|
|
To sum up, it is probably this ability: understand user intentions, provide user operation plan suggestions, generate marketing ideas, automated execution, automated supervision, and automated attribution. 2. Data is not a barrier. Feed the data as context to the big model so that the big model can understand you better and provide a differentiated experience. When we first built the I product, we started from the user's perspective and provided scenario functions to solve specific problems; secondly It is possible to see the data accumulated by users in various scenarios, such as videos, articles, and questions asked.
These are users' private data that LLM does not have. Judging from past experience, data is definitely not a barrier, but once you have the data, pass it to the LLM as context so that the LLM can understand you better. This differentiated expe Armenia WhatsApp Number rience is a barrier. For example, asking LLM to help you come up with a title is editing, but if you give LLM the product information together, the effect will be different. All applications that are not willing to be just I tools need to consider the I assets accumulated in the data, so that products can serve users more personalizedly based on assets. 3. What data should be accumulated to make I products? The data that needs to be accumulated is not the
question-and-answer pairs users have had in solving a certain scenario problem in a recent period of time (such as within a month), such as how to set up store coupons on Xiaohongshu. ——It requires enough best practices to be compiled into question and answer pairs. So in the end, I may not be a company that sells software systems, but may be an operator of intelligent operation systems. Through countless best practices and summaries of best practices, I will continue to input and operate this system to make it become more and more intelligent. . Big manufacturers have a lot of data, but they are not really strong
|
|
發表於 2024-2-20 18:55:05