(Credit: Bloomberg / Contributor via Getty Images)
Google is now using AI to write more than a quarter of its new code, according to Google and parent company Alphabet CEO Sundar Pichai, who shared the news during the firm's most recent earnings call.
Google is "using AI internally to improve our coding processes, which is boosting productivity and efficiency," Pichai said. "Today, more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers. This helps our engineers do more and move faster."
Using a generative AI model to write code isn't exactly a new idea—though it could ultimately result in fewer software engineer jobs, especially at more entry levels, while those who remain take on more work. Earlier this year, startup Cognition Labs announced its own AI "software engineer," dubbed Devin, which can execute entire engineering projects from start to finish with little to no human intervention. And last year, ChatGPT reportedly proved itself to have the skills of a Level 3 software engineer because it was able to pass a Google coding interview.
But using AI to write code could pose copyright issues or create security flaws if it's trained on licensed or old code. And some companies using AI-crafted code have already reportedly experienced outages and cybersecurity problems as a result, primarily due to a lack of adequate human oversight.
Google is continuing to go all-in on AI, from making AI Overviews appear at the top of Google Search to adding AI features to YouTube and developing a new Gemini model. Google is reportedly plotting a December release for its next Gemini model, which may be called Gemini 3.0, and could launch the same month as a new model from its main AI rival, OpenAI.
Google's next AI model may be an autonomous agent, meaning it can complete tasks by controlling a computer, opening different applications, typing, and clicking on a screen to finish a project or conduct research. This means Google could potentially offload even more of its internal projects to AI models in the future.
It this came to fruition, it would follow Anthropic's launch of a similar feature, Computer Use. But AIs can hallucinate or be tricked. Because of this, giving an AI this much power poses some major security concerns, and underscores the ongoing need for human supervision and oversight.
The increased use of AI tools also means higher energy use and data center costs, sometimes in areas with existing power shortages or unstable grids. AI's current high electricity and water use stand to take a substantial toll on the environment, and Google's emissions alone have already spiked 50% in the past year thanks to its AI use. Now, it plans to go nuclear in an effort to lessen that energy burden.
For now, AI is good for Google's bottom line. Alphabet's stock price is up about 2% in the past day, and up 36% in the past year. Nvidia has also seen massive gains in the AI era, but rivals like AMD haven't seen as much traction from the generative AI surge.