Stack Overflow CEO teases generative AI tools, addresses moderator protest

Stack Overflow CEO Prashanth Chandrasekar, in a recent wide-ranging interview with VentureBeat, focused on an optimistic view of the generative AI explosion. His positive outlook shines even as generative AI tools like ChatGPT have put the popular question-and-answer site for developers in the headlines on several fronts — including March data that showed traffic to Stack Overflow fell by nearly 14% year-over-year just as ChatGPT use was exploding.

Chandrasekar teased a suite of generative AI tools Stack Overflow will release later this summer — and said that 10% of the company is working on it.

>>Follow VentureBeat’s ongoing generative AI coverage<<

“There’s definitely a question around how we leverage this technology to deliver on our mission of helping build technology through collective knowledge,” he said. “This intersection between the power of community on one side and AI on the other side — from my standpoint, human-generated community content has taken us to this level, we have a large impact, but there are also so many problems we can solve by leveraging this technology.”

Stack Overflow will integrate generative AI both on the public website as well as its enterprise offering, Stack Overflow for Teams, he explained. “We want to do it responsibly and safely and have the right use case to solve specific user and customer problems,” he said. “It’s going to be very, very exciting — I can’t wait for the world to know and for our community to use the things that we’re about to announce.”

Stack Overflow developer survey showed increased use of gen AI

Chandrasekar also touted the latest Stack Overflow developer survey released last month, which gathered data from more than 90,000 developers. The findings showed that generative AI chatbot tools have had a major impact on developers. A whopping 83% of respondents said they have used ChatGPT over the past year, while GitHub Copilot was ranked as the top most-used AI developer tool.

Far from gen AI leading to developer layoffs, the number of software developers is going to explode thanks to generative AI, Chandrasekar predicted.

“We’re probably talking about 30 million official developers based on our own data sets, but that number is probably going to at least triple because there are so many people that are going to come into the field of writing code because it’s so much easier to begin to write a baseline level of quality of code,” he said. “And I think the watermark has just sort of gone up in terms of the expectations for everyone.”

Developers that don’t embrace gen AI tools will be left behind, he emphasized. “I think you want to definitely be adopting these tools to get more productive and efficient and learn faster in your role.”

Embrace of gen AI comes as moderators protest AI-generated content

Stack Overflow’s embrace of gen AI tools comes months after it instituted a “temporary policy” in December banning the use of ChatGPT on the site, saying it was due to to the “general inaccuracy” of the answers, as well as that such posts violate the referencing requirements of Stack Overflow. This led to the removal of thousands of posts, while hundreds of users were suspended for violating the policy.

But at the end of May, the company changed course, with a new policy requiring moderators to stop moderating AI-generated content simply for being AI-generated. And, in early June, volunteer moderators of Stack Overflow announced they would go on strike because of the company’s new AI content policy, which they said could lead to misinformation due to chatbot hallucinations.

The decision to ban ChatGPT was the “right decision” back in December, said Chandrasekar. But now, he explained, the company has slowly been working with the community, doing research and getting input.

“We have gotten feedback on how best to make sure that we will trust generative AI products, from a Stack Overflow standpoint, and that’s what we’re going to announce in about a month’s time,” he said. But he said that current tools like ChatGPT create many false negatives where comments are wrongly flagged as AI-generated.

“That is causing a really negative impact because when a legitimate human being coming on the website wants to ask a question, we don’t want to shoo them away,” he said. “We want to share the question, we want to serve our purpose and mission.”

Due to the nature of current gen AI tools, he continued, it is hard to “smoke out generated content.” So the decision to create a new AI policy, was “not an easy decision by any means — but we just thought that the cost of dissuading people from asking a question on Stack Overflow was too high. And it was happening very alarmingly high rate based on the analysis that we looked at.”

Stack Overflow pushes ahead with gen AI as it plans to charge for training data

Stack Overflow’s push into generative AI tools also comes a few months after it announced that it will charge AI tech giants for its data to be used in training AI models. In April, Chandrasekar said Stack Overflow would charge large AI developers for access to the 50 million questions and answers on its service.

He told VentureBeat that up until now, using Stack Overflow training for training AI models happened “in the background.” The legal framework is nascent and “there isn’t a lot of precedent on this topic,” he said.

“We’ve taken a public position around this subject, saying that we always want to be open and free with our data with our developer community. However, large AI companies building models — we absolutely want to work with them in a more formal way.”