Tag with "Mathematics"

OpenAI's claims about GPT-5 solving unsolved mathematical problems have been challenged by experts. While the AI found existing solutions in literature, researchers argue it didn't 'solve' anything new. The controversy highlights ongoing discussions about AI's true capabilities versus marketing hype.

Mathematician Terence Tao collaborated with AI model GPT-5 Pro to solve a complex differential geometry problem in just 11 minutes. The breakthrough highlights AI's growing role in mathematical research and demonstrates a new paradigm for human-AI collaboration.

ByteDance's Seed-Prover, an automated theorem-proving system, successfully tackled four complex problems at IMO2025, showcasing AI's potential in advanced mathematics. The system demonstrated speed, rigor, and innovation, though its model weights remain unreleased.

Google DeepMind's Gemini Advanced AI model has achieved a historic milestone by winning a gold medal at the 2025 International Mathematical Olympiad (IMO). The AI solved five complex problems using natural language processing, demonstrating advanced reasoning capabilities that could redefine AI's role in academic and commercial applications.

Huawei has unveiled its 'Ascend + Pangu Ultra MoE' system, capable of solving advanced math problems in just two seconds without relying on GPUs. The breakthrough highlights Huawei's advancements in domestic computing power and AI model training efficiency.

Google DeepMind unveils AlphaEvolve, an AI system that broke a 56-year-old math record and optimized Google's data centers and chip designs. The self-evolving AI also improved its own training efficiency by 23% and solved complex problems like the "kissing number" in 11-dimensional space.

**Summary** 1. Shanghai AI Lab has recreated OpenAI’s Olympiad tool, LLaMA-O1, open-sourcing it for the world. 2. The project harnesses Monte Carlo Tree Search, reinforcement learning, and AlphaGo Zero architecture. 3. LLaMA-O1 excelled in tackling mathematical Olympiad problems, outperforming most closed-source models. 4. The project uses Google’s Gemma2 as a base for training. 5. The mysterious GitHub account SimpleBerry is behind the LLaMA-O1 code, while another project, O1-Journey, from Shanghai Jiao Tong University, is also progressing fast.