Smartphones are getting increasingly powerful and there don’t seem to be many apps that can choke modern flagships. While the industry is looking towards even higher scores on benchmark apps, Google has reportedly seen the capable chips in a different way, that might actually bring value to users’ regular usage. You may already know that the Google Pixel 8 Pro and the Galaxy S24 series have some innovative and helpful AI features. Google even stressed the Pixel 8 Pro’s capability to run some relatively large AI models on-device.
Web page summarization could soon work on-device for the Pixel 8 Pro
Although Samsung’s Galaxy S24 series and the Google Pixel 8 Pro have access to many of those AI features from Google, a lot of them currently run on Google’s cloud servers and the list of on-device ones is “quite short”. Now Mishaal Rahman from Android Authority has discovered some evidence that implies that Google is working on running the web page summarization feature on-device for the Pixel 8 Pro.
The AI model that powers many of these AI features is Google’s Gemini. Gemini Nano, which is the smallest of the three Gemini versions (Gemini Nano, Gemini Pro, and Gemini Ultra) is actually the one that runs on-device on the latest flagships and this could be one of the reasons why the features are exclusive to the latest models.
Now, the Gemini Nano’s small size also implies its limited capabilities. This model is reportedly only capable of working with “small blocks of text”, although it is enough to summarize transcriptions of up to 15 minutes in the Pixel recorder. It should be enough to apply the web page summarization feature to most of the articles you read on the internet.
The Gemini Nano currently powers three AI features on the Google Pixel 8 Pro – generating summaries of the recordings, Gboard’s smart replies while chatting on WhatsApp, and message rewriting suggestions on Google Messages. On the Samsung side, the model runs many Galaxy AI features as well.
Running the AI features on the device should reduce the load on Google’s servers
Another reason behind not many smartphone models getting access to Gemini-based AI capabilities may be associated with the cost of running and maintaining the servers. If you think about YouTube, they have advertisements or subscription plans for you that also pay for those costs. The Galaxy S24 series devices have already been in discussions regarding a potential fee for the future to access the server-side AI features. However, that business model regarding the AI features has received a different response.
Running one of the Gemini-powered features on-device also essentially means utilizing a phone’s computing power that you own and manage. While for you, it may result in a faster response as it does not have to send and receive data through the internet as you use the feature. So, given your device can run the model, it’s worth running that on-device.
2024-03-15 15:07:59