Google’s artificial intelligence unit DeepMind has designed a new model that could revolutionize the field of algorithm development. The system called “AlphaEvolve” not only produces software; it can also analyze existing algorithms and make them more efficient. Google has started to actively use this technology within the company.

Google designs an artificial intelligence that can invent algorithms

AlphaEvolve is basically built on Google’s Gemini large language model infrastructure. However, it goes far beyond a classic chatbot. The solutions produced are analyzed and compared by the system’s evaluation mechanism.

Thus, as in natural evolution, the most successful solutions survive and more advanced algorithms are reached in each cycle. This evolutionary approach will allow the system to solve complex and difficult-to-solve programming problems more effectively.

While systems such as AlphaFold, previously developed by DeepMind, focus on certain areas, AlphaEvolve has a much more flexible structure and can work on any algorithmic structure. This multi-purpose design allows the system to take on a wide range of tasks from software engineering to hardware development.

Google has started to actively use AlphaEvolve in its own infrastructure. One of the first application areas was the task scheduling systems used in the company’s data centers. The improvements made to the clustering system called Borg with the changes suggested by AlphaEvolve saved 0.7% of computing resources.

This rate means a profit of billions of dollars when viewed on Google’s scale. So what do you think about this? You can share your views with us in the comments section below.