The emergence of chatGPT, the hundred model war of various manufacturers at home and abroad has kicked off, so do you need a product manager in the future? The author of this article raises such a question worth thinking about, let's take a look at the author's opinion!
Since the launch of chatGPT at the beginning of this year, the 100-model war of domestic and foreign manufacturers has begun. Just this week, Baidu released Wenxin Yiyan 4.0, CEO Robin Li bluntly said: This is the strongest Wenxin big model so far, and the comprehensive level is no less than GPT-4. In the environment where big models are "prevalent", will product managers be needed in the future?
1. Wenxin 4.0: The era of AI native, a new era of human-computer interaction
This is the strongest Wen Xin big model to date. It has achieved a comprehensive upgrade of the basic model, and has significantly improved the four major abilities of understanding, generation, logic and memory than the online version of Wenxin Yiyan, and the comprehensive level is no less than GPT-4 - Baidu CEO Robin Li announced the launch of Wenxin 4.0 at the Baidu World Conference.
1. Four major capabilities have been improved in a straight line
At the press conference, CEO Robin Li showed us cases of the four abilities of understanding, generation, logic and memory, for example, in terms of comprehension ability, ask Wenxin 4.0 on the case of provident fund long-distance loan policy, Wen Xin's words can deal well with complex prompt words such as out-of-order, ambiguous intentions, and subtext; In terms of logical ability, he can also easily solve Olympiad problems and summarize knowledge points. Overall, 4.0 is indeed a significant improvement over 3.5.
Here we will not compare how Wenxin 4.0 and GPT4 perform on various tasks, interested children's shoes can take a look:
https://mp.weixin.qq.com/s/Gb88_7c1aF1A8d2vR_ku-w
2. Baidu GBI – the first generative business intelligence product
As a product manager, I am more interested in the complete product landing based on the large model - Baidu GBI, the first generative business intelligence product in China, GBI applies the Wenxin large model in data analysis scenarios, supports natural language conversational interaction, performs data query and analysis and other tasks, and truly realizes the dream of "data analysis with your mouth".
In addition, GBI also has the ability to learn, we can add the expertise of the pendant category, so that it can understand the "black language" of certain industries and become an industry expert.
Second, think: Will AI replace product managers in the future?
In the face of the super understanding and generation ability of large models, I believe you all have the same question, do you need a product manager in the future? In my opinion: needed! The future is a need for product managers who can use AI.
So today let's try to use GPT to "turn a sentence requirement into" a complete PRD.
1. Case: From one-sentence requirement to full PRD
We often encounter certain requirements that are only one or two sentences described, and it takes a lot of brain cells for product managers to complete the requirements description. Here we assume that a sentence from a certain big guy is required: give the customer a few options in the input box.
1) Output the basic framework
As product managers, we can complete the requirements of this sentence by ourselves: in the input box (of the conversation page), (according to the user's current input), (in the form of a drop-down box) to provide customers with (5) related recommendation questions. The next step is to let GPT output the first draft of the PRD step by step:
Enter Prompt: Suppose you are a [Product Manager], and the product you are currently responsible for is [an intelligent conversation customer service bot]. Now plan [In the input box of the dialog page, according to the user's current input, provide customers with 5 relevant recommendation questions in the form of a drop-down box]. Please output the requirements document, which needs to include [requirement background, function description, data embedding, acceptance criteria] and other dimensions.
Output (GPT3.5 as an example):
2) Complete the details
Leaving aside the legibility of text descriptions, there may be a need to optimize from the documentation itself:
- Problem 1: Lack of flowcharts.
- Problem 2: The data embedding point lacks details such as trigger conditions for the buried point.
(1) Generate a flowchart
Input Prompt: Okay, then please add the relevant flowchart and output it in mermaid format.
Output (GPT3.5): The following is a simplified flowchart created using the Mermaid format that describes the flow of recommendation questions related to the input box of the intelligent conversational customer service bot:
graph TD
Subgraph user actions
A [User Input Problem] –> B [System Analysis User Input]
B –> C {match}
C –>|Yes|