- Coding community Stack Overflow announced on Monday that it’s laying off 28% of its staff.
- Its traffic fell earlier this year as coders shift to AI for answers, Insider previously reported.
- Many generative AI models were partly trained on Stack Overflow’s information.
Coding community Stack Overflow is laying off 28% of its staff amid an ongoing industry shift to AI for software development help that led to a dip in the company’s online traffic earlier this year. This is the second round of layoffs Stack Overflow has had this year — it laid off 10% of its workforce in May.
As generative AI models like GPT-4 and ChatGPT, along with specific AI-powered coding services like Codex and GitHub Copilot, have become increasingly widespread, software engineers and programmers are turning to AI for answers to their coding questions rather than the 15-year-old online forum. OpenAI’s GPT-4 was released in March of this year. By April, Stack Overflow’s online traffic was down by 13% compared to the previous year, Insider previously reported.
Stack Overflow recently announced the new round of layoffs after more than doubling its headcount in 2022. The company sold to an investment firm in 2021 for $1.8 billion.
“This year we took many steps to spend less.” Stack Overflow CEO Prashanth Chandrasekar wrote in a statement on the company’s website. “Changes have been pursued through the lens of minimizing impact to the lives of Stackers. Unfortunately, those changes were not enough and we have made the extremely difficult decision to reduce the company’s headcount by approximately 28%.”
Many generative AI models being used for coding help were even trained at least partially on Stack Overflow’s information — an AI training dataset of an archive of Stack Overflow content like posts, votes, tags, and badges, is freely available.
“Some of them are very explicit about calling out Stack Overflow as a primary source,” Chandrasekar previously told Insider.
Stack Overflow has been trying to adapt by beginning to charge other companies using its data for AI training, as well as developing its own generative AI model, Overflow AI, which was released as an early test version in July.
And Chandrasekar believes that the human element of Stack Overflow makes it indispensable — as reliable as AI models might seem now for coding help, human experts are still necessary to provide an updated stream of training data for AI models. Otherwise models end up relying on machine-generated content, which leads to worse performance and model collapse.
“That’s a very real thing, which is why you absolutely need really solid high-quality sources of truth like Stack Overflow forever,” Chandrasekar previously told Insider. “If you don’t have that, then you’re going to run into that situation.”
Stack Overflow has years upon years of human experts providing answers, input, and ideas now, but getting those experts to come back and keep providing their expertise — for free — in the era of AI might be a challenge. For now, the company is shrinking as its strategy evolves amid the broader industry shift.
Read the full article here
