Google launched its chatbot Bard last week. For a moment if you asked the chatbot about itself, it would claim that it had already been shut down by Google. The mistake led to a saga that included two AI-powered chatbots sourcing incorrect information to each other. I found myself at the center of this series of events, as one of those bots listed my article as a source.
Bard is an AI chatbot that’s powered by Google’s LaMDA. It competes with Bing Chat, which is powered by OpenAI’s GPT-4. Bard launched in preview last week, so people had a chance to play with it and test its artificial intelligence.
Twitter user and UX writer Juan Buis asked Bard “how long will it take before google bard will be shut down?” The tool then incorrectly stated:Â
“Google Bard is already shut down. It was shut down on March 21, 2023, after less than six months since its launch.
Google cited the lack of adoption as the reason for the shutdown and expressed their commitment to continuing to explore ways to use AI to enhance creative expression.”
Google, of course, had not shut down Bard. The tech giant is known for cancelling services — see killedbygoogle — but Bard is up and running in preview.
and here’s the Bing chatbot telling everyone Google Bard has already been shut down. It references a news report about Google Bard incorrectly saying it was being shutdown. We now have AI helping feed other AI bullshit 🥴 pic.twitter.com/hEiq2bPvyPMarch 22, 2023
Then, Bing Chat shared Bard’s incorrect information. The Verge’s Senior Editor Tom Warren asked Microsoft’s chatbot about Bard. Much like Google’s tool, Bing Chat stated that Bard had been shut down. But much to my surprise, Bing Chat cited my own article as a source.
I thought the piece, then titled ‘Google Bard says it’s ‘already shut down’ in obvious mistake,’ was clear that Google had not shut Bard. Bing Chat didn’t get the memo.
As AI is prone to do, Bing Chat’s response to the question “has Google Bard been shut down” or similar queries evolved quickly. While Bing Chat initially said that Bard had been shut down after about a day the chatbot would say that it could not comment on the topic.
Now, Bing Chat accurately states the following:
“Google Bard has not been shut down. There was an error where Bard incorrectly said it had already been shut down. This error is due to Bard sourcing information from the web.”
Would you like to take a guess as to which article is the source of that information? It’s the very same article that Bing Chat originally cited as the source when stating Bard had been shut down.
When asked for its sources, the first link shared by Bing Chat is my article on MSN. The fifth source is a direct link to my previous article.
A strange saga
This situation was at least odd. Some might say it’s worrying. I never guessed that I’d be part of two AI-powered chatbots feeding each other incorrect information. While I see the humor in the saga, there is cause for concern. Bing stated that Bard had been shut down as a fact. It then listed a link to my article on Windows Central. It didn’t show the headline of the article. which would have clearly indicated that Bard had not been shut down, or any other context.
What if someone saw Bing’s response and my piece as a source of fake news? My reputation and people’s view of Windows Central could be adversely affected.
While AI isn’t new, public-facing chatbots using the technology to this extent are. I predict that over time Microsoft and Google will refine how Bard and Bing Chat source information. But in the meantime, I’d prefer to not be listed as a source for a statement that directly contradicts my headline and article.
