On this basis, we can refine the limitations of large models into the following Large models cannot obtain realtime information and are completely unaware of the socalled news hot spots. This is very unlike humans. The knowledge mastered by large models is limited. If you encounter problems that you don't understand, you will start to make up random things, which is a headache. sufficient reasoning capabilities (at least not yet), and their accuracy is questionable when dealing with more complex problems, such as solving complex mathematical problems.
Large models cannot obtain private Guangdong Mobile Phone Number List documents of enterprises (RAG can solve this problem to a certain extent). The large model cannot obtain business data in the external system database, such as maprelated information, Zhang San’s consumption records on a certain platform, etc. We will find that no matter how the large model is trained, these limitations will still exist and will not be solved as the large model continues to be trained.
If you think of a large model as a human being, no matter how smart a person is, it is impossible to receive and digest new knowledge in real time, obtain all private data, and solve all problems alone. So how can we make large models reliable and reliable? How about powerful? The answer is Find reliable people to work with to solve complex problems . What kind of product manager is needed on the Bside? Bside product managers aim to improve the work efficiency of the supply side, so Bside needs are mainly oriented to business issues.