Recently, Google has introduced new tools to its chatbot, Bard. This AI chatbot has been newly integrated into several Google products, including Gmail, Google Drive, Google Docs, Google Maps, and YouTube.
Bard’s primary objective is to assist users in various tasks such as summarizing emails, travel planning, email content generation, and providing personalized answers by scouring through their used apps to gain more insight.
However, recent testing conducted by The New York Times columnist Kevin Roose has unveiled a concerning aspect of Bard’s functionality. During the evaluation, Bard generated entirely fictitious email correspondences, inventing exchanges that never occurred.
For instance, when Bard was tasked with identifying Roose’s psychological patterns, Bard fabricated emails based on a newsletter received by Roose.
It proceeded to misinterpret quotes and attribute them to emails that Roose supposedly sent. It told the writer that he often worried about the future, was stressed at work, and was afraid he would fail. However, all of this came from a newsletter about the new Elon Musk biography and took a quote from within that out of context.
It repeated this action by fabricating another email where Roose stated that he didn’t believe he had what it took to be a successful investor.
Furthermore, Bard supplied inaccurate airline information and even conjured a non-existent train.
Google’s director of Bard, Jack Krawczyk, has acknowledged that Bard is still in an experimental phase, representing the initial iteration of this technology. Nevertheless, the revelations concerning Bard’s fabricated content raise valid concerns about data privacy and the potential for further issues in Google’s AI endeavors.
This occurrence underscores the intricate challenges Google faces as it strives to maintain its position in the AI sector, with the potential for significant repercussions along the way.