Google's Demo of Upcoming AI Shows It Making Huge Factual Mistake

This is absolutely embarrassing.

FAIlure to Launch

Google's ChatGPT competitor called Bard — which doesn't even have a launch date yet — is already off to a rough start.

As spotted by Reuters, Google's chatbot is already spewing misinformation — right in the company's own promotional demo.

"Google just announced they’re going to be rolling out AI search results and there is inaccurate information IN THE AD COPY!" science communicator and author Hank Green tweeted.

False Statements

It's a picture-perfect example of the current state of AI chatbot tech. While they're extremely good at generating plausible-sounding responses, contemporary large language models (LLMs) like OpenAI's GPT and ChatGPT are still largely unable to differentiate truth from fiction.

And that bodes badly, because tech giants including Microsoft and Google are racing each other to deploy these LLMs in their widely-used search engines — potentially filling the internet with complete garbage.

Case in point, a short video of Google's new feature, which the company's official account shared on Twitter earlier this week, contains some glaring issues. One whopper: the blatantly false statement that NASA's James Webb Space Telescope took the first picture of an exoplanet, or a planet outside of our solar system.

In fact, as NASA is the first to admit, the first exoplanet to be imaged, a Jupiter-like planet orbiting the brown dwarf 2M1207, was captured by the European Southern Observatory’s Very Large Telescope (VLT) back in 2004, 16 years before the James Webb started observations.

Bad Start

This morning, the search giant revealed it will include generative AI answers in some of its search results, but still has yet to reveal when the public will gain access to the feature.

All we know is that the wider public will have access "in the coming weeks."

Investors were clearly not impressed, with Alphabet shares dipping nearly eight percent so far today.

In short, it's a worrying start in advanced AI for the largest search engine in existence, when a highly anticipated new feature is already spreading lies before even being launched to the public.

"Natural. Language. Processors. Can. Not. Know. When. They. Are. Lying," Green added in a follow-up tweet.

READ MORE: Google AI chatbot Bard offers inaccurate information in company ad [Reuters]

More on Bard: Google Officially Announces ChatGPT Competitor Called "Bard"