Problems that Deepseek cannot solve will be addressed by Haichuan. There are several hypothetical questions; feel free to improve them or provide answers.
Thread Content
First of all, I invite all fellow sailors to participate. Those who haven’t been invited are asked to add me as a friend; it’s not possible to invite them until that is done: lol. I would like to extend a special invitation to @qugd, @pzhmotor, @Zhongyuanren, @slayerstar, @leo_0088, @huachan2006, @li*nji, @AChuan, @zhjnavy015, @wamj6566, @cjzhou2, @Cleaner, with the order of ranking taking into account precedence. On February 6, 2025, a user asked whether deep ceek could be used to assist in chemical process design There are various opinions on this topic; some approve of it, some question it, and some express anticipation... The author’s view for the first time is as follows: It is merely an advanced search tool. It works fine if a database is available, but it becomes difficult in the absence of one. In essence, it makes it easier and more precise to search for existing knowledge. Its main purpose is to address the problem of chaotic knowledge information caused by an overwhelming amount of advertisements. Designing such a tool requires practical production experience, knowledge of safe production management, and experience in various areas. Relying solely on the ability to retrieve existing knowledge is far from sufficient. For creative work like design, no artificial intelligence can replace humans in the short term. Published on 2025-2-14 14:38:42. The author subsequently presented several questions to which DeepSeek could not provide answers. To facilitate better communication, the author decided to create a separate post and hopes that other users will contribute to expanding on this topic, as well as providing more examples of problems that DeepSeek cannot solve. Of course, the goal is not to raise these questions merely to demonstrate DeepSeek’s current limitations. Instead, the intention is to maintain or restore the in-depth discussion atmosphere that existed on Haichuan in the past, to delve deeper into various issues and phenomena, and to offer more answers and possibilities. This will help engineers in various fields of the chemical industry to continuously improve themselves and solve existing technical problems, thereby promoting the development and progress of the industry. If there is an oxygen pipeline with a working pressure of 0.225 MPa and a DN150 diameter, what is the maximum flow rate it can handle? And if the flow rate is 2500 Nm3/h, what precautions should be taken during operation? DeepSeek’s response: The server is busy; please try again later. Author: If more specific engineering constraint parameters can be provided for the question, please feel free to do so. Answers are also welcome, as well as revised question conditions that lead to new questions and answers.
I’ve seen a few typical box flowcharts created in DS.
For box flowcharts, there’s no need to use intelligent AI, right?
DS is merely a tool for retrieval and information optimization; professional tasks require extensive expertise to handle and train, so it cannot serve as a substitute in the short term. Whether it can become a substitute in the future depends on market demand and costs
Feel free to share your own independent insights.
1. The minimum role that DS plays in chemical engineering design is that of a high-level professional mentor； 2. Making good use of DS can at least significantly improve work efficiency and provide more opportunities for personal growth ； 3. The ability of DS to solve problems depends on an individual’s level of proficiency; just as with consulting experts, one must know how to ask questions in order to get answers ； 4. DS certainly cannot fully replace everything that humans do, such as the experiential aspects of a process, the sense of achievement, and innovation and creativity.
This post was last edited by *nht1 on 2025-2-27 at 15:15. There is a bit too much admiration for mentors in this approach; it’s only possible to conduct some analyses and integrate existing knowledge and data, and it’s often ineffective when dealing with specific problems.
The foundation of AI should still be the summarization of existing knowledge; try asking it to innovate, or try asking it to make predictions
Currently, I use DouBao, Kimi, and ds together. What I’ve observed is that they can be useful as online resources for assisting with research in chemical engineering – for things like material information and established process routes. However, for the same question, their responses can sometimes differ, so it is necessary for engineers to evaluate them carefully.
Using the Tencent IMA+deepseek approach, after importing the knowledge base, it can provide quite accurate calculations and recommendations regarding specialized knowledge. For example, select the OH2 pump based on flow rate and head, and calculate the wall thickness based on pressure and size.