Google shocks the world.

Google DeepMind's AlphaEvolve has just broken five Ramsey number records, the most challenging combinatorial problem in mathematics, by writing its own algorithm instead of waiting for humans to design it.

Google DeepMind's AlphaEvolve breaks records in the field of mathematics. Photo: Bloomberg .

Google DeepMind has just released a scientific report showing that AlphaEvolve simultaneously broke five Ramsey number sublimit records. This is one of the most difficult combinatorial problems in mathematics, with previous records standing for 6-20 years.

DeepMind CEO Demis Hassabis immediately shared the news, calling it "a major milestone for AI in mathematics." Turing Prize winner Yann LeCun also congratulated the research team.

The Ramsey number is a problem that has stumped even the greatest mathematicians. Paul Erdős, Terence Tao's teacher, once said that if aliens threatened Earth and humanity had to calculate the Ramsey number R(5,5) within a certain time limit or face extinction, the most reasonable choice for humanity would be to surrender. This statement reflects the extreme difficulty of the problem.

Specifically, AlphaEvolve improved the lower bounds of the five classic Ramsey numbers, including R(3,13) from 60 to 61, R(3,18) from 99 to 100, R(4,13) from 138 to 139, R(4,14) from 147 to 148, and R(4,15) from 158 to 159. Although each number only increased by one, increasing by one unit is more difficult than increasing the order of magnitude in many other problems. All five breakthroughs came from the same system.

Notably, AlphaEvolve doesn't solve problems in the conventional way. Instead of humans designing search algorithms and letting machines run them, AlphaEvolve reasons in its own algorithmic space. It uses the large-scale Gemini programming language to continuously improve its code, test it, score its performance, and retain the most efficient algorithms.

The DeepMind research team identified AlphaEvolve as having invented four different algorithmic groups for 28 R(r,s) values, ranging from random initialization methods to complex algebraic structures based on Paley graphs and quadratic residue graphs.

This isn't the first time AlphaEvolve has made waves. Previously, the system broke a 56-year record in matrix cores, optimized Google's data center operational schedules, and discovered simplified AI chip architecture patterns. When a system discovers algorithms to optimize its training process, the line between tool and creator becomes increasingly blurred.

Source: https://znews.vn/google-gay-soc-post1635566.html