In the face of the learning pressure brought about by the rapid iteration of AI, the author shares three effective learning methods: focusing on personal goals to focus, improving core competencies in collaborating with AI, and learning in action through rapid practice. These methods help authors maintain learning efficiency and reduce anxiety amidst the AI wave.
Recently, I talked to friends in the AI circle, and everyone intuitively felt that the speed of AI evolution this year is super fast, much faster than in 24 years, and various new AI products come out every day.
The speed of AI producing new products is faster than the speed of learning AI yourself, which is okay, and you can’t finish learning at all!
Every time I talk about this, I can feel that everyone’s anxiety is about to overflow. I have also experienced this stage in the past two years, every time AI comes up with something new, I want to follow it, but I find that in fact, a person’s attention is limited, and if you have to do anything, you must not be able to do anything well in the end.
So I began to compress my attention, do as little as possible, do things deeper, and gradually explore my own learning methods, and now I can be a little less anxious under the wave of AI~
In today’s article, I would like to share with you my learning methods for your reference~
I hope it can help everyone to be a little less anxious and surf under the AI wave more easily.
The core of this learning method is three things:
1. Focus on your goals and reduce the loss of attention
2. Do things that will make your model stronger
3. Practice quickly, and learn while doing is the most efficient
This set of learning methods requires you to first think clearly about what your goal is, what your core strengths are based on this goal, and then practice it frequently to increase your motivation and reduce anxiety through positive feedback.
01. Focus on your goals and reduce the loss of attention
Let me take myself as an example, last year when I first started learning AI, I was a hammer in the east and a stick in the west, anyway, I have to study and study any product, ChatGPT is easy to use to write copywriting, use chatgpt, AI drawing is easy to use, use AI drawing, and use it quickly when a new product comes out.
But in the end, you will find that you still don’t know how to use anything, you know every corner, but every piece is not done deeply enough, and then you can’t combine it with the company’s business, and you find that you have learned a lot but there is no result.
At that time, I thought about it for a while, thinking that this kind of learning was inefficient, or that I would learn something if the company’s business needed something, so I began to go down the line of the AI landing scene.
How can product managers do a good job in B-end digitalization?
All walks of life have taken advantage of the ride-hail of digital transformation and achieved the rapid development of the industry. Since B-end products are products that provide services for enterprises, how should enterprises ride the digital ride?
View details >
If the business needs to be a knowledge base, then learn how to make the AI knowledge base complete high-quality output; If the business needs copywriting automation, then study how to make AI achieve a 99% content production stability rate; If the business needs AI to empower the customer service team, then study how to apply AI+SaaS scenarios.
After making AI run a little smoother in business, I began to wonder if AI could empower me more.
So I began to research what scenarios I could combine with AI in my daily life, such as using AI to write prd, using AI programming to write scripts, and using AI to read papers.
Understand more AI landing scenarios to expand influence.
This is my own core goal.
So you will find that the content of my official account and my daily research are closely centered around the AI landing scene.
Around this goal, then I actually won’t follow the various new product information of AI, I just need to wait for everyone to finish the evaluation, and then take a quick look at the evaluation information to understand how this product is, and whether it can improve the effect of my existing landing scenes.
So I became an AI blogger who didn’t stay up late and couldn’t grab the first post, and the core reason was that the content had nothing to do with my goals.
I will write prompts, disassemble products, and research agents, and these things seem to be unrelated, but their ultimate points are all AI landing scenarios.
Focus on your own goals and reduce the loss of attention, so that you can do something deep and solid.
Many friends came to me after reading the various prompts I wrote and asked me, what idea did you use to come up with this prompt?
In fact, the answer is actually very simple: I need this prompt to help me solve the problem I am currently encountering.
When I write the thesis master prompt, I need to understand what Claude’s two papers are talking about, and by the way, I also need to prepare to do model training with our algorithm, a bunch of papers and new logic are waiting for me to learn there, I need an efficient learning partner.
I needed to let the AI be a mirror to find me enough information, and then it would present me with a variety of reference results, allowing me to make the final core judgment.
So how do you set goals for yourself?
The goal must be able to land and solve your real problems.
