Did Bard use ChatGPT data for training? Google denies allegations

Google has been accused of using OpenAI chatbot ChatGPT's data to train its own AI bot, Google Bard. Tech companies have deviated their focus to stopping ChatGPT from gaining more popularity. In line with this, some leading tech firms recently jumped on the AI chatbot bandwagon. Much to their chagrin, most newly launched AI chatbots including Microsoft's Bing Chat have failed to match OpenAI's offering in terms of popularity.

Google Bard is one of the newly launched AI chatbots that's struggling to catch up with ChatGPT. The search engine giant introduced its AI chatbot last month to compete with ChatGPT. However, Bard researchers have been catching flak lately. According to a report by Android Police, researchers of Bard are being accused of using OpenAI's technology data without consent to develop Google's AI bot.

Did Google Bard use OpenAI's technology data?

Generative AI models are made using a combination of machine learning and training data, and Bard is no exception. Google uses its LaMDA (Language Model for Dialogue Applications) as a machine language model. LaMDA comes in handy for forming the syntax of the chatbot. As far as training data is concerned, there's no dearth of open-source libraries that AI researchers can use, but Google is unlikely to have utilised these libraries.

However, the report suggests the public domain data was not enough to develop Bard. The Information claims Google took advantage of the ChatGPT data that was shared publicly by its developers via ShareGPT. To those unaware, ShareGPT is a website that allows users to share the OpenAI chatbot's responses. Former Google AI researcher Jacob Devlin reportedly warned the company's chief executive Sundar Pichai and other top executives that the company would violate OpenAI's terms of service by using ChatGPT data.

Aside from this, Devlin was concerned that Bard would provide ChatGPT-like responses if Google uses OpenAI's data. Notably, Devlin now works for OpenAI. Furthermore, the report claims the American tech firm told its DeepMind division to work with the Brain team on another initiative known as Gemini. In its statement to WinFuture, Google clarified that Bard wasn't trained on data from ChatGPT.

Google told the publication that "Bard is not trained on data from ShareGPT or ChatGPT." According to an earlier report by Business Insider, Pichai asked Google employees to put two to four hours into enhancing its Bard chatbot before its release.

OpenAI CEO's response to Google using ChatGPT's data

Taking to Twitter recently, OpenAI CEO Sam Altman announced that he is "not annoyed" at Google for using ChatGPT data to train its own AI chatbot. Google revealed it will make Bard available to select users in the US and the UK not long ago. People who are interested in testing the new AI chatbot will have to sign up on the waitlist before getting access.

Bard is likely to make a few mistakes occasionally since it is still in the initial stages of testing. In fact, Pichai even warned the employees about the AI bot's possible errors via an email, according to a report by CNBC. "Things will go wrong," the executive warned Google employees in the email. However, Bard is slated to get major upgrades in the coming days, Pichai told New York Times' Hard Fork in a podcast.

Aside from this, Pichai said Google is prepping to upgrade Bard to some of the company's more capable PaLM models. This will improve Bard's capabilities in multiple departments including coding and reasoning. Also, it will allow the AI chatbot to provide more accurate answers to maths questions. "So you will see progress over the course of next week," Pichai added.