GOOGLE has been identified as a company focused on searches, social networks or smartphones to define themselves as a company A company focused on artificial intelligence. At that time the movement was seen as something strange, because AI was not part of the daily conversation and could only be tested in giant laboratories such as Mountain View. Now that has changed, a lot, and in Spain and the rest of the world, AI is part of our day to day.
Google has managed to climb to the current wave After OpenAi opened it with chatgpt. Gemini has become a worthy rival and, little by little, he is planted not only to the company led by Sam Altman, but also Microsoft and other industry giants. And in certain ads, such as today, it shows that they created the foundations for this technology.
Google has just announced Alphaevolve, An evolutionary coding agent based on large language models for discovery and algorithms optimization General purpose. That is, Alphaevolve combines Gemini creative capabilities with automated evaluators. This means that the system is able to verify its own answers and uses an evolutionary framework to improve them.
The free android
This is an important leap because one of the pegas that generative artificial intelligences currently have is their possibility of hallucinating. That is, that the answers that give are not true, and it is something that Google has suffered in its own meats when it began to implement the AI in the search engine.
The tool itself, Alphaevolve, has helped improve the efficiency of Google data centers, chips design and AI training processes. It has even helped to find new solutions to unsolved mathematical problems. Alphaevolve Verifies and qualifies the proposed programs by automated evaluation metrics before executing themwhich is very useful in areas such as mathematics and computer science.
Example of use of Alphaevolve in the design of a chip
The free android
Google itself has used the results thrown by Alphaevolve to improve data centers, hardware and software with latest generation algorithms. Google sets as an example that “Alphaevolve discovered a simple but extraordinarily effective heuristic to help Borg to orchestrate the vast Google data centers more efficiently. This solution, in production for more than a year, continually recovers, on average, 0.7 % of Google global computer resources.” That is, that Google is now capable of making more calculations in its facilities without increasing the consumption of resources.
Alphaevolve too has helped correct certain inefficiencies in the development of processors that Google integrates into its mobiles. For example, in the tensor chipsets, which are the ones that the company develops internally. Suggested improvements make their development easier, from the point of view of engineers. Improvements have also been seen in the suggestions that have made Training times of Gemini models are reduced by 1%.
At the moment this is not available to the public, but Google has created A form so that academics and experts can try it soon. The company’s intention is that This AI applies, in the first instance, to mathematics and computer problemssince it specializes in generating solutions for any problem whose solution can be described as an algorithm and automatically verify.
