How to find a quick entry point for learning in the field of Artificial Intelligence?

Recently, I have read many articles about artificial intelligence and have deep feelings, here is a summary of what I have read and felt these days through this blog post.
In the field of Artificial Intelligence, many people find it very difficult to find entry points to learn and improve in the face of ever-changing new models, methods, and ideas. So, today I’m going to share how to quickly get started in the field of artificial intelligence to help you find some shortcuts to learning.


Focus on the knowledge backbone

In each sub-field of AI, there is a very limited amount of knowledge. By “backbone” I mean the important assumptions, ideas, and methods that make up the subfield. If you can master these backbones, then it is relatively easy to understand the other loose ends.
There is another reason why it is important to understand the backbone. On the scale of time, there are techniques and ideas that used to be very influential at one time, but as time changes, they are replaced by newer techniques that come later. Therefore, it will also be easier to see the timeliness of different types of technologies by just grasping the backbone.
So how do we go about understanding what a knowledge backbone is?

First, the “literature review” paper. The purpose of this type of paper is to provide a staged summary of a relatively new and rapidly changing field. In a sense, we can think of the literature review as another kind of “textbook”. Since a certain field is changing rapidly, a literature review of knowledge in this field will be very current. Therefore, when reading a literature review, we need to pay attention to the period of publication to ensure the freshness of the literature.


Second, “Tutorials” of top academic conferences. As we mentioned before in the sharing of how to learn the content of the conference, the lectures often contain some of the latest hot issues, and at the same time, there is a set of stage-by-stage results for the study of these issues, and at this time, the researchers who have made outstanding contributions to this field will summarize the content of these lectures and share them with the audience. In other words, paying attention to these lectures is equivalent to understanding the literature review of a certain field.
It is important to note that the method of focusing on the backbone of knowledge can be “recycled”. What does this mean?
For example, let’s say we want to understand what the basics of deep learning are. At this level, the main components are feed-forward neural networks, convolutional neural networks, and recurrent neural networks. Then, next, we need to understand roughly what these three different neural networks actually are and what kind of problems they solve. Layer by layer, for a particular neural network, such as convolutional neural network, we need to understand the main knowledge of “what is the meaning of convolution”, “what is the purpose of this structure”.
To summarize, no matter what you learn, the first idea is to look at what it is the backbone of knowledge, to develop such a habit of thinking, your learning efficiency will be greatly enhanced.

Follow renowned scholars

Another “shortcut” to getting started in AI is to follow the research of scholars in your favorite field.
In some subfields of AI, it is often the case that the work of a few scholars sets the stage for major developments in the field. Have you noticed that this is actually the backbone of the field in another sense? Tracking and studying the research results of these scholars serves two purposes: on the one hand, you can learn the latest developments; on the other hand, you can target your learning and achieve twice the result with half the effort.
If we collect about five scholars for a field, and then pay attention to the results of these scholars and their institutions or laboratories, then generally speaking, you read almost 10 papers, you will be able to grasp the general content of the development of this direction.

Learning a small number of models in depth

For many beginners in AI, a dizzying array of models can be overwhelming. At this time, instead of diving headfirst into the endless variety of new models, it is better to learn a few models in depth. This idea is also in line with the software field we often say “T-shaped” talent model, breadth is very important, but without the depth of the breadth, many times there is no value. With depth first, it will be easier to expand the breadth.
For example, from the industrial application point of view, we generally need to master just a few models. It can be said that if you master the linear model, tree model and neural network model, you can solve more than 80% or even more of the business needs. What we need to do, then, is to learn these three models in depth and in detail, rather than to understand dozens or even hundreds of models.

Summary

I’ve sorted out a few ideas for you to get started quickly in the field of Artificial Intelligence, and hopefully you’ll be able to quickly locate useful information to improve your learning process. However, one thing I want to say is that although what we’ve shared today is called How to Get Started Quickly, it’s even more essential that we have a smooth heart and an attitude of willingness to get down to the basics of learning in order to really grow in this field. In other words, there aren’t many shortcuts to growing into a competent data scientist or AI engineer.

Print Friendly, PDF & Email

Posted

in

by

Tags:

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *