The Future of Software Innovated with AI | VeriServe Academic Initiative 2021!

The 6th VeriServe Academic Software Initiative 2021, which started in 2016, was held on November 25th (Thursday) and November 26th (Friday).

In the present age where DX is accelerating, the speed of evolution of software services and products is accelerating due to the influence of the new coronavirus (COVID-19). We met and gave lectures.

The world in 2030 created by software | VeriServe Academic Initiative 2021

At the VeriServe Academic Initiative 2021, a total of 10 lectures were held under the theme of “The World Created by Software in 2030.” This is an event that considers the future of software from a perspective.

Day 1 Lecture content

Mr. Izumi, Ministry of Economy, Trade and Industry,
“Essence of Digitalization and Transformation to Growth Business – Current Status and Future of DX Promotion in Japanese Companies”
Mr. Shuho, VeriServe,
“Issues of Software Testing in Large-Scale Agile Development and Approaches to Solving Them”
Mr. Oshimi, CRI Middleware Co., Ltd.
“Innovation through game business and real-time communication”
Mr. Okada, VeriServe
“Future outlook for security systems – security concepts changing with AI “
Mr. Ishikawa, National Institute of Informatics “
Understanding the difficulty of AI quality and challenge with cutting-edge technology

Day2 Lecture content

Mr. Matsumoto, Fujitsu
“Digital Transformation Learned from Advanced Digital Countries -The World of After- Digital-“
Mr. Ito, VeriServe “Beyond
the Success of Test Automation -Issues after Automation and Future Prospects of Test Automation-“
Mr. Honda, Ideson
” Quality-oriented agile development -comparison with waterfall model development-“
Mr. Someda, HACARUS Mr. Suhara, VeriServe
“Vital points for applying QA4AI guidelines to actual projects” Mr.
Nawa, Cyber ​​Defense Japan
and practices to implement”

This time, we will report on the following three lectures!
Mr. Izumi, Ministry of Economy, Trade and Industry,
“Essence of Digitalization and Transformation to Growth Business – Current Status and Future of DX Promotion by Japanese Companies”
Mr. Okada, VeriServe,
“Future Prospects for Security Systems – Concepts of Security that Will Change with AI –”
National Institute of Informatics Research Institute Mr. Ishikawa
“Understanding the difficulty of AI quality and challenging with advanced technology”

The essence of DX is in data | Ministry of Economy, Trade and Industry Mr. Izumi talks about the current status and future of DX promotion

Mr. Noriaki Izumi, Director of Architecture Strategy Planning Office, Commerce and Information Policy Bureau, Ministry of Economy, Trade and Industry, explained the essence of digitalization and growth business from the following four perspectives.

• Intent and Overall Picture of DX Promotion Policy : DX Report 2.1
• Digitization Trends and Challenges to Transformation: A Case in the Manufacturing Sector
• The essence of digitization and the importance of data
• Creation of Digital Industry Considered by Architecture

1. Intent and Overall Picture of DX Promotion Policy : DX Report 2.1

   Based on the contents of “DX Report 2.1”, which he was also involved in producing, Mr. Izumi said that many people are saying that there are no precedents for modern top management in promoting DX. I explained it from the perspective of whether it is a corporate culture and whether precedents are really necessary when changes occur .

   In it, Mr. Izumi said, “Generally, management strategies are divided into two major strategies: “Better” to improve the current situation, and “Different” to bring about changes like online sales a while ago. , I want managers to think about which one to aim for with DX. ” he said.

   Also, “DX should be centered on ‘transformation’, but the ‘digital’ part has taken the lead. As a result, it is important to pursue customer value.” The intention of the DX promotion policy is to convey the idea.

2. Digitization Trends and Challenges to Transformation: A Case in the Manufacturing Sector

   Mr. Izumi introduced the manufacturing sector as an example of how digitization is changing the structure of society.
   In the manufacturing industry, the ability to rapidly develop and manufacture new products in response to market needs has shifted to the source of competitiveness .

   “Japan has Shinkansen technology, so they try to sell it to foreign countries, but the buyers won’t buy the Shinkansen alone.

   So, we thought about packaging the tracks, station buildings, IC cards, limited express tickets, and other sales techniques. We will lose to Europe, which offers everything, because we are asked for a package that can aim for a smart city.

