AI Breakthroughs Solve Centuries-Old Math Problems, Sparking Security Fears
Translated & summarized from Globes by baba
OpenAI and Double AI have announced breakthroughs in solving long-standing mathematical problems, including the longest path problem and versions of Hilbert's tenth and Riemann hypotheses. These AI-driven solutions, achieved at low cost, have alarmed mathematicians who fear a devaluation of scientific research and a decline in human mathematicians. Concerns also extend to national security, with potential implications for encryption and cybersecurity systems.
The story in 6 lines · by baba
- OpenAI and Double AI have solved complex, century-old mathematical problems using artificial intelligence.
- Mathematicians warn that AI's rapid problem-solving devalues scientific research and discourages future scholars.
- The breakthroughs raise serious national security concerns regarding encryption and cybersecurity.
- Double AI claims to have solved the longest path problem for under $2,000.
- OpenAI has published solutions to problems like Hilbert's tenth and a version of the Riemann hypothesis.
- Fields Medalists have cautioned against using research problems as AI benchmarks.
Two artificial intelligence companies, OpenAI and Professor Amnon Shashua's Double AI, have announced significant breakthroughs in solving complex mathematical problems that have stumped mathematicians for decades, and in some cases, over a century. These discoveries have sent ripples through the scientific community, raising concerns about the future of mathematics as a profession and posing serious security risks, including potential threats to encryption and national security assets.
Double AI, founded to tackle difficult scientific problems using AI agents and algorithmic verification, announced it had solved the "longest path problem" in graph theory, a challenge that remained unsolved for 38 years. The company presented a new algebraic formula that advances a mathematical constant and claims to have solved it for less than $2,000, a fraction of the cost incurred by major AI firms for similar problems. This achievement was reportedly accomplished using AI agents and a novel approach involving artificial brainstorming among these agents, with rigorous algorithmic checks to prevent hallucinations.
Meanwhile, OpenAI released a repository of 722 studies demonstrating solutions to numerous open problems in mathematics and computer science. Among these are solutions to a version of "Hilbert's tenth problem" from 1900, proving no computer algorithm can automatically determine if a given equation has rational solutions, and a proof regarding "Catalan's constant," a number mathematicians have studied for nearly 200 years. OpenAI also presented progress on the "Riemann hypothesis," a famous Millennium Prize Problem, by proving a weaker version of it.
These advancements come amid growing unease within the scientific community. Twenty-five Fields Medalists, the highest honor in mathematics, warned in an open letter about the trend of using research problems as benchmarks for AI models. They argue that the rapid, "mass production" of solutions by AI leaves insufficient time to understand the underlying theories, connect them to existing knowledge, or credit prior research, potentially devaluing the scientific process and discouraging future researchers.
OpenAI acknowledged these concerns and has established an independent advisory group, including prominent mathematicians and physicists, to guide the publication of AI results and maintain research standards. However, some, like Fields Medalist Timothy Gowers, express deeper worries about the long-term impact on the human scientific community, fearing that if machines solve problems more effectively, fewer young people will pursue careers in mathematics, leading to a potential decline in human understanding and transmission of knowledge.
Mentioned
