Post by account_disabled on Feb 25, 2024 1:58:38 GMT -5
There may not be many opportunities to start a business today, but there are still many great opportunities to do business. The times are calling for more full-stack technology and full-stack products, and the good times for individual developers are finally back. The content of this article focuses on the following issues: Several stages of AI products Data is not a barrier, combining data to provide differentiated experiences is a barrier What data should be accumulated to make AI products? How to ensure that the current investment in application development is not in vain? Perplexity’s Product Design Capabilities Adobe’s cancellation of acquisition of Figma is a change in human-computer interaction When will AI’s capabilities undergo a leap forward?
Product managers need to Back to the Future Several stages of AI products: AIGC → Copilot → Insight → Agent (automation) By building a beggar version of the product, we understand user intentions, and then provide Colombia Phone Number List AI training with customers’ best practices and success manuals, delivery cases, and customer service knowledge bases in the past few years to see if we can provide suggestions for user operation plans. To sum up, there are probably these capabilities: understanding user intentions, providing user operation plan suggestions, generating marketing ideas, automated execution, automated supervision, and automated attribution. Data is not a barrier. Feeding data as context to the big model,
so that the big model can understand you better and provide a differentiated experience is a barrier. When we first build AI products, we start from the user's perspective and provide scenario functions to solve specific problems; Secondly, you can see the data accumulated by users in various scenarios, such as videos, articles, and questions asked. 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 experience 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.
Product managers need to Back to the Future Several stages of AI products: AIGC → Copilot → Insight → Agent (automation) By building a beggar version of the product, we understand user intentions, and then provide Colombia Phone Number List AI training with customers’ best practices and success manuals, delivery cases, and customer service knowledge bases in the past few years to see if we can provide suggestions for user operation plans. To sum up, there are probably these capabilities: understanding user intentions, providing user operation plan suggestions, generating marketing ideas, automated execution, automated supervision, and automated attribution. Data is not a barrier. Feeding data as context to the big model,
so that the big model can understand you better and provide a differentiated experience is a barrier. When we first build AI products, we start from the user's perspective and provide scenario functions to solve specific problems; Secondly, you can see the data accumulated by users in various scenarios, such as videos, articles, and questions asked. 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 experience 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.