We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Hacker News·21d ago·5 min read
Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel
TL;DR
A new AI system called Star Fleet solves complex open mathematics problems using Lean 4 and multiple GPT-5.6 instances. It tackles Erdős Problem 123, which questions the sum of distinct coprime integers.
✦ Why It Matters
Engineers can explore using parallel AI agents to tackle complex mathematical problems in their own fields.
Key Takeaways
- Utilize Lean 4 for formal proof verification in your mathematical projects.
- Consider parallel processing with AI models for complex problem-solving.
- Leverage large databases of mathematical premises to enhance AI capabilities.
- Explore the implications of AI in automating theorem proving and research.
How It Works
Star Fleet operates by distributing tasks across multiple GPT-5.6 instances, each focusing on a different Erdős problem. It employs a unique approach by working on homogeneous exponent levels, which simplifies the divisibility conditions of the integers involved.
This allows the system to generate primitive subset sums while maintaining the necessary properties for formal proofs. The integration of advanced computational resources, including GPUs and a comprehensive database of Lean 4 premises, further enhances its efficiency and accuracy in solving complex mathematical problems.