Controversy between AIs: Google accused of training Bard with data from ChatGPT, but denies it!
The Information published allegations that Google is using ChatGPT’s responses to train Bard. The company denies the use of such data.
Google’s Bard did not have an impressive debut, but according to The Information, the company is making efforts to improve its Artificial Intelligence (AI) chatbot’s experience.
The efforts are so much that it forced the two AI research teams of Alphabet, Google’s parent company, to overcome years of intense rivalry to work together.
As a result, the Google Brain AI software engineers are working with employees from DeepMind, an AI laboratory that is an Alphabet sister company, to develop software that competes with OpenAI.
The new initiative, called Gemini, began after Google stumbled with the release of Bard.
Furthermore, The Information’s report also contains a third-hand claim that Google stooped so low as to train Bard using data from OpenAI’s ChatGPT, extracted from a site called ShareGPT.
Training Bard with the competitor?
According to The Information’s report, a Google AI engineer named Jacob Devlin left the company to join its rival OpenAI after trying to warn them not to use ChatGPT data because it would violate OpenAI’s terms of service, and its responses would seem too similar.
According to Devlin, the Bard team relied “extensively” on ChatGPT responses posted on ShareGPT, a site where users often submit responses they obtained from OpenAI’s chatbot.
One source told the publication that Google stopped using this data after Devlin’s warnings.
Thus, Google firmly denies that ChatGPT data is being used, but it does not deny that it has been used in the past.
In a statement to The Verge, Chris Pappas, a Google spokesperson, said, “Bard is not trained on any data from ShareGPT or ChatGPT.”
However, the spokesperson refused to answer whether Google had used this data to train Bard in the past. “Unfortunately, all I can share is our statement from yesterday,” he said.