Apple’s AI disastrously rewrote a BBC headline saying Luigi Mangione shot himself

Apple has just started rolling out a much-hyped set of AI features to its devices and we’re already seeing major problems. For example, the bbc has complained to Apple after an AI-based notification digest rewrote a bbc headline that the killer of UHC CEO Luigi Mangione had shot himself. Mangione did not shoot himself and remains in police custody.

Apple Intelligence includes a feature in iOS that attempts to alleviate user fatigue by grouping and summarizing notifications coming from individual apps. For example, if a user receives multiple text messages from a person, instead of displaying them all in a long list, iOS will now try to summarize the push alerts into a concise notification.

It turns out, and this should come as no surprise to anyone familiar with generative AI, that the “intelligence” of Apple Intelligence contradicts the fact that summaries are sometimes unfortunate or just wrong. Notification digests were first introduced to iOS in version 18.1, which was released in October; Earlier this week, Apple added native ChatGPT integration to Siri.

in a articlehe bbc shared a screenshot of a notification summarizing three different stories that had been sent as push alerts. The notice reads: “Luigi Mangione shoots himself; Syrian mother hopes Assad pays the price; “South Korean police raid Yoon Suk Yeol’s office.” The other summaries were correct, the bbc says.

He bbc has complained to Apple about the situation, which is embarrassing for the technology company but also risks damaging the reputation of the media if readers believe they are sending misinformation. They have no control over how iOS decides to summarize their push alerts.

“BBC News is the most trusted news outlet in the world,” said a bbc the spokesperson told the story. “It is essential to us that our audiences can trust any information or journalism published on our behalf and that includes notifications.” Apple refused to respond to the BBC questions about the problem.

Artificial Intelligence has a lot of potential in many areas, but language models are perhaps one of the worst implementations. But there are many corporate hopes that the technology will be good enough that companies can rely on it for uses such as customer service chat or searching large collections of internal data. But it’s not there yet; In fact, companies using AI have said they still have to do a lot of editing of the work it produces.

It feels somewhat unusual for Apple to deeply integrate such unreliable and unpredictable technology into its products. Apple has no control over the results of ChatGPT: the chatbot’s creator, OpenAI, can barely control the language models and their behavior is constantly modified. Short notification summaries should be the easiest thing for AI to do, and Apple is failing even at that.

At the very least, some of the features in Apple Intelligence demonstrate how AI could have practical uses. Better photo editing and a focus mode that understands which notifications should be sent through are nice. But for a company associated with polished experiences, bad notification summaries and a mind-bending ChatGPT could make iOS look unpolished. It seems like they’re jumping on the hype train to boost sales of new iPhones – an iPhone 15 Pro or newer is required to use the features.