008-Taking advantage of a hot topic, a DeepSeek training course for ordinary people inside

Written by
Caleb Hayes
Updated on:June-29th-2025
Recommendation

Master DeepSeek and start a new era of AI reasoning.
Core content:
1. Introduction and advantages of DeepSeek-R1 reasoning model
2. Accessibility and application scenarios of DeepSeek model
3. General formulas and questioning skills for communicating with DeepSeek

Yang Fangxian
Founder of 53AI/Most Valuable Expert of Tencent Cloud (TVP)

Take advantage of the hotspot

DeepSeek has become very popular recently, so we also wanted to take advantage of it and launched an internal DeepSeek training course for ordinary people.


#1

What is DeepSeek?

Large Language Model

DeepSeek: is an AI model from China .

But what everyone praises about DeepSeek is that it is an inference model. That is what everyone has heard of as DeepSeek-R1 , and its counterpart is ChatGPT-o3 . ( DeepSeek also has a general model like DeepSeek-V3 , which can be compared with the Doubao and Tongyi models we mentioned earlier.) The full-blooded version that everyone has heard of is the version that can call on the maximum computing power.


The advantages of the reasoning model: mathematical deduction, logical analysis, code generation, and complex problem solving. After using it myself, I feel that her Chinese copywriting ability is also very strong.


If you don't need complex reasoning and logic, it is more efficient to use the general model directly.


--------------

There are differences between the inference model and the general model, so there will still be some differences when communicating with AI .

For the inference model, you can actually forget about the prompt words. (There will be some instructions on how to use them later.) Here is an example:


Reasoning model (some say forgetting the prompt words requires sincerity): For example, I am a self-media operator


General model: Let’s say I am a professional copywriter in an advertising company.


That is to say, for the general model, I need to create a character for him to inspire him to have a conversation in the way of thinking of this character. Professionally, it is called structural guidance, which is to make up for what is lacking. The reasoning model directly says, "Say what you want."



#2

Where can I use DeepSeek smoothly?

Large Language Model



1 ) Directly open DeepSeek official website, https://www.deepseek.com/, and use it about 5 times a day. The first time is basically ok , and then it will show that the server is busy.

2 ) Directly search with Metaso https://metaso.cn/?s=bdpc ; overseas (fq required) use https://www.perplexity.ai/ (the advantage of the search tool is that it can be queried online)

3 ) I also recommend a chat called Wenxiaobai https://www.wenxiaobai.com/chat; or the one that just came out last week is Tencent Yuanbao, which directly connects to the DeepSeek model https://yuanbao.tencent.com/ 


4) Huawei mobile phones can use Xiaoyi https://consumer.huawei.com/cn/mobileservices/celia/


5 ) Because it is an open source model, you can use some dialogue tools to call the API (this is more professional and requires some hands-on skills)



#3

How to communicate with DeepSeek?

Large Language Model

In the past, the problem people encountered was how to search, but now the problem people encounter is how to ask questions. The emergence of any technology has lowered the threshold. So asking questions to AI is not as complicated as people think.

There is a general formula ( 7 elements)


  1. who I am,


  2. What do I have (some background information on you),


  3. What do I want to do with this content?


  4. Where or for whom should I use it?


  5. What should I do with AI and what effect do I hope to achieve?


  6. In what format to output (how long is required),


  7. (I was worried about what).



However, it is not necessary to write it in full. You can use some of it flexibly.

The results of the inference model output are divided into two parts: A, the inference process; B, the result of the inference

I think watching the process is also a very good process of thinking training. I will use a Valentine's Day marketing copy as an example to demonstrate the results. I asked this question: I want to post a Valentine's Day post on my LinkedIn account on February 14th, Valentine's Day. My customers are mainly European and American customers, so what content and theme should I use to attract their attention?

What I learned from AI’s answer:
1. How should I present a theme like Valentine’s Day on a commercial social platform?
2. What is the purpose of my posting this content?
3. How should structured content be written?
4. What kind of emotions should be expressed?
5. What should we pay attention to?
6. He also gave me a summary.
7, examples are also given.
8. The example also includes how the visuals should be designed.
9, and also includes how much attention this plan can get (with data indicators).
10. Also told me how to guide the interaction.
11. Also tell me the time of posting.

In the past, tools helped us improve efficiency, but now AI has enabled us to go from not being able to be fast to being able to be fast. So in this era, the problem is productivity.

--------------


Some notes


  1. A. Over-complexity, over-simplification;      


  2. B. Give up if you are not satisfied with the output; (adjust the prompt word)      


  3. C. Multiple rounds of dialogue and guidance      


  4. D. Use critical thinking to cross-validate (illusion): Magic can be defeated by magic.     


  5. E. Notes on document transfer: If the document is over 50 pages, it needs to be processed separately. PDF files cannot be used if they are images.      

--------------


Some usage tips


Words that can be used:

  1. A.         / Continue : Automatically continue generating when the answer is interrupted


  2. B.         / Simplify : Convert complex content into plain language
  3. C./Examples require real-world examples         ( especially when writing code )


  4. D.        / Steps : Let AI guide the operation process step by step


  5. E. / Check: Helps you find errors in the document



Methods that can be used:


1 ) When AI writes something professional, you can reply with a kindergarten formula: Use words that an 8- year-old can understand xxxx ; Please use a way that non-technical people can understand xx xx

2 ) When you are not satisfied with the content or expression of AI , you can use a reference object or person to help you: please write me an article in the style of Lu Xun ; please write me an article in the style of [Fauvism]

3 ) Let AI help you analyze your concerns and potential risks. You can write: If xxxx, xxxx will not happen; or if xxxx, what will happen in the next xxxx time;

4 ) If you want to simplify, you can just say: delete all the correct nonsense and give practical suggestions;


5 ) If you want him to think more deeply from multiple perspectives, write: Please use a critical thinking method or write: Please review it xx times xxxx;