Use of AI, quantum computing, and other next-generation computing technologies has been increasing rapidly. Keizo Azuma of Fujitsu's AI Services Business Unit, which is on the forefront of the AI business, talks about changes in the business landscape and the latest trends in technologies for solving social problems.
Next-generation Computing Shifts from PoC to Implementation
Most prior efforts to use AI only completed Proof of Concept (PoC) verification. Since 2018, AI became more used in business. The widening scope of use is notable. Initially, AI mostly handled part of operations, but recently there is a marked trend to use AI for the entire system (e.g., human resources and document management).
AI's importance to business grows as it spreads to new sections, divisions, and tasks. Quantum computing has also drawn more attention globally since 2017, though countries focus on different areas.
The United States (e.g., IBM, Google, and Microsoft) focuses on quantum gate systems like IBM Q. Canada, where government-backed companies (e.g., D-Wave and 1QBit) work on quantum annealing technology, has a high level of both hardware and software. In China, a huge national budget funds research on quantum-based encryption.
In Japan, besides Fujitsu, Hitachi, NTT Group, and NEC have been developing annealing machines that are not based on quantum gate systems.
Many companies compete to develop quantum computing technologies, but I believe Fujitsu's Digital Annealer is probably the only practically useful technology in the world.
Digital Annealer Efficiently Solves Large-scale Problems Impractical for Conventional Computers
Digital Annealer is a new architecture that uses a digital circuit design inspired by quantum annealing behaviors to solve combinatorial optimization problems quickly.
Conventional computers calculate each possible pattern to find the optimal build-up approach. Quantum computing derives the optimal pattern without such calculation, so quantum computers instantly reach answers that take conventional computers much time.
Among quantum computing technologies, Digital Annealer's strength is solving large-scale problems with an enormous number of combinatorial patterns.
In quantum computers, only adjacent quantum bits can be coupled together; distant bits cannot. Also, the contacts with adjacent quantum bits that can be coupled are limited; the number of contacts that can be coupled is the upper limit on processing.
By contrast, Digital Annealer uses conventional digital technology that does not have this limitation. A large number of inter-bit couplings form a totally coupled state to instantly exchange information, enabling Digital Annealer to quickly calculate large, complex problems.
Solving Combinatorial Optimization Problems in Daily Life to Make Society Better
Digital Annealer makes quantum computing easier to use in real business. The first-generation Digital Annealer processes data with 1,024-bit scale elements. In stock market investments, Digital Annealer can analyze 1,024 stocks, about half the stocks in the first section of the Tokyo Stock Exchange.
In December 2018, we started offering 8K bits (8,192 bits) as the second generation as a cloud service, and in Q2 of 2019, we will launch onpremises service. With the second generation, you can analyze all stocks in TSE's first section at once and derive the optimal portfolio.
Our daily lives have many large-scale combinatorial optimization problems; solving these will make society better.
For example, annealing advancements may greatly help prevent traffic jams. Today's car navigation systems show the best routes based on traffic conditions. Since such systems have similar performance, the same detour routes are displayed regardless of the vehicle you drive, so many vehicles crowd certain detour routes.
In the future, annealing technology can combine an enormous number of routes to show optimized routes per vehicle. If Digital Annealer can derive the optimal route per vehicle, there will be no huge traffic jams on holidays.
Also, Digital Annealer is helpful when a train line stops due to an accident. Once an accident occurs, many people wait on the platform and detour lines become crowded. If routes for detour lines based on overall optimization are displayed on each person's smartphone, replacement transport will not be so crowded.
Fostering Business with Universities and Companies
Digital Annealer can be applied to business, including finance, chemical and drug discovery, retail and logistics, and information systems.
Joint research with universities and PoC verification at companies have already begun.
Fujitsu Laboratories Ltd. and Waseda University made a Fujitsu Co-Creation Research Laboratory to solve actual combinatorial optimization problems with Digital Annealer.
We do joint research in many fields like finance, digital marketing, and logistics. We also established a Fujitsu Co-Creation Research Laboratory with the University of Toronto in Canada, an international leader in quantum computing and AI, in March 2018. We do joint research on traffic, networks, finance, and healthcare.
Corporate use of Digital Annealer is growing too. In manufacturing, small-lot production of multiple models is increasing, but it is not easy to manage each phase. Fujifilm Corporation uses Digital Annealer for technical verification of production line optimization.
In finance, investments are diversified with portfolios to reduce risk and increase returns. Mitsubishi UFJ Trust Investment Technology Institute Co., Ltd. (MTEC) uses Digital Annealer to verify optimized portfolios in asset management.
In marketing, the digital marketing approach of reflecting each individual's differences in taste, age, and sex in web ads is actively employed. For more efficient digital marketing, Recruit Communications Co., Ltd. does R&D on new marketing technology.
These businesses use Digital Annealer to instantly find the optimal pattern from among an extremely large number of combinations.
Aiming to Combine AI, Annealing Technology, and Supercomputers
ICT has been changing the shape of business for many years; it creates systems by transforming tacit knowledge into explicit knowledge.
However, some tacit knowledge has not yet been turned into systems. Examples include call center logs and histories of negotiations with clients. In many cases, these are left abandoned in text format.
Also, today we can use IoT sensors to collect various data for digitizing and processing. For example, we can grow good vegetables in the optimal environment by controlling environmental factors at a farm (e.g., temperature and humidity). IoT enables us to control things that humans did not perceive or monitor before for overall optimization.
In such processes, ICT must evolve into the next stage. AI can analyze abandoned tacit knowledge without formalizing it. Using ICT to process all data to optimize the entire society makes life better.
To this end, annealing technology, AI, and supercomputers are useful. Fujitsu has the world's most advanced technologies, like Digital Annealer, the K computer and AI Zinrai. We have an environment to fully leverage the synergy of annealing technology, supercomputers, and AI.
Fujitsu is one of the few companies in the world that independently works on these three technologies. Fujitsu contributes to society's development with new technologies.
- Keizo Azuma
- Head of the AI Services Business Unit, Fujitsu Limited