AI and Singularity

The term "singularity" is often used to refer to a time in the future when AI surpasses human intelligence. There is a wide range of predictions among experts on when singularity will take place. How will rapid technological advancements affect our lives and businesses, and how should we cope with such advancements?
[Fujitsu Insight 2017 Keynote Speech Report on "The Cutting Edge of AI and IoT Application"]

We are Currently in the Fourth Industrial Revolutsion, Triggered by AI

Iwao Nakayama
Corporate Executive Officer
Chief Evangelist, Global Marketing Department
Fujitsu Limited

If we take a look back in history, we see how various inventions have caused industrial revolutions. The cause of the first industrial revolution was the steam engine, the second was electricity, and the third was computers. Today, we are seeing the beginning of a fourth industrial revolution, triggered by AI. What does the future have in store in the next few decades?

We can expect a massive paradigm shift on a scale we have never seen before as a human species. In this paradigm shift, technology will rapidly transform, and the human experience will change so drastically that it will never be the same. This is what we call "singularity."

The Future of Singularity as Predicted by Ray Kurzweil

The person responsible for popularizing the use of the word "singularity" is Ray Kurzweil, who is currently in charge of AI development at Google. In the book that was published 12 years ago titled The Singularity Is Near, Kurzweil describes what kind of changes will take place in the time period between 2010 and 2040 due to the latest technological advancements.

In his book, he predicts that by the first half of the 2020s, AI will become capable of having intelligence that is equal to the level of a person who has received higher education. He also predicts that there will be small robots called nanobot machines that are the size of viruses, that will be used in medical care applications. He also predicts that by the late 2020s, virtual reality (VR) will be so advanced that it will be barely distinguishable from reality.

By the 2030s, he predicts that all people and businesses will have to be redefined to keep up with the changes. One of the factors that will bring about these changes is the capability for "mind uploading." In other words, humans will be able to use network connections to directly transfer thoughts from the brain. This means that humans will become software databases much like smartphones.

By directly inserting nanobot machines into the brain, we will be able to generate VR without the use of external devices, fusing the real world we see through our eyes with the virtual reality inside our brains. Such an age will become reality by 2030. Furthermore, we will be able to directly send information to our brains in real time, making it possible to experience the sensations of another person remotely. For example, when trying to perfect your golf swing, it will become possible to transmit the data that Hideki Matsuyama's (a professional golfer) brain sends to his muscles when he takes a swing.

By the 2040s, we will be spending most of our lives in virtual reality, much like in the movie The Matrix. Instead of waking up in the morning to go to school or to work, we will stay in our rooms with nanomachines inserted into our brains, and interact with people in virtual reality and create new businesses. We will enter an age that we cannot even fathom today.

We can predict that singularity will become reality due to the combination of various factors, but the most important role will be played by AI.

The essential factors for the creation of singularity

Ten Tips for Implementing AI Successfully According to McKinsey & Company

How should we cope with AI as we go forward? Let's take a close look at the 10 tips for implementing AI successfully, according to research by McKinsey & Company.

1. Do not be excessively enthusiastic and take in everything at once. For now, not all companies are utilizing AI.

This day in age is often described as experiencing an AI boom, but the percentage of companies around the world that have implemented AI is only 20 percent, and companies that are in the experimental stages is 41 percent. However, it is expected that these companies will fully implement it in the near future. Especially in the communications and financial fields, costs spent on AI are expected to increase by 15 percent.

2. There is credibility in the claims that AI will increase sales and profits.

Results show that out of three companies that have implemented AI, one has achieved increase in profits. Furthermore, it is expected that the profit rate will increase over 5 percent compared to competitors. AI is directly influencing the increase in profits.

3. Without the support of top management, transformation caused by AI will not succeed.

The key for succeeding in AI implementation depends on whether top management understands what AI is, and what kind of profits it yields. Companies that have succeeded did so in part because they received more support from top management than those that did not succeed. It can be said that companies that have failed to implement AI did so due to insufficient support.