For example, a product manager sets the goals of AI learning, which can be these:
- Changed careers to become an AI product manager, promoted and raised
- Do a good job in the company’s AI business, and get a promotion and salary increase
- Let AI help you do more work, and you can have more free time
- Write a product that you like in your heart and explore for more possibilities
After reading it, do you think these targets are tacky, hahaha.
But these goals are clearly implemented and can solve problems, not “pseudo-goals” driven by anxiety.
What is a pseudo-target?
- Proficient in all types of AI products
- Become a person who will not be eliminated in the AI era
- Don’t be pulled down by everyone, keep a horizontal line
These goals don’t actually help you, but make you more anxious.
What is the problem you are solving at the moment, what is your goal?
02. Do things that will make the model stronger and you will become stronger
I myself took some detours when I used cursor last year, when cursor was not as versatile as it is now, it still needed people to be the core, AI to assist people to complete various programming, and finally to launch products.
With my programming knowledge, I encountered many problems that I couldn’t solve on my own, and there were actually two ideas in front of me at that time:
1. Learn some programming knowledge by yourself and see if you can solve these problems
2. Find a way to solve problems with AI to empower programming more efficiently
According to my current thinking, I will definitely choose 2, but at that time I thought that Composer was so bad, after all, it still had to be human-driven AI programming, so I should find a way to learn a little programming knowledge.
So I went to learn how to draw beautiful pages in Vue, how to use Docker, and so on.
The learning process was very unsmooth, I found that I had to learn a bunch of basics, and then slowly grind to understand what the principles were.
But this year’s cursor’s Agent model superimposed on the large model has brought earth-shaking changes, and the problems I faced before can actually be solved by the model, and there is no need to learn it myself at all.
So what is the core competitiveness of my cursor? This was one of the questions I was most confused about at the time, and I didn’t have an answer in my head.
Until I heard Mr. Xiao Pai’s sharing, he talked about the idea of making AI products.Do things that will make the model stronger as much as possible, so that the product will not be worthless because of the model iteration.
I suddenly realized why my cursor programming ability did not improve for a while, because I fell into a misunderstanding, and I went to study things that were meaningless when the model became stronger.
I quickly cut back my focus to think about how to use AI to solve problems, and I can see that the AI programming ideas shared with you now follow the principle that the stronger the model ability, the better the product effect developed.
What is “the model becomes stronger and the other becomes stronger”, let me give you a few examples:
- Collaboration with AI: Determine which can be solved by AI and which must be solved by humans
- Ability to ask good questions: No matter what AI evolves into, the quality of the results will always depend on the questioner
- Depth of understanding of business: This is the cornerstone of a good product, and a good product must have a deep understanding of the business
- Aesthetic ability: AI can draw many beautiful pictures, but what is good and what is suitable for this scene is up to people to judge
These abilities will not be replaced by models, they are all human areas of responsibility, and the stronger the model, the greater the value of these abilities.
03. Practice quickly, and learn while doing is the most efficient
For a long time before, my idea of learning new things was like this:
But this efficiency is actually quite low, I am looking for more and more tutorials, and I am doing less and less.
Now the idea is more to learn by doing, do things first, see what your problems are, and then find ways to solve them:
In this way, positive feedback also comes very quickly, and you can notice that you are improving every day, so you are more motivated to persevere.
For example, when I was making the “Prompt Management Assistant”, I didn’t read the API documentation of the multi-dimensional table at all, and I didn’t know how to connect the data between the plug-in and the Feishu multi-dimensional table.
So I started to pull the AI to start doing it, solving problems while doing it, and the problems I encountered in the middle were one after another, and I couldn’t think of it at all:
- Feasibility analysis of plug-in and Feishu multi-dimensional table data docking
- Find the Multidimensional Tables API documentation to solve the authorization problem
- How to reauthorize the token after 2 hours of authorization
- Authorization issues with safe link redirects
- Style compatibility issues between Dark Night Mode and Normal Mode
- Different browser plug-ins are installed
By making the “Prompt Word Management Assistant”, my understanding of Feishu multi-dimensional tables has become one step deeper, and I have also gone over to listen to more application scenario cases of Feishu multi-dimensional tables.
If I didn’t do it right away at the beginning, but planned and looked for tutorials first, then this matter would be far away.
Do it first, don’t be afraid of mistakes and failures, just iterate a few more times.
I have finished sharing three things with you, I hope this article can help you relieve your anxiety in the AI wave and grow faster~
