{
  "id": 14512930,
  "title": "AI Came for Math. Now It’s Chasing the Rest of Science",
  "url": "https://urgent.news/2026/10/14/ai-came-for-math-now-its-chasing-the-rest-of-science-14512930",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-14T18:15:00.000Z",
  "source": {
    "name": "Wired Business",
    "slug": "wired-business",
    "url": "https://www.wired.com/story/ai-is-coming-for-science-next-watch-out/"
  },
  "original_language": "en",
  "account": "Last week, MIT hosted a presentation by Neil Thompson, an expert on computing and artificial intelligence progress. In a brief half-hour talk, Thompson convincingly argued that several scientific disciplines are poised for disruption by AI. As mathematicians grapple with the impact of AI on their field, Thompson's latest research suggests that fields like probability, statistics, economic theory, high-energy theory, gravitation, cosmology, and quantum physics could soon face similar challenges. According to Thompson's study, between mid-2024 and mid-2026, the most advanced AI models improved from solving approximately 2% of theoretical problems in these fields to around 35%. This rapid progress appears to have no signs of slowing down, with Thompson predicting that \"a lot of theory questions\" will soon be transformed by AI's capabilities. To conduct the study, Thompson, alongside colleagues Gabriel Manso and Claudio Mayrink Verdun, analyzed problems presented in research papers published between December 2025 and May 2026. They tested the ability of various OpenAI models—o1, o3, GPT-5, GPT-5.1, GPT-5.4, and GPT-5.5—to solve these problems. Crucially, the models had to devise solutions independently, without relying on pre-existing answers from the training data. By blocking internet access, the researchers ensured the models couldn't simply look up solutions online. The study focused on fields where AI could generate answers without requiring simulations or experiments, although Thompson acknowledges that the results may have implications for other theory-heavy areas of scientific inquiry. In recent news, OpenAI announced that an unreleased model had solved a significant mathematical challenge known as Navier-Stokes, a problem concerning fluid motion behavior. This revelation sparked controversy after a mathematician from NYU accused OpenAI of \"gazumping\" them after learning of their progress. Since then, OpenAI and other companies have announced numerous solutions to long-standing math problems, creating a crisis in the field. Some mathematicians have called for restrictions on AI use, emphasizing that the process of solving problems often contributes as much as the final answer itself. Thompson believes that advances in AI could both positively and negatively impact science, noting that theoretical breakthroughs can drive practical advancements. He also highlights that greater AI utilization may lead to more progress through theoretical analysis. Furthermore, prominent AI researchers have established new labs dedicated to AI for scientific purposes, while established AI labs are expanding their scientific research initiatives. These efforts aim to explore how AI can assist in both theoretical analysis and real experiments. While AI might diminish the need for theorists to solve key problems, experts argue that human involvement remains crucial for generating ideas, connecting dots, and verifying the soundness of AI-generated solutions.",
  "summary": "AI may soon be the default way to solve theoretical problems in economics, high-energy physics, and other fields, MIT researchers tell WIRED.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Wired Science",
        "title": "AI Came for Math. Now It’s Chasing the Rest of Science",
        "url": "https://urgent.news/2026/10/14/ai-came-for-math-now-its-chasing-the-rest-of-science",
        "published": "2026-10-14T18:15:00.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}