The Mountain View, California, company, which is owned by Alphabet Inc., hasn't said how many people will be given access to Bard yet (Image: Google) The race to create the best conversational AI chatbot has intensified since ChatGPT launched last November, sparking what some call the AI war. But Google may have crossed a line with its own chatbot Bard, according to a recent report that claims the tech giant is skimping on the training data for its AI system.
In a report, The Information has claimed that Google is forcing its DeepMind division to help the Google Brain team gain an edge over AI powerhouse OpenAI with a new initiative called Gemini. The report also suggests that Google was so desperate that it trained Bard using data from ChatGPT, scraped off a website called ShareGPT.
A Google AI engineer named Jacob Devlin reportedly tried to warn Google to not use ChatGPT data since it would violate OpenAI’s terms of service – and because Bard’s answers would look too similar if Google did so. Following this, the engineer left the company to join OpenAI.
Google did stop using ChatGPT data following his warnings according to the report, though. However, per a Google spokesperson who spoke to The Verge, the search giant never used any data from ShareGPT or ChatGPT in the first place.
Despite Google’s efforts, Bard has seen quite a rocky start. In its first demonstration early last month, Bard made a factual error about the discoveries from the James Webb Space Telescope, leading to Google parent Alphabet losing $100 billion as stocks tumbled.
Still, whether or not Google actually did the deed, it’s interesting to see that Google is exercising its control over Deepmind, which has been seeking more autonomy from Google for a long time.