It’s early days in the generative AI chatbot wars, but OpenAI’s ChatGPT has a marked advantage over Google’s Bard right now. The distance between the two chatbots grew even further last week when OpenAI debuted GPT-4 — the language model powering Bing Chat. Now, a report from The Information suggests that in order to jump-start its own AI chatbot’s progress, Google is using data from ChatGPT to train Bard. Google has since denied these allegations.
Google says ChatGPT isn’t training Bard
According to the report, Google’s parent company Alphabet is forcing two competing AI research teams to work together on Bard. The joint effort, named Gemini, combines employees from the AI research lab DeepMind and Google’s Brain AI group.
In an effort to speed up Bard’s development, a former Google researcher alleges the Bard team “appeared to be relying heavily on information from ShareGPT, a website where people publish conversations they’ve had with ChatGPT.”
This is obviously a serious allegation, but Google denies that it’s true.
“Bard is not trained on any data from ShareGPT or ChatGPT,” Google spokesman Chris Pappas told The Verge. The publication asked if Google had ever used ChatGPT data to train Bard in the past. “Unfortunately, all I can share is our statement from yesterday,” he said.
The report states that Jacob Devlin, the researcher mentioned above, left Google to join OpenAI after warning the Bard team not to use ChatGPT data. He told them that not only would the Bard answers look too similar, but that it would violate OpenAI’s terms of service. Another source told The Information that the team stopped using the data after that.
As all these AI chatbots grow and evolve, it’s inevitable that they will eventually interact with one another. That said, lifting data from one chatbot to train another might be a bridge too far. If you want to see Bard in person, you can sign up now to try it out.
