March 28th.MicrosoftAfter inviting Copilot Pro users to test in December 2024, non-Pro users in the U.S. are now invited.Tested in Microsoft Edge browser Copilot Vision Function.
1AI cited the results of a test conducted by tech media Windows Latest, which showed that although the function can recognize page content, there are obvious defects such as response interruptions, limited interaction, and a narrow field of view, which is only suitable for extracting basic information and does not yet have practical value.
The Copilot Vision feature, previously available only to paid Copilot Pro users, has been extended to free users in the U.S., but will need to be invoked from the Edge browser sidebar.
The procedure is that the user searches for "Copilot Vision" in Bing and clicks on the matching results, accepts the terms and conditions, and then activates the "Glasses" button via the voice icon in the sidebar, and the button is highlighted to indicate that the current web page can be scanned.
The media outlet found that the feature is still "half-baked", and during the test, there were many times when the answer was interrupted or stuck in a loop, and it was necessary to repeat the question in order to get an incomplete answer. For example, when asked to describe the content of a web page, Copilot stopped responding for 15 seconds and then suddenly switched answers.
In addition, in terms of interaction, it is not possible to perform actions such as clicking or playing videos, and even rejects the user's command to "stop talking", stating that it is "unable to access elements of the page", and that it can only recognize content in the visible area of the screen.
In terms of information processing, it relies on the existing text on the page, is unable to actively retrieve web data, and is unable to correlate previously recognized relevant information after scrolling the page.
According to the media, the current Copilot Vision is more like a "half-baked" product, with problems such as narrow scanning range and confusing logic that significantly reduce its usefulness. Microsoft needs to optimize core features such as scrolling and cross-page searches to realize the potential of the AI assistant. It may be free to collect user feedback, but at this stage it's only recommended for simple page summaries.
