Google develops AI capable of creating and “fixing” software applications by editing executable code

According to Google, using artificial intelligence to “generate” executable code will reduce reliance on human programmers while avoiding mistakes made by them.

Started within the Alphabet X research division, the project codenamed Pitchfork is in the “generative artificial intelligence” category and aims to algorithmically generate images, videos, executable code and more using artificial intelligence. If it succeeds, the initiative could have major implications for Google, i.e. how developers create software application code.

Google’s plan is to help developers work more efficiently with a small team of software engineers, doing some of the work using artificial intelligence-based tools.

The new AI will be able to learn the programming “style” of engineers involved in various projects, generating new code in their style. Another important role will be in the quality validation process, with AI assisting in identifying bugs, possibly rewriting the code portion of the application in question.

Not so ambitious that it can generate executable code from scratch, the new AI is intended more to reduce the need for humans to create and update application code.

Google/Alphabet is not the only company making major advances in the Generative AI technology sphere, with Microsoft developing its own Copilot tool, capable of generating code suggestions by analyzing the creative process of software developers. Apparently, software engineers using this tool can auto-generate up to 40% of the new code created over the course of a working day, dramatically reducing the turnaround time of the projects they work on.