Most people probably wouldn't have been able to reveal Bard's mistake, even though it's actually quite egregious, especially for a web-trained conversational AI. The stumbling block also escaped all the Google staff, in charge of preparing the launch video, while it did not go unnoticed by many experts and astronomy enthusiasts .
Indeed, Brad stated that the James Webb space telescope (Jwst) would have been used to take the first photos in history of a planet located outside the solar system, or exoplanet. A completely false statement, given that the first exoplanets were photographed thanks to powerful ground-based telescopes even 14 years before the launch of the James Webb.
What made Brad fall into error was a poor understanding of logic grammatical of the information found on the web, which therefore does not live up to expectations. In fact, as Bruce Macintosh, director of observatories at the University of California and among the first to photograph an exoplanet from the ground, pointed out on Twitter, the chatbot would have confused the phrase "the first photo of an exoplanet taken by the James Webb space telescope" with “The first photo of an exoplanet was taken by the James Webb Space Telescope”.
A misunderstanding that is easier to understand in English, due to the almost identical form of the two sentences, and which has dramatically led an artificial intelligence astray, designed precisely to be able to avoid falling into these typically human errors of distraction . After receiving the error report, as reported by Reuters , Google said that this problem underlines the need to continue with new updates to the Bard system.
"This fact highlights the importance of a rigorous process of verification, which we're launching this week with our team of trusted testers," a Google spokesperson said. "We'll combine external feedback with our own internal testing to make sure Bard's responses meet a high level of quality, safety, and accuracy. with real-world information".