthropic AI model for the proof.
• The AI system improved the proof twice before final verification.
• Lean, a proof assistant, checked and accepted the proof.
• Meta's Tactus glove features capacitive tactile sensors and IMU-fused pose tracking.

Why it matters

Researchers in quantum computing now have a longstanding conjecture resolved, potentially advancing the field and influencing future research directions. Developers of AI systems and robotics gain access to new, high-quality hand manipulation data from Meta's Tactus glove, which can improve the dexterity and effectiveness of AI agents in physical tasks. These developments provide scientists and engineers with new tools and results that can accelerate progress in both theoretical and applied AI research.