Stack Overflow signs agreement with OpenAI to provide data for its models

OpenAI is working with Stack Overflow, a question-and-answer site for software developers, to make its generative AI models better at tasks that have to do with programming.

The two companies say that OpenAI’s models, including those used by its ChatGPT chatbot platform, should get better over time at answering questions about programming because of the partnership that was announced on Monday. At the same time, OpenAI’s knowledge of building new generative AI features into the Stack Overflow platform will help Stack Overflow.

By the end of June, the first set of features will be live.

Stack Overflow, which initially blocked ChatGPT responses on its site because it believed them to be spam, is making a significant change with its partnership with OpenAI.

Since April of last year, Stack Overflow has been testing generative AI features to create models that “reward” developers who contribute to the platform. The company released a conversational search tool in July. It allows users to ask questions and receive answers from Stack Overflow’s database of more than 58 million questions and answers. The company also released tools for businesses to fine-tune searches based on their own documentation and knowledge bases.

Concerns about the accuracy of AI-generated information, the volume of information available, and the privacy of individual contributors’ data on the platform led some developers on Stack Overflow to oppose the changes.

At least some of those worries were valid. GitClear has examined over 150 million lines of code added to project repositories in the past few years. GitClear discovered that the use of generative AI development tools leads to an increase in the number of errors in codebases. In other places, security researchers have said that these kinds of tools can make bugs and security problems in software projects even worse.

However, developers are still using generative AI tools for at least some coding tasks, even though they seem to have some problems. A Stack Overflow poll from June 2023 found that 44% of developers use AI tools in their work now, and 26% plan to do so soon.

This has caused Stack Overflow to have a bit of an existential crisis. Reports say that a lot fewer people are visiting the platform since last year when powerful new generative AI models came out. Stack Overflow provided the training data for many of these models.

To cut costs, Stack Overflow is now looking for licensing deals with AI providers.

The company’s deal with OpenAI, which didn’t say how much it would cost, comes after Stack Overflow teamed up with Google to add Stack Overflow data to Google’s Gemini models and work with Google to add more AI-powered features to its platform. Stack Overflow clarified at the time that the deal was not exclusive, and it proved to be accurate.

CEO of Stack Overflow Prashanth Chandrasekar has said that 10% of the platform’s nearly 600 employees are working on its AI strategy. He has also said that the strategy’s potential to bring in extra money is key to ensuring that Stack Overflow can keep attracting users and keeping the quality of its information high.

The head of Stack Overflow said in a press release this morning, “Stack Overflow is the world’s largest developer community.” “Through [our] industry-leading partnership with OpenAI, we want to change the way developers work by using the power of community, best-in-class data, and AI experiences to make work more efficient and encourage teamwork.” Our goal with OverflowAPI, as part of our work to advance the field of socially responsible AI, is to set new standards with checked, reliable, and correct data, which we will use to build and deliver technology solutions to our users.