Neil Thompson presents AI's potential to revolutionize multiple scientific fields
Researchers have found a significant leap in AI capabilities, with top models going from correctly solving about 2% of theoretical problems to around 35%.
A leading expert on computing and artificial intelligence has made a compelling argument that several scientific fields are on the cusp of being revolutionized by AI.
Neil Thompson presented his findings at MIT last week, highlighting the potential for AI to disrupt areas such as probability, statistics theory, economic theory, high-energy theory, gravitation and cosmology, and quantum physics. His research suggests that these fields will be significantly impacted by the rapid advancements in frontier models of artificial intelligence.
According to Thompson's study, between mid-2024 and mid-2026, there was a notable leap in AI capabilities, with top models going from correctly solving about 2% of theoretical problems in these areas to around 35%.
Researchers at Thompson's lab have been investigating how advancements in computer science influence innovation and society. Their work involves analyzing the impact of AI on various fields, including mathematics. The study's findings suggest that recent developments in AI capabilities are not limited to math alone.
To test the abilities of top models, the researchers isolated them from external influences by blocking internet access. This ensured that the models relied solely on their internal workings and did not consult pre-existing solutions or online resources. In doing so, they discovered that these models could solve complex problems independently, without relying on training data provided during development.
The study's results indicate a significant leap in AI capabilities between mid-2024 and mid-2026. Top models showed a notable improvement in solving theoretical problems, rising from around 2% to approximately 35%. This rapid progress has sparked concerns that similar breakthroughs may soon occur in other math- and theory-heavy areas of scientific inquiry.
The controversy surrounding OpenAI's breakthrough in math has spilled over into other areas of scientific inquiry, sparking concerns about the role of artificial intelligence in driving progress.
A mathematician at New York University accused OpenAI of "gazumping" them after learning of their rapid progress in solving longstanding math problems. This accusation highlights the high stakes involved in AI research and the potential for rival groups to be left in the dust. The incident has added fuel to a broader debate about the use of AI in scientific inquiry.
The pace of innovation shows no signs of slowing, with OpenAI and other researchers announcing hundreds more solutions to math problems. This rapid progress is being mirrored in other areas of science, where theoretical advances are driving practical breakthroughs.
Researchers are now investigating how to integrate artificial intelligence into real-world experiments, allowing for a more hands-on approach to scientific inquiry alongside theoretical analysis.
This shift in focus is crucial, as AI may soon be capable of solving complex problems independently, potentially rendering human expertise less necessary. However, experts stress that this does not mean the elimination of human involvement altogether, but rather a redefinition of roles within the scientific community. As Manso notes, even with AI's potential to generate solutions, humans will still be essential for verifying and interpreting results.
Facts based on reporting originally published by Wired.
You may republish this story, in full or in part, if you credit News Central Site and link to it (licence CC BY 4.0). Photos are not included.
