According to a report by CNBC, Benoit Dupin, Apple’s senior director of machine learning and artificial intelligence, made an unexpected appearance at this year’s AWS re:Invent conference in Las Vegas.
During his talk, Dupin detailed how Apple incorporates Amazon’s custom AI chips into a variety of its cloud services. He also mentioned that Apple is assessing the potential of Amazon’s latest AI chipset to pre-train its Apple Intelligence models.
In his remarks at the event, Dupin highlighted that Apple has utilized AWS processors, such as the Graviton and Inferentia, for over ten years, enhancing services like Siri, App Store, Apple Music, and Apple Maps. This collaboration has reportedly resulted in a 40% increase in efficiency when compared to traditional x86 processors from Intel and AMD.
“We enjoy a robust partnership, and the infrastructure is both dependable and capable of supporting our customers globally,” Dupin noted. Prior to joining Apple in 2014, he held the position of vice president at Amazon’s A9 Search Technology group.
Dupin also affirmed that Apple is exploring the latest AWS AI training chip known as Trainium2. He indicated that Apple anticipates a “up to 50% improvement in efficiency with pre-training” using the Trainium2 chipset.
The application of the Trainium2 at Apple would focus on pre-training AI models rather than being utilized directly for Apple Intelligence functionalities.
Apple Intelligence features are driven either on the device itself or through the cloud, utilizing the company’s Private Cloud Compute platform. This particular infrastructure is based on Apple’s proprietary Silicon chips.
Earlier this year, Apple revealed in a research document that it leveraged Google’s Tensor chips for training its AI models. It was significant at the time as Apple chose Google’s Tensor processors over the more commonly used Nvidia chips by other companies.
Follow Chance: Threads, Bluesky, Instagram, and Mastodon.
: . More.