4. There is no need to push AI implementation forward without external support. Gain the capability and capacity through cooperation with other companies.

There is no need to push AI implementation forward with just the staff in your own company. Even companies like Amazon and Google have developed partnerships with other companies to boost their AI skills. Many of the companies that have implemented AI early have purchased the necessary solutions from other companies depending on what their objectives were. Companies that have implemented AI by developing all AI solutions themselves are actually in the minority.

5. Do not succumb to the temptation of leaving AI implementation up to just the IT team.

If the project is left entirely up to the internal IT team, an AI system that does not meet the original objective criteria may be implemented. The AI system will be built by those who do not understand the setting in which it will be used, and in the end, a totally ineffective system will be built. Therefore, the project must be driven forward through cooperation between IT and the teams that will use the system.

6. Adopt the portfolio approach to accelerate AI implementation.

In order to implement AI, a portfolio approach is necessary that is divided into the following three patterns: for the purpose of utilizing existing, verified technological solutions (short-term), to verify usability for main business purposes using incomplete technologies (mid-term), and for high-impact purposes utilizing cutting-edge technology (long-term).

7. Machine learning is a powerful tool, but it does not suit every purpose.

When developing solutions through AI, machine learning is often used. However, this is not a tool that suits every purpose. It is necessary to search for the most suitable tools for each objective.

8. Before implementing AI, it is crucial to build digital capabilities.

Results show that companies that have abundant experience in digitalization have higher chances of increasing profits using AI. Many companies that have implemented AI early on had already succeeded in digital transformation by investing in digital capabilities including cloud data and big data.

9. Be bold.

Companies that take on major challenges and assertively pursue strategies report far more promising profit prospects than companies that lack strategy.

10. The greatest obstacle is people and operational processes.

Going forward, there is an increasing need to clearly distinguish between the roles of AI and people. Decision-making management to promote optimum decision-making is more important than process efficiency. There must be a re-education program for employees in preparation for the coming age of AI.

What is the Content of the 600 AI Projects in Japan?

Next, we introduce the more than 600 AI-related projects that are currently in progress, and their specific AI applications and technologies.

The concept behind Fujitsu's AI, called "Human Centric AI Zinrai," based on the Japanese term shippujinrai, meaning "with lightning speed," is human-centric. In other words, it's AI that puts humans in the center. The number of projects running in Japan has already surpassed 600. In Zinrai, in order to handle each of those projects, Fujitsu has prepared APIs suited for different objectives. They are designed to be able to handle various project needs, such as traffic image recognition, conversation translation, demand forecast, delivery planning, and human resource procurement.

When classifying the 600 projects into industries of application, the percentage of manufacturing is 40 percent, distribution is 27 percent, and public/regional projects is 17 percent. Utilization in the fields of application is as follows: 19 percent in call centers, 13 percent in creative manufacturing, 12 percent in knowledge application, 11 percent in demand forecasting and marketing, and 9 percent in facility maintenance. When observing these applications by usage purposes, mobility, knowledge, social infrastructure, manufacturing, facility maintenance, and call centers make up the bulk.

Fields in which negotiations are underway for Zinrai

What Solutions Can AI Implementation Provide for Companies?

By classifying these many examples, we examined the various needs that customers have, and prepared 17 types of solutions to various challenges. We will introduce some examples in detail.

A Business System Capable of Natural Conversations with Customers

This technology automatically learns interaction methods from legacy conversations, to cope with the diversity and ambiguity that are characteristic of the Japanese language. Tokio Marine Nichido implemented this as a customer-interaction business system capable of natural conversation, and evaluated it highly for its effectiveness.

In one example of the use of this technology, when a customer comments "I wanted to go to Hawaii," during an interaction, the system answered, "Hawaii sounds nice. Where are you going today?" by recognizing that it was a wish, and not the actual destination. Natural conversations such as this was made possible.

