Google Deepmind's AlphaProof Nexus has autonomously solved nine open Erdős problems, including two unsolved for 56 years, at a cost of just a few hundred dollars per problem in computational inference.
The system represents a significant advancement in AI-assisted mathematical discovery. Among the nine problems solved, two had eluded mathematicians since 1968, demonstrating the potential of machine learning to tackle longstanding theoretical challenges.
AlphaProof Nexus differs fundamentally from competitors like OpenAI's natural language approaches. The system uses the Lean compiler to automatically verify each proof step, eliminating ambiguity and ensuring mathematical rigor. This verification layer adds computational overhead but guarantees the validity of solutions.
The Erdős problems—named after prolific mathematician Paul Erdős—represent some of mathematics' most difficult open questions. Their resolution, even partially, indicates progress in bridging human mathematical intuition and machine-driven problem solving.
However, results come with caveats. The overall success rate stands at just 2.5 percent, meaning the system fails far more often than it succeeds. This low conversion rate reflects the fundamental difficulty of the problems and the challenges in scaling AI reasoning to complex mathematical domains.
Inference costs of a few hundred dollars per problem remain economical compared to human mathematician time, but the low success rate raises questions about practical applicability. Researchers would need to attempt many problems to achieve one solution.
The breakthrough underscores AI's evolving role in scientific research. Rather than replacing mathematicians, systems like AlphaProof Nexus may serve as tools for exploring solution spaces and identifying promising research directions. The automatic verification through Lean addresses a critical concern in AI-generated mathematics: ensuring correctness without human oversight.
Google Deepmind's work builds on earlier AlphaProof systems and reflects broader industry investment in AI reasoning capabilities. The results suggest that specialized AI architectures, combined with formal verification systems, can contribute meaningfully to theoretical mathematics—an area previously dominated entirely by human intellect.
Open-weight AI companies—those releasing freely available models—are attracting major acquisition interest from tech giants. The trend reflects growing capital investment in the business model of distributing AI models at no cost.
Google Deepmind has upgraded its Co-Scientist AI system to autonomously plan experiments, operate lab equipment, and publish scientific papers. The Gemini-based multi-agent platform demonstrated experimentally validated results across materials science, chemistry, and medical AI development.
Uber's weekly AI agent requests have grown nearly tenfold since February, yet the company has held spending flat since April after exhausting its entire 2026 AI budget in Q1.
A recent paper shows artificial intelligence often diagnoses and treats patients better than human physicians. The findings are prompting difficult conversations within the medical community about the profession's evolving role.