# AI Models Rapidly Advancing in Theoretical Scientific Research Frontier AI models are solving complex theoretical problems across physics, statistics, and economics. - **Source:** [Science Latest](https://brifing.si/source/ebc20f22-8d1a-4ae4-ba4c-efcd0853ba19) - **Published:** 2026-10-14 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/2f9d0130-13e0-4939-b478-eecf0991124a ## Scientific Theory Under Transformation Neil Thompson from MIT highlights that artificial intelligence is rapidly disrupting theoretical scientific fields. His recent study indicates that disciplines like quantum physics, cosmology, and economic theory are facing significant shifts. Between mid-2024 and mid-2026, AI performance on theoretical problem-solving jumped from 2 percent to 35 percent, signaling a persistent upward trajectory. ## Rigorous Testing Methodology Employed Researchers Gabriel Manso and Claudio Mayrink Verdun tested OpenAI models including o1, o3, and various GPT-5 iterations. They utilized research papers published between December 2025 and May 2026 as ground truth. By blocking internet access and selecting post-training cutoff papers, they ensured models solved problems independently rather than retrieving memorized data. ## Expanding Beyond Pure Mathematics While the initial shock waves hit mathematics, the study suggests these impacts will ripple across all theory-heavy scientific domains. The research focused specifically on areas where AI can derive solutions without needing physical simulations or experiments. This evolution suggests that the automation of theoretical inquiry is becoming a broader scientific phenomenon. ## The Navier-Stokes Controversy OpenAI caused significant friction in September by announcing a solution to the Navier-Stokes fluid motion challenge. The announcement sparked controversy after an NYU mathematician accused the company of preemptively claiming credit for work involving other academic partners. This event highlights the growing tension between AI labs and traditional academic research communities. ## Crisis in Mathematical Methodology The influx of hundreds of AI-generated solutions has triggered a crisis within the mathematics community. Some experts argue that the process of solving a problem is as valuable as the final result itself. Consequently, there are growing calls to implement formal limits on AI usage to preserve the integrity of mathematical inquiry. ## Balancing Innovation and Verification Despite the rapid progress, AI models remain prone to errors, as evidenced by OpenAI withdrawing several solutions from its latest dataset. Researchers emphasize that human experts are still essential to verify outputs and connect complex dots. The ease with which a model can confidently present an incorrect answer necessitates rigorous human oversight. ## Future of AI-Driven Science Major AI laboratories are now ramping up scientific research, with some founding dedicated labs focused on AI for science. These initiatives aim to empower models to conduct real-world experiments alongside theoretical analysis. This shift could accelerate scientific progress, provided that the transition remains grounded in sound verification and expert-led conceptual frameworks. ## The Role of Human Expertise The transition toward AI-assisted science does not eliminate the need for human scientists. Experts remain critical for interpreting solutions and ensuring theoretical soundness. As AI takes over routine problem-solving, human researchers will likely pivot toward higher-level conceptualization and verification, ensuring that the integration of machine intelligence enhances rather than degrades scientific rigor. ## Sources covering this story - https://www.wired.com/story/ai-is-coming-for-science-next-watch-out/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._