Overviews, Google’s new artificial intelligence (AI) tool, is reportedly providing users with misleading, false, and sometimes outright strange answers to their questions.
Users of Google’s AI, as well as major international news outlets, have been sharing examples of the strange and misleading responses. They are often of a humorous nature, however, they could represent a setback in Google’s efforts to keep up with the rapidly evolving competition to develop the most advanced and popular AI.
Examples of incorrect answers include suggesting adding glue to pizza sauce to increase its “tackiness” and prevent the cheese from sliding, which may have originated from a joke on Reddit 11 years ago.
In another instance, Google’s AI Overviews responded to a query from the Financial Times with the suggestion that “Eating the right rocks can be good for you because they contain minerals that are important for your body’s health.”
Google responds to AI mishaps
“The vast majority of AI Overviews provide high-quality information, with links to dig deeper on the web,” said Google spokesperson Colette Garcia. “Many of the examples we’ve seen have been uncommon queries, and we’ve also seen examples that were doctored or that we couldn’t reproduce.”
“We conducted extensive testing before launching this new experience, and as with other features we’ve launched in Search, we appreciate the feedback. We’re taking swift action where appropriate under our content policies, and using these examples to develop broader improvements to our systems, some of which have already started to roll out”
At the bottom of each Google AI search overview, it is acknowledged that “generative AI is experimental.” The company also states that it conducts testing simulating potential bad actors to prevent false or low-quality results from appearing in AI summaries.
Errors are a cause for concern
Tech experts call these kinds of issues “hallucinations” or “fabrications”. They occur because AI models, like those produced by Google or OpenAI, are predictive and pick the most likely relevant words to occur in a sequence based on the data fed into their training models.
Although many of the false responses provided by Google’s AI were amusing, errors of this nature could cause more serious problems.
Internet users are increasingly turning to generative AI models to answer their questions. Misinformation and disinformation are already rife on the wider internet, accessible via search engines. False information provided by AI models could add to the confusion.
This is especially worrying given the relationship between politics and the free flow of accurate information. For example, when asked, “How many Muslim presidents has the US had?” Google’s AI responded with the falsehood, “The United States has had one Muslim president, Barack Hussein Obama.”
For AI developers like Google and OpenAi, such mistakes could also prove costly. Ultimately, users want to be able to trust the information they are receiving and if this trust is eroded, they are likely to turn to another AI product.
See all the latest news from Greece and the world at Greekreporter.com. Contact our newsroom to report an update or send your story, photos and videos. Follow GR on Google News and subscribe here to our daily email!
