How difficult is the qualification that is popular with AI engineers? I will introduce it in a list

1. What are the recommended qualification criteria for AI engineers?

When considering a qualification recommended for an AI engineer, the most important thing is to be able to confirm and learn the skills required for an AI engineer through studying that qualification. Therefore, let’s first confirm the AI-related technologies required for AI engineers.

1.1 Typical AI (artificial intelligence) related technologies

The techniques required for AI engineers include the following:

■ Deep learning

Deep learning is a field of artificial intelligence research, and is a machine learning method that uses multiple layers of neural networks. A neural network is a mechanism that processes information created by modeling the structure of the human brain. By using this neural network, it has become possible to make a machine recognize an object by following a process like when a human recognizes it, instead of the conventional recognition by mechanical processing. is.

■ Cloud AI

Cloud AI is simply AI used on the cloud. The amount of data that you want to process with machine learning tends to be large. It is a method of sending the large amount of data to the cloud and learning using the computing device of the cloud. By using the cloud, there is an advantage that you do not have to prepare a large amount of data and a powerful arithmetic unit yourself. Furthermore, each major cloud service is providing an environment that makes it easy to use AI one after another. For example, GCP (Google Cloud Platform), a cloud service provided by Google, provides a development environment for AI engineers and data scientists called AI Platform.

■ Edge AI

Edge The edge of AI is a “terminal”. It is a mechanism that processes AI by moving it directly on a terminal such as a smartphone, which was previously supposed to be used on the cloud. For example, Edge AI, which does not require much data transmission / reception to / from the cloud, is useful for processing that does not allow communication delays. In addition, not communicating much with the cloud has the advantage of reducing the amount of data communication and preventing information leakage.

I summarized the comparison between cloud AI and edge AI in a table.

1.2 Criteria for selecting qualifications recommended for AI engineers

After confirming the required technology for AI engineers, we will introduce the recommended qualification selection criteria that take that into consideration.

■ Practicality

What is important for qualification study is that the study content is practical? That is the point. For example, “Python3 engineer certification data analysis test” is recommended as a qualification in which the content of the test range is particularly consistent with the range required for AI engineers. This qualification is still a new qualification and has almost no history, but highly practical libraries such as “NumPy” and “pandas” are included in the question range. Therefore, you can gain knowledge of data analysis practice while preparing for the exam. For example, there are other libraries that will be asked in the “Python3 Engineer Certification Data Analysis Exam” as follows.

• NumPy
• pandas
• Matplotlib
• scikit-learn

By confirming the test range in this way, we are examining its practicality.

■ Number of examinees

All qualifications directly related to AI have a short history, and there are still many qualifications that are not well known. Even so, it turns out that the qualification that attracts a large number of examinees is highly recognized, and it is easy to prove objective skills when acquiring that qualification.
For example, the cumulative number of examinees for the two qualification exams “G-test (for generalists)” and “E qualification (for engineers)” conducted by the Japan Deep Learning Association (JDLA) is as follows.

• G-test: 60,887
• E qualification: 4,150

■ Versatility

Considering versatility, we also pick up contents for the entire IT engineer, such as those that ask basic knowledge of databases and IT in general, and qualifications related to statistics. This is because basic IT knowledge and mathematical knowledge are also important as prerequisites for learning AI.

2. 10 Recommended Qualifications for AI Engineers

Next, I will introduce each of the recommended qualifications in detail.

The G-test conducted by the Japan Deep Learning Association (JDLA), a general incorporated association, is a test to confirm whether you have the ability and knowledge to learn the basics of deep learning and utilize it appropriately.

■ Difficulty of G-test

The guideline for the difficulty level of the G-test has not been announced at this time. However, since the pass rate is as high as 65.9%, it is easy to think that it is a simple test compared to the IT passport test, which has a pass rate of about 50%. However, for the IT passport exam, which is mainly for students and IT inexperienced people, it is people who have a high interest in AI from the beginning who know the existence of the G-test that has just begun and are taking the exam. .. Therefore, it seems that the test is not simply easy.

Like the G-test, the E qualification conducted by the Japan Deep Learning Association (JDLA) has the ability to select an appropriate method and implement it in a program after correctly understanding the theory of deep learning, and has a deeper knowledge of AI. It is a qualification to prove that you are doing it.

■ Difficulty of E qualification

The pass rate for the E qualification is as high as 71.9%, which is the same as the G test. However, this result is only the result of the examination by those who have completed the certification program, and it is the qualification that those who have deep knowledge about AI will come to take it in the first place.
According to the E-qualification examinee questionnaire (valid respondents: 68) surveyed by “Study AI Co., Ltd.”, the most cited “applied information technology” was the “examination that seems to be close in difficulty”. In the “personal examination”, it was 39.71%. Furthermore, among the questions that include the Oracle Master Platinum, Project Manager Examination, and PMP Examination, 26.47% of the respondents answered that it was “more difficult than this list”, which is the second highest after applied information, indicating that it is a fairly difficult qualification. Probably.

2.3 Image processing engineer certification

The Image Processing Engineer Certification is a private qualification conducted by the Graphic Arts Society. In the field of image processing, which is one of the ways to utilize deep learning, it is a test to evaluate the acquisition status of the knowledge necessary for design and development. Although the number of examinees is small, it is a long-established qualification that has been available for 15 years.

■ Difficulty of image processing engineer certification

The image processing engineer certification is carried out by the mark sheet method, and it passes with a response rate of 70% or more. The pass rate is about 60 to 70% for basic and 30 to 40% for experts. Considering that the pass rate of the IT passport exam is about 50% and that of applied information engineers is about 20%, it seems a little easier than that. However, in the first place, this is a test for engineers and R & D related people who are interested in image processing. Therefore, the actual difficulty will be a little higher than the image you receive from the numbers.

2.4 Actuary Qualification Exam

The actuary qualification test is a qualification test conducted by the Institute of Actuaries of Japan. An actuary is also called an “actuarial science person” in Japanese, and refers to a person who analyzes financial products and calculates insurance premiums using mathematical mathematics such as advanced probability theory and statistics. There is nothing directly related to AI in the current test range, but much of the mathematical knowledge dealt with there can be applied to data science. Furthermore, most of the mathematical processing that actuaries carry out is what AI is good at, and it is also a field where the introduction of AI is expected in the future.

■ Difficulty of actuary qualification exam

Actuaries are occupations that cannot be identified unless they pass the qualification test, but it is said that the number of registered people is currently around 1500. Supply is not keeping up with demand because it requires an esoteric understanding of mathematics. It’s a fairly difficult exam, but once you get through it, you won’t have any trouble incorporating mathematics into your program anymore.

2.5 ORACLE MASTER

Oracle MASTER is a certification system (vendor qualification) implemented by Oracle Corporation Japan, the provider (vendor) of the Oracle database. It is a qualification that can prove knowledge and skills regarding the operation and management of Oracle databases.

■ Difficulty of Oracle Master

According to the “IT Skill Standards (ITSS)” established by the Ministry of Economy, Trade and Industry, the difficulty level to which each grade of Oracle Master belongs is as follows. It is certified as difficult as the IT passport exam in bronze and as difficult as the advanced division exam in Platinum.

Source: IT Skill Standard V3 Download | Information-technology Promotion Agency

2.6 Statistical tesThe statistical test is a private test conducted by the Statistical Quality Assurance Promotion Association, but it is a test supported by the Japan Statistical Society and various ministries and agencies. It covers the knowledge necessary for using statistical data from basic to advanced, and it is a test that can confirm very important knowledge when handling statistical data processing as an AI engineer.