Technology Sharing

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

ChatGPT was born with a bang at the beginning of the year, leading a new wave of AI - big models. To be honest, some people have been talking about big models since last year. At that time, big companies were all doing it internally, such as Google, Facebook, Tencent, Alibaba, etc. I was actually opposed to it at the time. Every company burned so much money, manpower and material resources to build a big model, and it was likely to be a mess in the end. Then the startup OpenAI released ChatGPT and led the way. This also reveals some rules. Disruptive innovation is likely to be born from startups, because big companies are too comfortable and everything they do is assembly line. Every major technological change is a war, and bubbles, scams, and burned money are inevitable. Then I have to reflect on why I couldn’t realize that this was a disruptive innovation like Teacher Howie did when he was working on gpt3.

After ChatGPT came out, it is not advisable to make large models. Unless you are a Musk-level player, it may be a waste of time. I am optimistic about the vertical field model. Today, OpenAI's dev day revealed the GPT Store (similar to Apple's App Store, openAI is really like Apple...)

This means that the time has come to start developing vertical models based on a specific large model.

I recently talked with big names in Silicon Valley and academia. They all use GPT4 and Github Copilot as daily development tools. Then I recently learned a set of prompt engineer courses by Professor Andrew Ng, which is very friendly to beginners. The most complicated programming is to define functions and for loops. But the most difficult part is the prompt part, which is obviously not at the same level as the few words we use in ChatGPT every day. It is even more difficult than writing code normally. (The red part is the prompt part) So you have to use it in conjunction with the tools generated by the prompt, or save some commonly used templates.

This lesson simply introduces the basic functions of the big model to give you an idea of ​​how it works. I have worked on NLP projects before, and overall it is much more convenient than the traditional machine learning process. Since I am a quant trader, I can gradually transfer data analysis to the big model, no rush.

Finally, should I use API or ChatGPT? Because I am currently a user of large models, not a developer, and his GPT Store has just started, so I will use ChatGPT first. Price is also an important factor. According to a friend who uses the API, he now spends several hundred yuan per month. The charging strategy used by OpenAI is similar to Amazon Cloud. You think it is very convenient and cheap at the beginning, but you will feel heartache when you pay the monthly order.

So my attitude towards big models is to communicate more with people in the industry, not to fall behind, and see how to better integrate with production. Then learn while using, and don't fomo~

AI big model learning benefits

As a warm-hearted veteran of the Internet, I decided to share my valuable AI knowledge with everyone. How much you can learn depends on your perseverance and ability. I have shared important AI big model materials, including AI big model entry learning mind maps, high-quality AI big model learning books and manuals, video tutorials, practical learning and other recorded videos for free.

Due to limited space, only part of the information is displayed. You need to click the link below to get it.

2024new CSDN Big gift package: AGI Large model learning resource package》Free sharing

1. Full set of AGI large model learning route

A learning journey in the era of AI big models: from basics to cutting-edge, master the core skills of artificial intelligence!

How to get into the big model?

To fully understand the big model, you must first understand the current status of the LLM big model on the market, learn Python language, prompt prompt engineering, and then deeply understand Function Calling, RAG, LangChain, Agents, etc.

Many people don’t knowIf I want to learn big models by myself, what route should I take?

The latest and most comprehensive resources for all large models, including [Learning Roadmap], [Supporting Self-study Video + PDF], [Interview Questions]I have sorted it out for you here.Get it via the card below!

"Artificial Intelligence Big Model Beginner's Learning Gift Pack", you can scan the QR code below to receive it for free!

1. Growth roadmap & learning plan
To learn a new technology, as a novice you must first learn the growth roadmap. If the direction is wrong, all your efforts will be in vain.
For those who have never been exposed to network security, we have prepared a detailed learning growth roadmap & learning plan for you. It can be said to be the most scientific and systematic learning route. You will have no problem following this general direction.

2. Video Tutorials
Many friends don’t like obscure texts, so I have also prepared a video tutorial for you. There are 21 chapters in total, and each chapter is a condensed version of the essence of the current section.

3.LLM
LLM (Large Language Model) is the most popular and concerned one

Development prospects: Big models have a wide range of applications in natural language processing, image recognition, speech recognition and other fields. With the advent of the big data era, big model technology will continue to develop and provide more development opportunities for programmers.

Skill requirements: To become an excellent large model programmer, you need to have the following skills:

• Master the knowledge related to deep learning, such as neural networks, convolutional neural networks, etc.;
• Familiar with programming languages, such as Python, C++, etc.;
• Understand big data processing technologies, such as Hadoop, Spark, etc.;
• Have a good foundation in mathematics and statistics to better understand and optimize large models.