Apple tested Amazon’s artificial intelligence chips to see if they are able to train their language patterns. The Cupertino company said it is evaluating Amazon’s new Trainium2 chip for pre-training tasks on its own models. Apple added that initial tests have yielded positive results for effectiveness.
According to CNBCApple took an important part in the AWS Reinvent conference organized by Amazon Web Services. Benoit Dupin, director of machine learning and artificial intelligence at Apple, said that the technology uses the AWS cloud to support services such as Siri or Apple Maps.. However, the announcement that has caused a stir is related to artificial intelligence chips developed by Amazon.
Dupin stated that Apple is in the early stages of evaluating Trainium2a chip designed specifically to support larger generative AI workloads. The manager added that the first numbers show Increase efficiency by 50% thanks to pre-training.
The relationship between Apple and Amazon is not new. Dupin said that Apple has relied on AWS to improve its services for a decade.. Similarly, the technology company used chips such as Inferentia and Graviton for search services. This makes training with Trainium2 a logical step for Apple, which was one of the first to test Amazon’s new chip.
This is Trainium2, the Amazon chip that Apple uses to train its artificial intelligence models.
The Trainium2 (Trn2) chip is hardware designed specifically to accelerate the training of large-scale language models (LLMs). According to Amazon, Trainium2 uses an MPU-based architecture. able to effectively perform mathematical operations required during the learning phase. chip supports models with billions of parameters and uses low-power methods which guarantee energy efficiency.
“Trn2 offers 30-40% better value for money than current generation GPU-based EC2 instances. A single Trn2 instance combines 16 interconnected Trainium2 chips with NeuronLink’s ultra-fast, high-bandwidth, low-latency inter-chip interconnect to deliver 20.8 maximum petaflops of calculations, ideal for training and deploying models with billions of parameters.”
Apple isn’t the only one using Trainium2 to train AI models. Amazon confirms it has entered into an agreement with Anthropic use the chips on Claude. The technology company said it is building a cluster called Project Rainier that will expand Claude’s training with hundreds of thousands of Trn2 chips connected to EFA networks.
Using Trainium2 to power Apple’s AI makes perfect sense. In addition to partnering with AWS, the technology company has bet on chips developed by its competitors. We see the most obvious example in the example Google processors to teach the Apple base model family.
Source: Hiper Textual
I’m Ben Stock, a highly experienced and passionate journalist with a career in the news industry spanning more than 10 years. I specialize in writing content for websites, including researching and interviewing sources to produce engaging articles. My current role is as an author at Gadget Onus, where I mainly cover the mobile section.
