Google refused to use ChatGPT data to train Bard

In response to Open Ai’s ChatGPT, major tech firms have never been more eager to slow down; some, like Microsoft, jumped on board quickly, while others, like Google, were left lagging behind. Last month, in response to the well-liked generative AI from Open AI, the business finally unveiled Bard. Recently, allegations have surfaced that its employees used data from Open AI’s technology without authorization to aid in the development of Bard.

Bard is based on a combination of machine learning models and training data, just like other kinds of generative AI. In the instance of Bard, Google created the chatbot’s syntax using its Language Model for Dialogue Applications, or LaMDA. There are many open-source libraries for training data that AI researchers can use, and there is no question that Google has looked everywhere for them. It seems, though, that the information in the public realm was insufficient.

According to The Information, Google used data from publicly available ChatGPT answers shared through ShareGPT, a website where users can share ChatGPT chatbot responses, to train Bard. After telling senior managers, including CEO Sundar Pichai, that using ChatGPT data would be against OpenAI’s terms of service, Jacob Devlin, a Google AI researcher, departed the company. Devlin, who now works for OpenAI, was also worried that Bard might receive similar responses if Google accessed ChatGPT’s data.

According to the article, Google instructed its DeepMind division to work with the Brain team on a brand-new project known internally as Gemini.

In a letter to WinFuture, Google stated that Bard’s training did not rely on information from ChatGPT. According to the business, “Bard is not trained on data from ShareGPT or ChatGPT.”

Google was also approached by Android Police for a statement, but we haven’t received one yet.

Google has struggled and even crunched as it tries to develop Bard on an advanced timeline, balancing moral dilemmas and stagnation. Pichai reportedly urged Google employees to take up to four hours out of their day to fix some bugs in its ChatGPT, according to a prior story from Business Insider.

If Bard was trained using ChatGPT data before it was released, and whether it still uses such training, are both still unknown at this time. According to Google, neither situation applies.