Follow me and remember to star⭐️ so you don’t get lost
The way humans interact with machines is undergoing a significant transformation, evolving from simple conversationsChatbot to autonomous Agents. Today, I will share some new developments from Zhipu in this area.
Zhipu, as a leading company exploring Agents, has achieved many remarkable results. First, let’s talk about AutoGLM, which can autonomously perform over 50 steps in long operations, like a super intelligent assistant. Moreover, it can execute tasks across apps, meaning it can navigate between different applications to accomplish various tasks for us.
The upgrade of AutoGLM allows it to tackle more complex tasks. For example, it can understand long instructions and execute them quickly, even faster than manual operations in multi-step and loop tasks. Additionally, its cross- App capability acts like a bridge, connecting different applications for smoother task execution. Don’t you find it cumbersome to give commands to smart assistants every time? AutoGLM solves this problem by supporting customizable phrases for long tasks. For instance, we used to say in detail “Help me buy a cup of Luckin coffee, coconut latte, Wudaokou store, large, hot, with a little sugar“, but now we can just say “Order coffee“.
Now let’s talk about the new experience on the Web, where Zhipu’s Qingyan plugin has launched the AutoGLM function, truly ushering in a new era of “fully automated” internet use. It supports autonomous operations on dozens of websites, including search engines, Weibo, Zhihu, Github, etc. For example, during a live demo, the Qingyan plugin automatically completed “Search for Mango TV on Baidu, open ‘Xiaoxiang Renjia’, play the latest episode, and send bullet comments on the ending“, all without human intervention. Isn’t that amazing?
In addition to mobile and browser applications, Zhipu also has new breakthroughs on the PC side. The GLM – PC is a significant technical exploration, based on Zhipu’s multimodal model CogAgent. Currently, the first phase of internal testing scenarios has been opened, featuring many practical functions. For instance, the meeting proxy can help us schedule and participate in meetings, and send meeting summaries, which is incredibly useful for those who frequently attend meetings. There is also a document processing function that supports document downloading and sending, as well as understanding and summarizing documents, making document handling much easier in the future. Furthermore, in terms of web search and summarization, it can search for specific keywords on designated platforms like WeChat Official Accounts, Zhihu, Xiaohongshu, etc., and complete reading and summarization, saving us a lot of time. Additionally, it has remote and scheduled operation capabilities, allowing us to issue commands remotely via mobile to let it autonomously complete computer tasks, or set a future time to execute tasks while the computer is on. There is also an invisible screen function, which allows it to autonomously complete work on an invisible screen while we work, freeing up screen usage rights.
However, we must also view these new technologies objectively. Although GLM – PC operates similarly to humans and theoretically can perform many applications designed for humans, the complexity of PC tasks and the nature of tasks we perform on PCs are often quite complicated. Currently, the capabilities of large models are still some distance away from truly replacing human office work. In the current version, users still need to input very precise commands. However, we must believe that with the continuous development and advancement of technology, these issues will gradually be resolved.
📬 Submission Channel: • Submission Email: [email protected]