Databases
Cloud giant says humans remain accountable, even when code gets an assist from the machines
Google is encouraging its database developers to lean "heavily" on AI coding tools as it ramps up contributions to open source projects such as PostgreSQL.Earlier this year, Google announced a raft of new contributions to PostgreSQL, the open source database that has become a popular RDBMS for developers building new applications in the cloud.Sailesh Krishnamurthy, VP of Databases, Google Cloud, told The Register that the company was using AI coding tools to accelerate its contributions to open source database systems, although each developer remains responsible for their individual contributions.
"We do encourage folks to use AI heavily ," he said. "We are seeing huge amounts of productivity improvements internally. In the end, we have individual engineers take accountability for our contributions. Whether you have a piece of code that is completely drafted by AI, or not even part of what you're pasting into your development environment, you have a whole spectrum where AI is used in different places. Either way, the accountability remains on behalf of the person who's done it."
AI coding tools can be especially suited to developing contributions to open source projects because the codebase is publicly available and has been used to train the generative models, he said."That's how models have a better sense of the code, as opposed to many proprietary pieces of code, which are inside the firewall."PostgreSQL was designed to be extensible. As such, it can be a system well suited to vibe coding to get new ideas off the ground quickly, Krishnamurthy said."The sweet spot is where you have maybe an interesting academic idea that is well understood, and you have a codebase that's well understood, and you're trying to say, well, I want to take this idea and I want to take this piece of code and build an extension for it. That's a great example where you have something isolated – the blast radius is small – and you can go and use AI to interpret the code. Our own engineers are using AI quite heavily, but also judiciously."
