Apple Details Data Collection To Bolster AI And How It Affects Your Privacy

Apple has joined most of the tech world in riding the artificial intelligence wave, though the company is struggling mightily to integrate Apple Intelligence into its products. Embarrassingly, its AI has often failed to even get small features, such as summarizing headlines, working the way most expect it to. According to a new Bloomberg report, those shortcomings are because Apple only trains its models using synthetic data.

The company is hoping to right the ship by leveraging a technique it developed, dubbed differential privacy. By using this technique, which happens directly on a user’s device, it will be able to compare its synthetic data with real world data generated by users. Apple says that a “participating device will send only a signal indicating which of the variants is closest to the sampled data on the device.” This comparison will also only take place if a user has opted-in to sharing data analytics.

It's always interesting to see how Apple manages to access user data in privacy preserving ways. Even if this doesn’t end up helping the company improve its AI, it might prove to be a useful technique for other companies to consider in an effort to roll out new AI-enabled features. As AI becomes more and more prevalent in people’s lives, there will be more questions about privacy that will crop up. Having this technique available can go a long way with helping people become more comfortable with using AI tools.

Time will tell if this new approach to gathering training data for its models will work as intended, moving Apple closer to where it needs to be with its AI efforts. The company has been betting big on AI as a potential new source of revenue, as sales of its popular iPhones have missed the mark in recent years.