NTT R&D Forum 2024 will be held for five days from the 25th (Mon) to the 29th (Fri) at the NTT Musashino Research and Development Center. Prior to this, a keynote speech and preview for the media was held on the 21st (Thu). In the keynote speech, NTT Executive Officer and Head of Research and Planning Kinoshita Shingo took the stage and introduced the efforts of NTT laboratories under the title “IOWN INTEGRALS.” Here is a summary of part of the speech.

Keynote speech by Head of Research and Planning Kinoshita Shingo: “IOWN INTEGRALS”

There are two meanings behind “IOWN INTEGRALS.” INTEGRATION means “integral” or “essential,” but we want to apply IOWN to various fields by putting variables such as networks, security, and AI into the function IOWN and build it up. We want to make the value of IOWN indispensable to humanity and the earth.

NTT East and West launched APN IOWN1.0 in March last year, but we are often asked what the difference is between APN and dark fiber.

Since the core part of the network is already established, APN functions immediately once access is connected. It can be opened in a shorter period of time than dark fiber.

In addition, dark fiber cannot be easily changed once it is laid, but APN makes it easy to change bases. For example, today it is this base, and tomorrow it is that base, so there is a high degree of freedom in changing the connection destination.

In addition, it has various other features, such as low management costs and easy long-distance transmission.

Currently, with a view to optimizing RAN using APN, we are considering whether we can improve efficiency by applying APN to 5G equipment. Base stations and antenna equipment are installed on the assumption that there will be a maximum number of people in residential and commercial areas. However, during the day, many people move to commercial areas, and residential areas are often empty. It is inefficient to operate equipment when there are no people. That’s where APN comes in. By connecting the equipment in residential and commercial areas with APN, the functions of the equipment in the residential area are transferred to the commercial area during the day, and the functions of the equipment in the commercial area are transferred to the residential area at night. By doing this, it may be possible to halve power consumption.

DCI, the computing infrastructure of the IOWN era, can achieve high efficiency and low power consumption by dividing computing resources as finely as possible and operating only the computing resources that are necessary for data.
In addition, electrical wiring had limitations on highly efficient computing systems due to the limited connection distance, but by using optical wiring, the connection distance of computing resources can be extended, realizing a large-scale, highly efficient computing system. DCI is expected to be used for data processing that is large, diverse, and dynamic.
For example, in image analysis AI that processes a large amount of camera footage in real time, optimal resource allocation based on the difference in population between day and night can be used to achieve significant power efficiency. In the future, it is expected that APN will be used to apply this technology to multiple data centers geographically distributed.

By efficiently connecting data processing platforms distributed by APN, it is expected that a large-scale, high-performance, and power-efficient ICT infrastructure can be realized.

DCI-2 is being developed for commercialization around 2026 in IOWN2.0. DCI-2 aims to improve power efficiency by 8 times by connecting CDI servers, which have computer resources subdivided into board units, with optical switches using photonics-electronics convergence devices and optimally controlling them with a DCI controller.

We will launch “Physics of Intelligence” as a basic research project on AI. This is a project jointly developed by NTT Research Inc. of the United States and Harvard Medical Center, and will solve AI as a natural science.

We don’t know what’s inside AI today. Not knowing what’s inside is also a barrier to development. For this reason, we will investigate the internal mechanisms of AI.

For example, if you ask a generative AI to draw a pink lizard, it will draw it well, but if you ask it to draw a pink panda, it will not be able to do it at all. Why is this? Using this difference as a hint, we will mathematically solve what the imagination of AI is.

We aim to be the world’s best research institute in terms of research. Currently, we are focusing on completing IOWN and Tsuzumi. As for social implementation, we will promote valuable social implementation from both the perspective of market-in and product-out.

NTT was ranked 11th in the world last year in terms of the number of papers it has published, but has now risen to 9th. We will move closer to 5th place soon and eventually to 1st place. If we narrow down the field, there are also research fields where we are ranked 1st or 2nd in the world. We have maintained our 1st and 2nd place in the world in fields such as “optical communications,” “information security,” “neural function analysis,” and “quantum computing.”

NTT Research Inc. also published a large number of papers in fiscal 2011. In particular, NTT Group papers accounted for 14% of papers presented at the world’s most prestigious international conference on cryptography.
Japan is ranked 13th in the world in terms of the number of patent applications for generative AI. The United States and China are strong in this field, but we would like to make up ground.

※Translating Japanese articles into English with AI