Learning Paths for Investing in AI

Source: DaoShuo Blockchain

During the online discussion on Saturday, some netizens left messages asking what books they can read, what magazines they can subscribe to, and what methods they can use to learn in the field of AI?

I think the study of AI should be based on our purpose.

For me, the purpose of learning AI is very simple: not to become an expert in this field, nor to make a living in this field in the future, but simply to understand the development of this field so that I can find suitable investment opportunities in this field.

To study for this purpose, I think the most important thing is to understand the logic of AI so that you can have a rough judgment on new things that will emerge in the field of AI in the future.

To understand the logic of AI, I think you can start with some books that introduce the basic principles of AI.

In this regard, there is a book online that even Altman recommended, "This is ChatGPT" (written by Stephen Wolfram).

This book starts with the simplest basic concepts and introduces the mathematical principles and working methods of large language models. As long as you know basic addition, subtraction, multiplication and division, you can read this book.

If you find it difficult to read, you can just read the first few chapters. You can get a general understanding of the principles of the large language model without reading the following chapters.

After understanding these principles, we can know why the training of large language models requires GPUs, data, and algorithms, and we can also know what role and in what aspects GPUs, data, and algorithms play in the training process of large language models.

If we think about it further, we will know what kind of optimization NVIDIA has made to the GPU in order to improve the training efficiency of large language models, why NVIDIA has acquired some small companies in history for those optimizations, and what those acquired small companies do.

Following this logic, I roughly understand why many so-called decentralized computing powers on the market are "pseudo-projects" - it's not that the direction of decentralized computing power is wrong, but that it is difficult to design an ideal decentralized computing power system under the NVIDIA framework.

To truly realize this kind of system, I think the design of GPU must be reconstructed. If we must use NVIDIA's framework to build such a decentralized computing system, the constructed system can only be regarded as an experiment or demonstration at best, and it is difficult to become a strong competitor to the centralized computing system.

Once we have an understanding of the basic principles of AI, there is no need to delve deeper into the mathematics. Next, I will focus on the application scenarios and development trends of AI. In this regard, I read "Inflection Point: Standing on the Eve of AI Subverting the World" by Wan Weigang.

The good thing about this book is that it is very imaginative and has basic logical support, which allows us to rationally speculate and imagine what the future world full of AI will look like.

Except for these two books, I haven’t read any other books specifically. The rest of the time, I basically read various articles on the Internet (such as WeChat public accounts and Twitter) and followed various new developments. Then we enrich and expand our understanding of AI based on the new information provided in these articles and updates.

For example, we know that the current ChatGPT is a large language model, which mainly trains AI to understand language. But human intelligence is rich and colorful. In addition to language, we have many other ways to perceive the world. Many articles in the field of AI will introduce the development of other types of AI, such as behavioral models, spatial models, etc.

This knowledge can enrich our horizontal understanding of AI and let us know that the development of AI spans so many fields. Some of these cross-fields are still in the research stage, while some have shown promising results. In the next few years, they will likely give birth to their own "ChatGPT". And when these new “ChatGPTs” emerge, how much cloud, computing power, and GPUs will they require?

These can greatly enrich our understanding and imagination of investment in the AI ​​field.

In addition, I suggest that you read more summaries and sharings of the development of the AI ​​field by some well-known venture capitalists.

For example, I recently read some insights from Sequoia Capital on the development of the AI ​​field, which mentioned the "intelligent agent economy" that may emerge in the future, that is, the economy that may be formed by the interaction between AI agents.

When talking about this economy, Sequoia Capital emphasized that it must have three elements:

The first is permanent identity; the second is seamless communication; and the third is security.

After reading these three elements, I immediately thought of blockchain. Aren’t these three elements the killer features of blockchain technology?

The encrypted wallet is the permanent identity of AI, the interaction based on blockchain smart contracts is seamless communication without interference, and the decentralized and censorship-resistant characteristics ensure the security of AI agents.

The above are some of the methods I use to learn and understand AI, for your reference.