AI experts share: Beginners to understand deep learning puzzles

Introduction: Gao Yang is currently a data scientist and artificial intelligence expert in a startup company. Former senior big data expert in Huanju Times, with rich experience in machine learning and deep learning.

In Mr. Gao Yang's previous sharing, many students asked Mr. Gao Yang various questions about the frontier technology field of introductory deep learning. For some important questions, Mr. Gao Yang made specific responses. The Truex X technology education platform is committed to benefiting more technicians. With the authorization of Mr. Gao Yang, we will share the professional Q&A of Mr. Gao Yang here.

Today's sharing will focus on the confusion of beginners in deep learning: whether traditional technical programmers should transform into artificial intelligence, and how to do it more suitable to develop under the AI ​​wave.

Recently, a classmate sent a dialogue message on the Singularity Big Data public account. The content of the message is as follows:

[Can you analyze how the traditional technical programmers who are already working are more appropriate under the AI ​​wave? Whether or not to transform under the AI ​​wave AI; or continue to specialize in the original direction, and learn to understand AI by yourself.

Because:

I. The AI ​​learning curve is steep and self-learning is costly and difficult.

II. If you do not have actual project experience, you cannot provide endorsement when applying for an AI position, and you are just a beginner.

III. If the company has an AI project, they often apply for AI-related personnel directly. The company’s internal post adjustment is not suitable

for IV. The previous work experience of turning to AI will be void

. Not valued in the company, very easy to be replaced! 】

I think, such a confusion is a lot of beginner students will have. Then I will tell you my understanding of these issues.



Problem I. The AI ​​learning curve is steep, and the cost of self-study is high and difficult.

This difficulty exists objectively, and the learning curve of AI is indeed relatively steep. Compared with some engineering application products, such as MySQL, Hadoop, etc., these products can basically be installed and some basic operations can be used first, and the concept is clear and simple. Of course, if you become an expert in these fields, you will have to go through a series of systematic studies and the accumulation of engineering experience. It's just that these fields are relatively mature, and it's OK to learn according to the standard routines, and there are many Chinese learning materials, so the difficulty of learning is much less.

The difficulty in artificial intelligence comes from several aspects. On the one hand, the most cutting-edge technologies and theories in the world come out first in English, such as important papers released at conferences. These papers not only test everyone's English reading ability, but also contain a lot of proper nouns and technical terms. For many beginners, even if these nouns are correctly translated into Chinese, they still cannot understand what they mean. So, starting the third floor when the first two floors of the building are not well established will be extremely painful and fruitless.
However, this problem can still be overcome by other workarounds, for example, there are now many Chinese books. The emphasis of these books is different. For beginners who have difficulty understanding English papers, they might as well look for some easy-to-understand materials. After all, there are many Chinese materials that are not cryptic and scripted, but use very elegant examples. Interpreting the way AI solves problems from the programmer's point of view is still a great reminder for learning this subject.

Therefore, since it cannot be done in one step, then climb up from the bottom of the mountain.

Question II. If you don't have actual project experience, you can't provide endorsement when applying for an AI position, you are just a beginner.

Endorsement is not based on the endorsement of the project itself. So far, there are very few artificial intelligence projects that can truly be regarded as complete. Many projects are also old bottles of new wine. I have seen more than one project that is an ordinary data mining project or a BI project, and it starts to "cheat money" with an artificial intelligence shell. Coming out of such a project, I don't think even the project can endorse itself.

Endorsement is up to you - there are many, many AI engineering projects on GITHUB, which you can use to study. Can you read them? Can you explain clearly how they are implemented? Can you find the corresponding location faster and make changes when necessary? Can you use your own abilities to make beneficial improvements? If the answer is yes, these competencies will be on display during your interview. Don't say "I can, but I can't say it", this problem will never exist. As long as you are willing to work hard, have a clear idea, and the project is always the same, the interviewer will know your level with just one mouth.
Therefore, the kung fu is done by yourself, don't expect someone to endorse you when you go from zero to one.

Question III. If the work company has an AI project, it is often directly applied for AI-related personnel, and it is not appropriate to adjust the internal position of the company.

This problem may not be certain. Position adjustment within the company is relatively free in many companies, but the premise is that you have to let your leaders confirm that your new position will be more valuable than your current position. Can you understand the logic of this? Can you get your leaders to figure out this logic? If it can, the rest is not a problem.

Therefore, you must work hard at ordinary times, accumulate a lot, and always be prepared to understand your own strength and ability, so as to seize the opportunity when there is an opportunity.
Question IV. Turning to AI to void previous work experience

This is just a question of giving up, giving up, and giving up. What you need to do is to measure which part is more cost-effective for you according to your own situation. AI is just a new field. No one can guarantee that you will get better results if you do this. Of course, if you don’t transform, no one can guarantee that you will be able to retire, right? Everyone’s situation is different, so it’s impossible to make everyone swarm to study AI and promise that AI will earn millions in the future after all… It sounds like a cult is bewitching people.

In addition, the previous work experience I do not think is completely invalid. You will always sum up something in your work, and you will always have some experience. The so-called accumulation is not only written on the resume, I wrote Java code for 6 years, but more about the understanding of the objective world, or the ability to fulfill my own philosophical views on new things.
Therefore, there are differences between people. We must admit that, and we must make reasonable choices according to local conditions, and take the opportunity to summarize and summarize experience in a higher dimension, so as to make ourselves more invincible.

Question V. Worrying that the wave of AI is too strong, so that traditional programmers can only survive in the cracks. Not valued in the company, very easy to be replaced!

Don't worry too much about this, at least it won't be a problem for the time being. The so-called, the ruler is short and the inch is long. Programmers survive by doing things that machines can't. Let’s make an inappropriate analogy. Now that the degree of industrialization is so high, the production capacity is very strong, but many jobs that seem to be less advanced have no effect. Do you see the aunt who takes out the garbage being unemployed? Is the delivery guy unemployed? Are the chefs who cook in restaurants out of work? Will programmers lose their jobs if they're not unemployed? Don't you think what programmers do is less fungible than what they do?

AI has its own advantages, and it must have its own limitations. Many of these limitations are the irreplaceable advantages of human beings themselves.
So, do your job well, especially those things that machines can't do, and you'll be more valuable and irreplaceable.