Google Denies Bard Was Trained With ChatGPT Data

Google’s AI service Bard hasn’t exactly made an impressive first impression, and according to The Information, the company is pushing its DeepMind division to work with the Google Brain team to defeat OpenAI with a new initiative termed Gemini in an effort to turn around the fates of its AI chatbots. The report from The Information also raises the possibly shocking third-party claim that Google descended so low as to train Bard using information from OpenAI’s ChatGPT, which was collected from a website called ShareGPT. The report claims that a former Google AI researcher opposed using that data.

A spokesperson for Google, Chris Pappas in a press brief vehemently and unequivocally denied that the data is used. Chris says  “Bard is not trained on data from ShareGPT or ChatGPT.” 

Further detail to verify if Google utilized ChatGPT data to train Bard in the past or not has proven abortive. “Unfortunately, all I can share is our statement from yesterday,” Pappas says.

According to The Information’s research, a Google AI engineer named Jacob Devlin attempted to advise Google not to exploit that ChatGPT data because it would infringe OpenAI’s terms of service and its replies would look too similar. He then left Google and joined its rival OpenAI right away. Google ceased using the data after his warnings based on the source’s advice.   Perhaps it disregarded that aspect of the training as well.

Whether Google ultimately used the data or not, it’s surprising to know that DeepMind, which has been making attempts to disassociate itself from Google for several years, maybe getting a call from Google.