2025, DeepSeek brings the first year of AI application!

A few random thoughts, record them down:

1. In 2025, big models will become the core driving force behind products, and big models themselves will no longer be products. Therefore, 2025 is the year of full acceleration of hardware + AI, or the full outbreak of software and hardware integrated products that can be used directly, and everyone has to race to get ahead and take a position first. AI hardware products that directly reach users or customers, as well as empowerment base products that empower applications, accelerate hardware interactive experience, AI fluency (to the point of moistening things silently), emotional value, and provide data storage analysis feedback closed loops, will have great opportunities!

This means that two products will explode:

1. AI products that directly face users or customers;

2. As enabling base products for chips, various functional modules, accelerators, etc., a large number of outstanding companies will emerge!

2. The difference between DeepSeek and OpenAI is a bit like the relationship between a teacher and a top student:

The gentleman is: take every question you ask seriously, and replay his understanding of you to you repeatedly, sometimes replaying far more than you think, and then he will explain it to you layer by layer based on your question, "Based on my understanding of you, do you think my answer is okay? There may be more ideas, we can continue to discuss in depth." The whole process is very immersive and enjoyable, and it feels like fighting with a master. He respects all your questions, no matter who you are, he will explain them in detail, in layers, and in depth. If you think he has a problem, he will immediately analyze and optimize himself, and give you a more in-depth and thoughtful result. The logical level is very strong, and he will unravel the mystery and tell you in a leisurely manner, just like communicating with a real person, and the result far exceeds your expectations.

A top student is one who is knowledgeable and has a good memory, but weak in logical reasoning. No matter what you ask him, he can give you an answer with certainty and you have no idea whether he is right or not, or how he came up with the answer. He makes you feel that he is very good, but you dare not use his answer, and you dare not even copy it, because he knows everything and speaks clearly, but he has no depth, no thinking, no process, and he just pursues a result.

Having a teacher who is a top student is superficial.

DeepSeek's autonomous thinking chain, the entire process of decomposition-planning-problem-reflection-reasoning, actually helps AI applications move from simple knowledge base learning interaction to the analysis and decision-making stage. Company bosses can really introduce it into their enterprises, and individuals can also explain why.

Therefore, the emergence of DS not only makes AI products popular and recognized in China like "Internet celebrities", but also provides consulting companies and solution integrators around Deepseek applications with a large number of business opportunities and feasible markets in 2025. A large number of small vertical models based on various industries and companies will appear. 2025 is destined to become the first year of AI application!

All industries will actively or passively use Deepseek to deploy vertical private domain models. Hardware + Deepseek to form real AI products will be the biggest industrial opportunity and venture capital opportunity, and this biggest opportunity will first explode in China. DS is really amazing!