The author also notes that, despite the WebNN spec outlining numerous exciting use-cases, the API itself does not appear to offer more than basic functionality such as loading, running, and performing operations on a model. They question whether there is another aspect of the API that would make these use-cases feasible in browsers.
Key takeaways:
- The WebNN API aims to enable running AI models directly in the browser.
- The concept of having AI in browsers is intriguing, but the idea of downloading and storing large models for each website or web app seems inefficient.
- The author suggests it might be more practical to have models built into browsers, with a web API providing standardized ways to interact with the underlying models of each browser.
- Despite listing out various use-cases, the WebNN API appears to only provide basic functionality such as loading, running, and performing operations on a model.