   In Europe, led by Germany’s Industry 4.0, not only the cooperation of industrial systems but also alliances between companies are formed, strengthening the ability to make international infrastructure proposals, so we have developed a good system like Japan. We are in an era where we can no longer sell on our own. ”

   “This situation is seen in various fields, but it is most noticeable in the manufacturing field, and there is a possibility that the systems and machines of Japanese companies will lose their international competitiveness in the future. ”

   In addition, regarding DX in the manufacturing industry, there are many patterns that misunderstand the essence, saying, “When automating human work, the work process is mechanized as it is, but in the end there are cases where the role of people does not change. Instead, we need to redesign machine-centric processes and redefine human roles. ”

3. The essence of digitization and the importance of data

   Regarding the essence of DX, Mr. Izumi said, “Until now, cases that were mainly said to be IT were the addition of IT and cyberspace to physical spaces such as offices, but from now on, they will be optimized in the digital world. The point is whether we can think of a form that minimizes the movement in the physical space by allowing the system to act a little in the physical space.”

   In addition, regarding the importance of data, he said, “Some companies that use data select from the data they have, but we analyze what data is really necessary and select data. It is important to

   In addition, many managers do not think about data degradation, and correctly judging when and where the data can be used will lead to correct data utilization.

   By using data correctly, we will be able to create cases that have a great impact, just as analysis using data is now commonplace in Japanese professional baseball. ”
   he said.

4. Creation of Digital Industry Considered by Architecture

   Regarding the current state of DX, “About 40% of companies consider themselves to be top runners, but when we actually look at the companies, only about 8% are capable of DX.

   The background to this is the current situation where we are focusing only on improving existing products and it is not going well, so first of all, what kind of alliances should we form and what kind of incentive structure should we have? I think that it is important to approach with a “middle out” that spreads up and down.

   Then, top management creates an ecosystem by involving others, and creates a new customer experience along with the incentive structure. We call the middle of the middle-out strategy as architecture. ” he said.

5. Regarding the creation of the digital industry, he said, “Even if the same product is provided, it is possible to optimize it through mass optimization by shifting from sales to service provision. For example, a software engineer who provided , By providing functions that have already been completed, the burden placed on the user side is reduced and productivity is improved.

   In such cases, service providers can attract new value from intellectual property research, and users can focus on competitive areas, creating a good cycle. ”

   Mr. Izumi argued that the state in which vendor companies and user companies share a multi-side platform and become members of the ecosystem will be a successful pattern of DX in the creation of the digital industry.

Gone are the days when developing a good service would sell

“The essence of digital change is that the era of selling good products and good services is over.

In doing so, it is important to conduct “backcasting” to embody a stretched future image, rather than “forecasting” that aims to extend the existing system.

Therefore, I think that utilizing a large amount of data in response to an architecture that reflects the vision will lead to the promotion of DX. ”

Security Future Outlook | How AI Will Transform Security

Mr. Shuji Okada of the VeriServe Solutions Division explained the outlook for security systems based on his own knowledge.

The topic was divided into three parts, and we talked about the future of security systems from the following perspectives.

1. Transformation of social structure brought about by advances in IT
2. A new security concept that responds quickly to changes in social ideas
3. What is a security system to counter new threats?

1. Transformation of social structure brought about by advances in IT

   Regarding the future social structure, Mr. Okada said, “With the diversification of IT, DX will advance, and in the future, the spread of AI-based automation in various fields such as RPA and autonomous driving will likely change the social structure itself. There is a nature.

   In addition, the realization of smart cities is approaching through the spread of services such as car sharing, which until now were individually owned, and the combination of services. ”
   he said.

2. A new security concept that responds quickly to changes in social ideas

   According to Mr. Okada, with regard to the security issues that arise as a result of the changes in social structure discussed in 1., Mr. Okada said, I think there is a possibility that it will be necessary to do something, and that a new sense of ethics that accompanies each space will be born.

3. ” Regarding the concept of security, he said,
   “Until now, the three elements C・I・A of Confidentiality, Integrity, and Availability have been emphasized. On the other hand, in the future, “Safety”, “Health”, “Information, Intellectual property right”, “Ethics”, “Legality” I believe that it will change to the six elements S, H, I, E, L, and D of “Defense (corporate and national defense)”
   .

4. What is a security system to counter new threats?

   Mr. Okada said that cyberattacks such as malware and DoS should always be taken into consideration when implementing AI, but in the future, the composition will be AI vs. AI. talked.

   “In a future where cyber-attacks have become more sophisticated and sophisticated, the key is to take the initiative. On the other hand, assuming that attacks cannot be defended against, we must minimize the damage and share information immediately. I think it will be important.”

   In addition, as a function of security systems that will be required in the future,
   “Even today, AI is being used in various areas such as individual identification by face recognition and fraud detection systems that perform machine learning on credit card usage patterns. is an AI that specializes in a specific area, and is a ” weak AI ” with a limited functional area and depth of learning level.

5. “ Strong AI ‘ ‘ that derives and executes is required.
   “In order to create ” strong AI
   “, “AI is not omnipotent, it has strengths and weaknesses, so we have to understand it correctly. It goes without saying that it is fast.However, it has disadvantages such as the ambiguity of responsibility and the inability to derive 100% correct values. needs to be improved,”
   he said.

