Are smartphones strong enough to 'shoulder' a giant AI model?

In early March, Google surprised everyone by announcing that only the Pixel 8 Pro would be equipped with the new AI model "Google Gemini". Meanwhile, the Pixel 8 and Pixel 8 Pro have quite similar configurations, Google does not allow the Pixel 8 model to support the use of the Gemini AI model, because the company believes that the reason why the device has not been used yet is because of the limitation issue. hardware processing.

This is quite confusing because Google originally positioned the Pixel 8 as a smartphone focusing on AI models. The company once developed an AI model for smartphones called "Gemini Nano", but in the end only Pixel 8 Pro benefited.

By mid-March, Google announced on the Pixel user support forum that the Pixel 8 would be supported with the Gemini Nano AI model through a quarterly Android update in June. However, it was inconvenient for users. Pixel 8 must enable this feature manually in the "Developer options" menu. This causes many users, especially those who are not tech-savvy, to miss this useful feature.

According to Google's vice president of software services and devices - Seang Chau, integrating AI models into smartphones will consume a significant amount of RAM. The reason is because AI models need to use RAM continuously to be ready to meet user needs at any time.

This is an issue that needs careful consideration, because the RAM capacity on smartphones is limited. Using too much RAM for AI models can affect the overall performance of the device, causing it to operate slowly.

However, Seang Chau also said that Google is trying to optimize the AI ​​model to minimize RAM consumption. At the same time, the company is also researching new hardware solutions to increase RAM capacity for smartphones.

Marvin Fry

Update 01 April 2024