Apple says its artificial intelligence models were trained on Google’s custom chips | Wilnesh News

apple The artificial intelligence models that power Apple Intelligence’s artificial intelligence system are pre-trained on processors designed by Google, it said on Monday, a sign that big tech companies are looking for alternatives. Nvidia When it comes to training for cutting-edge artificial intelligence.

Apple’s Choice Google’s The domestic tensor processing unit (TPU) used for training is in A technical paper just published company. Separately, Apple also released a preview version of Apple Intelligence for some devices on Monday.

Nvidia’s expensive graphics processing units (GPUs) dominate the market for high-end AI training chips and have been in such high demand over the past few years that it has been difficult to procure the quantities needed. Open artificial intelligence, MicrosoftAnthropic all use Nvidia GPUs in their models, and other technology companies, including Google, Yuan, Oracle and Tesla They are being snapped up to build their AI systems and products.

Meta CEO Mark Zuckerberg and Alphabet CEO Sundar Pichai both made comments last week suggesting that their companies and others in the industry may have overinvested in artificial intelligence infrastructure, but acknowledged that the business risks of not doing so were too great. high.

“The disadvantage of falling behind is that you won’t be able to master the most important technologies in the next 10 to 15 years,” Zuckerberg said at a conference. podcast With Bloomberg’s Emily Chang.

Apple makes no mention of Google or Nvidia in its 47-page paper, but says its Apple Foundation Model (AFM) and AFM server are trained on a “cloud TPU cluster.” This means Apple rents servers from cloud providers to perform calculations.

“This system enables us to efficiently and scalably train AFM models, including on-device AFM, AFM servers, and larger models,” Apple said in the paper.

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

Apple was later than many of its peers to announce its artificial intelligence plans, and soon after OpenAI launched ChatGPT in late 2022, Apple loudly embraced generative artificial intelligence. The system includes several new features, such as a new look for Siri, better natural language processing, and artificial intelligence-generated summaries in text fields.

Next year, Apple plans to launch features based on generative artificial intelligence, including image generation, emoji generation, and an enhanced version of Siri that can use users’ personal information and perform actions within apps.

Apple said in Monday’s paper that the AFM on the device is trained on a single “slice” of 2,048 TPU v5p chips working together. This is a state-of-the-art TPU, first launched in December. According to the paper, the AFM server was trained on 8,192 TPU v4 chips configured to work together as eight slices over the data center network.

According to reports, Google’s latest TPU will cost less than $2 per hour if booked three years in advance. Go to Google website. Google first launched TPUs for internal workloads in 2015 and made them available to the public in 2017.

Still, Google remains one of Nvidia’s largest customers. Google uses Nvidia’s GPUs and its own TPUs to train artificial intelligence systems, and sells access to Nvidia’s technology on its cloud.

Apple has previously said that inference, which is taking a pre-trained artificial intelligence model and running it to generate content or make predictions, will occur in part on its own chips in Apple data centers.

This is the second technical paper on Apple’s AI system, following the release of a more general version in June. apple says then The company uses TPU when developing artificial intelligence models.

Apple is scheduled to report quarterly results after the market close on Thursday.

watch: How generative AI’s huge electricity consumption is overburdening our grid