On Tuesday local time, Microsoft officially celebrated the historic year of Copilot (formerly known as Bing Chat/Bing chatbot). At the same time, the official also drew a set of pie for global users - revealing the features that Copilot has just launched or will soon be launched in 2024.
For the name Copilot, most users may be a bit confused. Bing Chat was first launched on February 7th this year and is a chatbot similar to ChatGPT, using the same underlying technology.
Subsequently, Microsoft's product managers and engineers rapidly popularized AI functionality to Edge browsers, Windows 11 operating systems, and Microsoft 365 subscription services at an extremely crazy pace. Faced with the proliferation of various product lines, Microsoft announced at the Ignite Developer Conference two weeks ago that all Bing Chat services would be renamed Copilot.
Take a look at the new features
Microsoft reiterated on Tuesday that Copilot will use its partner OpenAI's latest GPT-4 Turbo model in the coming weeks. In addition, Microsoft is also working hard to integrate the multimodal capabilities of GPT-4 with its image search business, and this new feature will also be launched soon.
Microsoft also revealed that it will soon launch a "rewrite" menu for Edge browser users. In theory, users can select text content from most web pages and command AI to "rewrite a copy.". In addition, Code Interpreter, which is used for complex processes such as data analysis, programming, and visualization, has also entered the internal testing phase.
Meanwhile, the image generation function based on the DALL-E 3 model has been launched. After a brief trial, although there are still visible flaws, there has been significant progress in understanding text and image expression.
In addition to these features that most users have already seen with OpenAI, Microsoft has also drawn a new pie today - Deep Search.
Microsoft introduced that this feature is to help search engines better understand "what users really want to search for.". Unlike existing keyword searches, deep search provides options for deeper and richer exploration of internet knowledge.
Simply put, it means utilizing the comprehension ability of GPT-4 to transform simple questions into more systematic and comprehensive descriptions. At the same time, based on the ranking system of Bing search, provide a more accurate ranking of search results that meet the requirements.
Microsoft's example is that when asked how Japan's loyalty card credit system (consumer credit system) operates, the system greatly enriches the description of this question through the GPT-4 model.
In this example, the above question would become: Provide an explanation of the working methods of various loyalty card programs in Japan, including the benefits, requirements, and limitations of each program. Examples of popular loyalty cards from different categories, such as convenience stores, supermarkets, and restaurants. Compare the advantages and disadvantages of using loyalty cards with other payment methods in Japan, including current rewards and benefits. Highlight the most popular services and participating merchants.
Meanwhile, in many daily search scenarios, users often can only provide more vague descriptions - such as "how Japan's points system operates.". At this point, deep search will also list potential different possibilities, including loyalty card points, high skilled talent immigration points, public transportation fees, etc., allowing users to choose to expand on their own.
Microsoft emphasizes that due to the complexity of deep search, performing a deep search may take up to 30 seconds. Technology giants also firmly declare that such waiting is worth it.
Easter egg: Based on its usage since its launch in February this year, Microsoft has summarized a batch of the most popular AI prompt word use cases among consumers.