A naukri.com initiative
VentureBeat
20h
14
Image Credit: VentureBeat
Meet AlphaEvolve, the Google AI that writes its own code—and just saved millions in computing costs
- Google DeepMind has unveiled AlphaEvolve, an AI agent that creates new computer algorithms and implements them in Google's computing systems.
- AlphaEvolve uses Google's Gemini language models and an evolutionary approach to refine and enhance algorithms automatically.
- The system has been integrated into Google's data centers, chip designs, and AI training systems, improving efficiency and solving complex mathematical problems.
- AlphaEvolve discovered an algorithm that boosts Google's computing resource efficiency by 0.7% and optimized Google's hardware design for Tensor Processing Units.
- It also improved the matrix multiplication kernel, cutting overall training time by 1% and reducing energy consumption for AI systems.
- AlphaEvolve developed new matrix multiplication algorithms and surpassed a mathematical record that had stood for 56 years.
- This AI system matched or improved state-of-the-art solutions in various mathematical areas, including geometry and number theory.
- AlphaEvolve uses an evolutionary approach to propose and refine code changes, exploring novel solutions that humans may not have considered.
- The system's potential extends to material sciences, drug discovery, and other fields requiring complex algorithms.
- Google DeepMind plans to launch an Early Access Program and envisions broader applications for AlphaEvolve beyond Google's infrastructure.
Read Full Article
Like
For uninterrupted reading, download the app