Interactive question answering with AI

Allocation of Nursery School Admissions

Next is an example of a municipality in Saitama Prefecture that performs admission screenings for nursery schools. There is currently a nation-wide problem of parents not being able to put their children in the nursery schools of their choice.

In the past, 20 to 30 staff members spent several days assigning children of each household to nursery schools based on information such as the annual income, family structure, and various other factors. The task of assignment was extremely difficult, with factors to consider such as making efforts not to place a child in a different nursery school than their sibling in the same household. However, after implementing Zinrai, the task that took many staff members several days can now be completed in a few seconds.

Efficient allocation of nursery school admissions

Predictive Detection Utilizing Fiber Optics

Next is an example of a project in the field of social infrastructure: predictive detection of equipment abnormalities. Temperature gradation indicators visualize areas where heat is being generated, using optical fibers with special coating as IoT sensors on machines and equipment. For example, this system is capable of using AI to automatically detect whether there are any anomalies such as leakage or clogging in piping or corrosion in facilities such as power plants and factories.

Predictive detection of equipment abnormalities

Cutting-edge AI Technology, "Deep Tensor"

Fujitsu has a technology called Deep Tensor. In the course of 60 years in the history of AI, after machine learning was invented, one of its algorithms led to the development of deep leaning, leading to technologies that can beat the world champion of chess in a game. Deep Tensor is another algorithm of machine learning.

Up until now, big data used in AI mainly dealt with numbers, text, images, and sound. However, what makes Deep Tensor unique is how it can use data of the various interactions between humans and objects in the real world that are made into graphs. Fields of applications include cyber-attack countermeasures, fintech, and in silico drug discoveries.

Defense Against Malware

Today, most companies are having to deal with the problems of hacking by external entities. By using Deep Tensor, malware behavior can be detected and presented in graph format. Abnormalities are discovered from the graphed communication logs, and the system determines whether they are common behavioral patterns of malware, or should be considered threats in any way.

Distinguishing malware attacks

Estimation of Damages in the Internal Structure of Bridges Using AI Technology

When performing periodic inspections of tunnels and bridges, it is often difficult to determine the extent of damages in the internal structure. At first glance, a concrete pillar may look stable, but the inside may be heavily eroded. Fujitsu has developed AI technology that is capable of determining the existence of damages in the insides of bridges through data acquired from sensors placed on the bridge. Vibration data taken from 180 sensors attached to the bridge is modeled through deep learning, enabling it to calculate the degree of irregularity and change. This makes it possible to provide results for the workers to determine the safety of the bridge.

Estimation of damages in the internal structure of a bridge

What are the Strengths of Fujitsu's AI "Zinrai"?

Fujitsu believes that the purpose of Human Centric AI Zinrai is to answer the various needs of customers. Where do Fujitsu's strengths lie? We believe that it is the fact that Fujitsu has Japan's greatest integration power.

Fujitsu's engineers are very dedicated, and only release products after high levels of completion are reached. This is how we hope to meet the needs of our customers as best we can. Today, various companies around the world are offering AI solutions. However, AI systems themselves do not vary among the many providers. With AI technology at the base, the aim is to customize that technology to meet individual needs. This is where the power of integration becomes necessary.

Although Fujitsu currently has 600 projects, this number will quickly rise above 1,000, 2,000, and more. By receiving a variety of requests from customers who want to create and perform new things with AI, we provide a wide range of solutions, which lead to new problem-solving ideas. We hope to increase the number of solutions we provide to the current 17 varieties to 30, 50, and more. Fujitsu is dedicated to creating AI solutions that meet the needs of not only large companies, but also small to medium-sized businesses, and the lives of all the people in the world.

  • Iwao Nakayama
    Corporate Executive Officer
    Chief Evangelist, Global Marketing Department
    Fujitsu Limited