Google DeepMind introduced AlphaEvolve, an AI system combining language models with evolutionary algorithms for mathematical discovery.
AlphaEvolve contributed to resolving the Kissing Number problem, raising the lower bound from 592 to 593 in an 11-dimensional space.
The AI system enhanced Strassen's matrix multiplication algorithm, reducing the number of operations required from O(n^3) to O(n².81), leading to significant performance gains in AI and big data processing.
AlphaEvolve utilizes reinforcement learning and evolutionary search to discover more efficient methods for matrix multiplication, improving algorithms via mechanisms like mutation, crossover, and natural selection.