ChromeThe team announced that it is exploring the concept of "built-in AI", which isBrowserThere will be built-in AI models to enable on-device AI for browser features and web platform APIs.
According to the official introduction, with built-in AI, your website or web application will be able to run various AI tasks for basic models and expert models without worrying about deploying and managing the above models. In particular, Chrome will launch Gemini Nano, the smallest version in the Gemini model family, runs natively on most modern desktops and laptops equipped with Chrome.
It is understood thatup to dateChrome hasClosed betaIt has a built-in Gemini Nano large language model (4bit3.25B), which can be used for summarization, translation, writing and other tasks, and can run completely locally and offline for free. Join the early preview program to experience the built-in AI API in its early stages
Benefits of built-in AI for web developers
Compared to building your own model in on-device AI, built-in AI has the following advantages:
Easy to deploy: When the browser distributes the model, it takes into account device capabilities and manages updates to the model. This means you are not responsible for downloading or updating large models over the network. You do not need to deal with storage eviction, runtime memory budgets, transmission costs, and other challenges.
Use hardware acceleration: The browser's AI runtime is optimized to take full advantage of available hardware, whether it's the GPU, NPU, or falling back to the CPU. As a result, your app getsoptimalperformance.
Benefits of running on device
Using built-in AI methods, it becomes easy to perform AI tasks on-device, which in turn provides the following advantages:
Process sensitive data locally: On-device AI can improve your privacy story. For example, if you process sensitive data, you can provide AI capabilities to users with end-to-end encryption.
Smooth user experience: In some cases, not having to make a round trip to the server means you can deliver near-instant results. On-device AI can be the difference between a viable feature and a suboptimal user experience.
Greater access to AI: Your device can take on some of the processing load in exchange for greater access to features. For example, if you offeradvanced If your product already has AI features, you can use on-device AI to preview those features so that potential customers can see the benefits of your product without paying extra. This hybrid approach can also help you manage inference costs, especially for common user flows.
Use AI offline: Your users can use AI capabilities even when there is no internet connection. This means your websites and web apps can be used offline or with variable connectivity and function as expected.
Application website:
https://docs.google.com/forms/d/e/1FAIpQLSfZXeiwj9KO9jMctffHPym88ln12xNWCrVkMY_u06WfSTulQg/viewform?pli=1
Official documentation:
https://developer.chrome.com/docs/ai/built-in?hl=zh-cn