Apple will update AI-generated news alerts feature that spews out inaccuracies

Many are urging Apple to cancel its new AI feature generating inaccurate news alerts for iPhone users altogether. But the company is choosing to update the technology rather than pause it.

The Apple Intelligence feature, launched late last year, took only a few days to demonstrate that it’s not capable of producing reliable information in a consistent and trustworthy manner.

At first, one of the notifications incorrectly summarized The New York Times’ story and said that Benjamin Netanyahu, Israel’s prime minister, was arrested – when he actually wasn’t.

Apple AI notification summaries continue to be so so so bad

[image or embed]

undefined Ken Schwencke (@schwanksta.com) November 21, 2024 at 9:22 PM

Later, just two days after the AI tool was activated in the United Kingdom on December 11th, the BBC, Britain’s public broadcaster, complained to Apple that an AI-generated summary of its headline falsely told some readers that Luigi Mangione, the individual accused of killing UnitedHealthcare CEO Brian Thompson, had shot himself.

The BBC contacted Apple to raise their concerns, and the group Reporters Without Borders (RSF) called on the company to remove the technology altogether.

These notification summaries group together and rewrite previews of multiple recent app notifications into a single alert on users' lock screens.

“AIs are probability machines, and facts can’t be decided by a roll of the dice. RSF calls on Apple to act responsibly by removing this feature,” said Vincent Berthier, head of RSF’s technology and journalism desk.

“The automated production of false information attributed to a media outlet is a blow to the outlet’s credibility and a danger to the public's right to reliable information on current affairs.”

Apple chose not to comment back then, and, unsurprisingly, the problem persisted. Just last week, the company’s AI generated news alerts falsely claiming that Luke Littler had won the World Darts Championship hours before it even began, and that the Spanish tennis star Rafael Nadal had come out as gay.

Now, the company has finally acknowledged the issue. In a statement sent to media organizations, Apple said it was working on a software change to “further clarify” to users when the notifications have been generated by Apple Intelligence.

The notification summaries, generated by AI for individual users, already have a small icon on them to indicate that they aren’t normal, unaltered alerts from the news organizations’ apps. However, some if not most users don’t even know what the icon means, and Apple seems to admit it.

“A software update in the coming weeks will further clarify when the text being displayed is summarization provided by Apple Intelligence,” said the company, adding that the feature was optional anyway and that users can disable the summaries on their devices.

Anyway, Apple’s explanation means that the feature isn’t being removed. Having rushed to roll out Apple Intelligence into the market in order to keep up with its rivals, the firm has rather chosen to try and make the users know more about the drawbacks of the feature.

Moreover, these mistakes are actually unavoidable, and no software updates will help. That’s because the large language models – also used for Apple’s notification summaries – are not capable of understanding the content they’re summarizing.

These models work by predicting portions of words based on what came before. Besides, even though the models are still being trained and are improving, so far, it simply can’t reach zero error percentage when myriads of summaries are generated every day.