The initial demo of Google's new AI chatbot, Bard, contained a factual inaccuracy, which has put the chatbot's development in jeopardy. A rival to OpenAI's ChatGPT, the bot was unveiled on Monday and was slated to become more broadly available to the public in the coming weeks. When questioned in a demo by Google about recent findings made by the James Webb Space Telescope, Bard provided three bullet points of information, including the claim that the telescope captured the very first photos of a planet outside of our own solar system. This assertion was untrue, though.
The first photograph of an exoplanet was acquired in 2004, not by the James Webb Space Telescope, as some well-known astronomers have been keen to point out.
Grant Tremblay, an astronomer, noted on Twitter that while Bard is "impressive," AI chatbots like ChatGPT and Bard often confidently assert false information. The computers do not search a database of verified facts; instead, they are trained on vast amounts of text and evaluate patterns to identify word sequences. As a result, they have the ability to "hallucinate" and fabricate data, earning the label "bullshit generators" from one AI scientist.
Tremblay later posted a screenshot of a comparable search on "old" and "busted" Google Search, which was able to get the right result. During regular trading, Alphabet's shares swiftly reacted and fell as much as 9% with volumes that were roughly three times the 50-day moving average.
The same AI that powers ChatGPT will be used in a more sophisticated form in a new version of Bing and the Edge browser, Microsoft revealed earlier this week. On the other hand, Google has only made Bard available to "trusted testers". The business asserts that, as a result of errors like this, Google will thoroughly evaluate Bard before making it available to a larger public.
In response to the mess that Bard created on Wednesday, Google recognised the value of a thorough testing procedure. The programme will mix external feedback with internal testing to ensure that Bard's solutions satisfy a high level for "quality, safety, and real-world information," a company spokeswoman told The Verge.