Toward the realization of future security systems

Mr. Okada talked about the three issues facing the realization of future security systems: policy, human resources, and technology
. “
In terms of human resources, there are few people who can envision future visions and strategies for security.
” I need to create a situation where I can judge.
“

Mr. Ishikawa, National Institute of Informatics About “Understanding the difficulty of AI quality and challenging with advanced technology”

Fuyuki Ishikawa, Associate Professor, Architecture Science Research Division, National Institute of Informatics, explained the concept of “AI quality” centered on two domestic guidelines on the theme of the difficulty of ensuring AI quality.

At the beginning of the lecture, Mr. Ishikawa touched on the performance limits of AI from the examples of Google Photos and Amazon .

“There was an incident in the past where a black person was tagged as a gorilla in Google Photos. ” was resolved by not using the word. For this reason, even companies like Google (tech giants with strengths in technology) cannot avoid AI performance limits. ”

“In addition, AI continues to learn, and it is not possible to recognize what it should not do. Amazon’s adoption of AI made decisions that were disadvantageous to women, so there were times when I stopped using it.

As a result of learning, it is possible that they learned about discrimination that was allowed in the past.Amazon is a company with a high percentage of men because there are many engineers.Since there are few women, AI cannot learn trends and discriminate. It is possible that it has become something like discrimination even though it was not meant to be.”

Difficulty of machine learning

Mr. Ishikawa said, “There are many engineers who are facing the difficulty of AI in the third AI boom,” and in 2018 he asked major companies, “When introducing machine learning in software development, ( I introduced a questionnaire that asked how difficult each item will be. 

According to the responses (see the slide above), 20-40% of the engineers answered that “a fundamentally different new way of thinking is required.” Based on this, Mr. Ishikawa

“Many people seem to accept that we need to change our way of thinking when it comes to assessing what customers want, assessing quality, and correcting and updating them.

From the point of view of quality assurance, it is not possible to prepare the correct answer even if you want to test it, or the cost of preparing it is high. Machine learning inevitably comes with difficulty because it is not always possible to capture the

and the difficulty of machine learning, and moved on to the concept of AI quality.

How to understand AI quality as seen from guidelines | What is AI quality that satisfies needs?

“Considering the engineer’s mission to meet the needs of customers, it is not enough to just produce numbers. For example, even if the accuracy rate of defective product inspection is 96%, it is possible that there are only a few defective products. Therefore, it is necessary to discuss the criteria for metrics*.

Also, when considering the quality of AI, from the viewpoint of meeting needs, it is necessary to ask questions such as I want you to guess this kind of defect pattern'' orIs there data for XX?” You are expected to use words (not just numbers). ”

* Metrics: Data collected and converted into easy-to-understand data (numerical values) by adding calculations and analyses, rather than in its original form.

(Quote:  MetricsManagement for making a right decision )

“There are two quality guidelines in Japan that include these stories. One is the QA4AI Guideline (AI Product Quality Assurance Guideline) , which provides a checklist for evaluating data, predictive models, and the entire system.

The object of evaluation is not only the deliverables, but also the consistency as a process . This is because it is important to continuously make improvements.”

“Another guideline is the AIQM (machine learning quality management guideline) . The subject that should be recognized is vague (see slide below). Describe it in a concrete form using words and mathematics.

In addition, we consider the data design, such as what kind of data is good in the first place, and examine whether the collected data is sufficient.The balance of the entire data is examined. This guideline also includes measuring the rate of oversight and the reliability of the program itself.”

At the end of his lecture, Mr. Ishikawa introduced an approach from the technical side of using AI to test quality.

“The technology called Search-Based Testing is a type of machine learning that designs tests and uses trial and error to improve the correct answer score. It is also called evolutionary calculation.

Meta (former name: Facebook) uses this technology for crash detection.It is used to find crashes with fewer tests and send them to engineers.In addition, a technology called DeepXplore automatically scans various images. We can find it and test the AI.”

Understanding and discussing the imperfections of machine learning

“When it comes to machine learning, I think the key points are ‘incompleteness’ and ‘uncertainty.’ There are guidelines like the one I introduced, but I’m still confused.”

“(AI) has a great impact on society depending on the scene in which it is used, and there are many things that you cannot understand unless you try or create it for the first time. It is important to do what we can and go through trial and error as quickly as possible. ”

In conclusion

This time, we reported on “VeriServe Academic Initiative 2021”!
Many issues have arisen with the promotion of DX, and it was clear that the current situation is that the level of requirements for AI quality and security is increasing.

In addition, from those who make DX policies, to those who actually develop it, and those who test the quality of the service, we will talk about the issues and current situation from the upstream to downstream perspective of DX, and learn more about DX and AI. It was an opportunity.