How Apple used help from Google to train its AI models

Apple’s chief executive, Tim Cookannounced at a conference on Monday (10) an agreement with OpenAI to include your powerful artificial intelligence (AI) as part of your Siri voice assistant.

But in the fine print of a technical document that Apple published after the event, the company makes clear that the Google, from Alphabethas emerged as another player in Apple’s efforts to catch up in the field of AI.

To build Apple’s base AI models, Google engineers used their own framework software with a variety of hardware, specifically their own graphics processing units (GPUs) and chips available only in Google’s cloud, called graphics units. tensor processing (TPUs).

Google has been building TPUs for about ten years and has publicly discussed two versions of its fifth-generation chips that could be used for AI training; the fifth-generation performance version offers competitive performance over Nvidia’s H100 AI chips, according to Google.

Google announced at its annual developer conference that a sixth generation will be released this year.

The processors are specifically designed to run AI applications and train models, and Google has been building a cloud computing hardware and software platform around them.

Apple and Google did not immediately respond to requests for comment.

Apple did not discuss the degree of dependence on Google’s chips and software compared to hardware from Nvidia or other AI vendors.

However, using Google’s chips typically requires the customer to buy access to them through its cloud division, much like customers buy computing time from Amazon.com’s AWS or Microsoft’s Azure.