- DeepMind's AlphaEvolve Develops the Fastest Matrix Multiplication Algorithm Since 1969
- Era Where AI Finds Better Algorithms Independently, Changing Roles of Developers
[Unblock Media] An era has dawned where AI autonomously discovers and designs better algorithms. The new AI system, AlphaEvolve, by Google DeepMind, is gaining attention for its potential to change the landscape of technology.
On the 16th, Mario Nawfal concisely summarized the practical achievements and potential of AlphaEvolve on his X (formerly Twitter), sparking discussion. This system symbolizes the possibility of 'meta intelligence,' where AI designs AI, rather than being merely an optimization tool.
AlphaEvolve has achieved the greatest performance improvement in matrix multiplication algorithms since 1969, going beyond mere theoretical possibilities. It has also shown meaningful results in various practical instances, such as improving data center task scheduling, enhancing chip design efficiency, and optimizing the speed of large language models (LLMs).
Rather than replacing developers, AlphaEvolve changes their role from 'designer to collaborative partner.' Nawfal emphasized, “Developers previously wrote code line-by-line, but now they can work with AI that explores millions of solutions.” This change amplifies developers' capabilities while creating a new kind of software design environment.
The algorithm-designing capability of AlphaEvolve can be utilized in scientific research, drug development, supply chain optimization, financial systems, and even the evolution of AI itself. Nawfal described it as “a thinking tool that can solve difficult problems faster,” rather than a mere tool.
The recent release of AlphaEvolve by DeepMind is seen as a signal of ‘AI-designed intelligence,’ where AI designs and improves its own intelligence. In his final post, Nawfal added: “AlphaEvolve shows us a new future where AI goes beyond following commands, to thinking, improving, and creating independently.”