Despite its early success with its digital assistant Siri, Apple Inc. has trailed its big technology rivals in artificial intelligence. Now, according to one report, it’s working on a dedicated AI chip to help remedy that situation.

A report from Bloomberg today said Apple’s processor, internally called the Apple Neural Engine, would provide improved services such as facial and speech recognition to iPhones, iPads and potentially other devices. AI technologies such as deep learning neural networks have in recent years vastly improved tasks such as speech recognition and language translations on devices ranging from smartphones to new smart speakers such as Amazon.com Inc.’s Echo and Google Inc.’s Home.

As Google, Microsoft Corp., Facebook Inc., IBM Corp., Baidu Inc. and others have hired hundreds of AI and machine learning experts, they’ve appeared to pull ahead of Apple in what looks to be one of the foundational technologies of this century. Apple has recently stepped up hiring and undertaken a steady series of acquisitions in AI. And last October, the company hired Russ Salakhutdinov, a Carnegie Mellon University associate professor in machine learning and computer science, to direct its AI research.

It’s not yet clear exactly how the chip would be used, but the report says it would handle AI work currently done on more standard central processing units and graphics processing units. The highly parallel processing of the latter chips, from the likes of Nvidia Corp., has been instrumental in crunching the massive amounts of data needed to train and run neural network models. Offloading the tasks not only would improve Apple’s services but also possibly improve battery life, since GPUs can use a lot of power.

The report also implied that the Neural Engine also could help Apple make headway in new applications of AI in which it has shown interest, such as self-driving cars and augmented reality. As a result, it was only a matter of time before Apple needed dedicated chips for AI. “The question wasn’t if Apple was going to do a specialized AI chip, but when,” said Patrick Moorhead, president and principal analyst at Moor Insights & Strategy.

Apple isn’t alone in developing dedicated chips for AI, though others have been more focused on powering servers in data centers and the cloud. Google last year announced its Tensor Processing Unit, which it uses internally for the likes of search results, language translation and image recognition. During its Google I/O software developer conference last week, it introduced a new version of the chip for its cloud computing service.

The week before, Nvidia also introduced a powerful new chip tuned for AI tasks. And mobile chip maker Qualcomm Inc. has added an AI module to its latest Snapdragon chips. What’s more, other companies such as Intel Corp. and several startups, some of which Intel bought, are trying out other chip designs optimized for various AI workloads.

Moorhead said the apparent intention to create a chip to do processing on the devices, rather than in servers, reflects Apple’s philosophy on AI. “Apple needs to do more AI on its phones than Google because of Apple’s strict privacy policy,” he said. “For instance, when Apple does face recognition today, the face isn’t shipped to the cloud like Google does, it’s done on the device. This protects the consumer’s privacy. Therefore, offloading AI from the GPU and CPU is a good thing and more important to Apple.”

It’s not yet clear when the Apple chip would be ready, but it’s possible the company could discuss it at its Worldwide Developer Conference in early June.

Photo: JESHOOTS/Pixabay