Google Denies Claim That Bard Was Trained by Stealing ChatGPT Data

Bold Claims The rumor mill has churned out a pretty huge claim against Google’s new Bard AI chatbot — but the tech giant is denying, denying, denying. This mini-debacle begins with the allegation, published by The Information, that Google trained had trained its would-be Microsoft competitor chatbot on data from Microsoft’s partner OpenAI’s ChatGPT — a claim that would, if corroborated, be pretty embarrassing for Google, but one that a company spokesperson denied unequivocally in a statement to The Verge. According to someone with knowledge of the situation who spoke to The Information, this claim is said to have originated with Jacob Devlin, a former Google AI researcher who, per the allegation, learned that his then-employer was scraping OpenAI chatbot data from a site called ShareGPT to build out Bard. Devlin reportedly tried to raise concerns up the food chain about using the OpenAI data, primarily because it would violate the rival firm’s terms of use and also because, you know, it would be pretty obvious if the answers sounded the same. Another source told the publication that Google stopped using the ShareGPT data, though it’s unclear whether that part of the training data was thrown out with the bathwater, too, or if none of the scuttlebutt is true. As The Information reported, Devlin jumped ship for OpenAI itself soon after voicing his concerns, though it’s probably best not to read too much into that. Past Precedent Hilariously, Google was in the ancient times of 2011 on the other side of this…Google Denies Claim That Bard Was Trained by Stealing ChatGPT Data

Leave a Reply

Your email address will not be published. Required fields are marked *