Top Google AI researcher Jacob Devlin quit earlier this year after warning Alphabet CEO Sundar Pichai and other top executives that the company’s ChatGPT competitor, Bard, was being trained on data from OpenAI’s chatbot.
A recent report from The Information has now revealed the news, citing a source with firsthand information of the situation, as well as another who had been briefed on it.
According to the report, Devlin told the executives that he suspected the team working on Bard was using information from ShareGPT, a site where users post exchanges with ChatGPT.
Devlin then cautioned executives that training on ChatGPT conversations could result in Bard sounding too similar to OpenAI’s chatbot. The researcher and other Google employees thought it was a violation of OpenAI’s terms of service.
The report continued, that Google ceased using the data to train Bard after Devlin alerted executives to the problem. Devlin joined OpenAI shortly after departing Google in January. He was one of several AI researchers who left Google for competitors at the start of the year.
Devlin, who worked at Google for more than five years, was the lead author of a 2018 study paper on training machine learning models for search accuracy, which contributed to the AI boom. His study has since been incorporated into the language models of Google and OpenAI.
Over the years, OpenAI has recruited dozens of former Alphabet employees. Google and OpenAI have been engaged in an AI arms race since the company’s chatbot made news in November for its ability to do everything from write an essay to provide basic code.
Google released Bard to a small set of users in the United States and the United Kingdom earlier this month. But Google’s race to catch up to OpenAI doesn’t end there as Alphabet’s two AI teams, DeepMind and Google Brain, have reportedly joined forces to contend with OpenAI.