Google is working on artificial intelligence who himself writes, corrects and updates the codes.
According to the website insiderthe project, nicknamed Pitchfork, started in the X research division of major tech Alphabet and has already moved to the Labs group, which houses long-term programs such as virtual and augmented reality.
Based on machine learning, Pitchfork integrates Google’s pursuit of so-called generative artificial intelligence, which uses image creation algorithms and video, for example.
While this is an impressive piece of technology, it could impact the future of code developers, both inside and outside the company.
According to documents seen insider, the system was created to teach code as well as write and rewrite itself. The tool must be able to design in order to learn programming styles and write new features based on this knowledge.
The original goal was to create a tool that could update the codebase of a programming language. python programming from Google for newer versions.
“The idea was: how do we move from one version to another without hiring all these software engineers?” an unnamed employee said.
However, over time the goals have changed. Now Google needs a general purpose system that can still reduce the need for people to write and update code without sacrificing feature quality.
“The team is working closely with the research team,” the official said. “They are working together to explore different use cases to help developers.” According to him, the project is led by Olivia Hatalsky, a longtime Google employee.
Artificial intelligence on the rise
Google is working on several AI projects. OUR deep mind, a subsidiary of Alphabet, already has an AlphaCode system that uses artificial intelligence to generate codes. However, today the program is focused on coding at a competitive level.
There are reports that Google is working on a platform that offers code snippets based on what developers enter. It is very similar to Copilot from GitHub, which is owned by Microsoft.
According to website Bloomberg, developers using Copilot can generate up to 40% of the code. GitHub hopes to double that number by 2027.