Google Asks DeepMind For Help To Improve Bard Chatbot, But Denies Use Of OpenAI Data

JAKARTA - Bard, a chatbot made by Google, is reported to have failed to achieve good performance. Therefore, the company enlisted the help of DeepMind's artificial intelligence (AI) division to help the Google Brain team beat OpenAI with a new initiative called Gemini, as reported by The Information.

The report also mentions the surprising notion that Google used data from OpenAI's ChatGPT taken from a site called ShareGPT to train Bard. A former AI researcher at Google has reportedly warned the company against using the data as it violates OpenAI's terms of use and an answer from Bard will look too similar.

However, Google categorically denies that Bard was trained using data from ShareGPT or ChatGPT. "Bard was not trained on using data from ShareGPT or ChatGPT," spokesman Chris Pappas told The Verge.

According to a report by The Information, an AI engineer at Google named Jacob Devlin left the company to join rival OpenAI after trying to warn Google not to use the ChatGPT data.

He said the use of the data violated OpenAI terms and Bard's answer would look too similar to ChatGPT's answer. Another source says that Google stopped using the data after a warning from Devlin.

Regardless of whether Google uses the data or not, it's interesting to see that it might turn to DeepMind for years trying to become more independent from Google.