NTT launches AI collection physics and a 4K inference AI

NTT Research announced at the annual upgrade event that it started a new collection of artificial intelligence basic research, called “” Artificial Intelligence Group Physics.

Physical artificial intelligence became a big deal in 2025, with Nafidia leads the charge To create artificial data for self -driving cars and human robots so that they can reach the market faster. NTT Research launches a set of physics of artificial intelligence (PAI) to reach the plane.

The new independent NTT Research group exits its physical intelligence laboratory (PHI) to develop our understanding of the “Black Square” of artificial intelligence to improve the results of confidence and safety. NTT Research, which has an annual R&D budget worth $ 3.6 billion, is a section of NTT, the Great Communications Company in Japan.

Last year, NTT created its vision of “intelligence physics”, which was initially formed in cooperation with the Harvard University Center for Brain Science, the main contributions made over the past five years, and continuous cooperation with academic partners.

The new group will lead Hidenori Tanaka, NTT research scientist and physics expert, neuroscience, and machine learning, in the broader pursuit of human/artificial intelligence.

The new group will continue to progress in a multidisciplinary approach to understanding the leading Amnesty International for the team over the past five years.

Early, the PHI laboratory realized the importance of understanding the nature of the “Black Box” of artificial intelligence and machine learning to develop new systems with great improved energy efficiency. As artificial intelligence has now advanced at an amazing rate, merit and safety issues are also decisive to industry applications and the governance of artificial intelligence adoption.

In cooperation with prominent academic researchers, artificial intelligence group physics aims to address similarities between biological and artificial intelligence, increasing the complexities of artificial intelligence mechanisms and building confidence that leads to more harmonious integration of human cooperation and AI. The goal of this is to have a better understanding of how artificial intelligence in terms of training, accumulation of knowledge, and decisions so that we can design artificial intelligence, secure, safe and worthy of confidence in the future.

This approach repeats what physicists have done for many centuries: People understood that things move when the forces are applied, but it is physics that revealed the fine details of the relationship, which allowed humans to design the machines that we know today. For example, the development of the steam engine has reached our understanding of the thermal dynamic, which in turn enabled the creation of advanced semiconductors. Likewise, the work of this group will constitute the future of artificial intelligence technology.

The new group will continue to cooperate with the Harvard University Center for Brain Science (CBS), led by Professor Harvard Venkatish Mourthi, and with an assistant professor at Princeton University (and former NTT research scientist) Gotam Reddy. She also plans to cooperate with a assistant professor at Stanford University, Syria Ganguli, who participated in Tanaka with many papers. The main team of the group includes Tanaka, NTT Research Mia Okawa and NTT after a PhD research, EKDEP Singh Lubana.

Previous contributions include:

• The widespread toning algorithm (more than 750 quotes in just 4 years)
• An algorithm to remove bias for large language models (LLMS), recognized by the National Institute for Standards and Technology (NIST) for his scientific and practical vision; and
• New visions in the dynamics of how to learn artificial intelligence concepts

To move forward, I have a three -sided artificial intelligence group physics. 1) It intends to deepen our understanding of the mechanisms of artificial intelligence, all that is better to integrate ethics from the inside, and not through a mixture of precise control (i.e. forced learning). 2) Borrowing of experimental physics, will continue to create a systematic control of artificial intelligence and monitor learning behaviors and predicting AI step by step. 3) It aspires to heal the breach of confidence between artificial intelligence and human operators by improving processes and controlling data.

“Today it represents a new step towards society’s understanding of Amnesty International by creating NTT Research in the artificial intelligence group,” Kazu Jumi, President and CEO of NTT said in a statement. “The emergence of artificial intelligence solutions and rapid dependence in all areas of daily life had a profound impact on our relationship with technology. With the continued growth of artificial intelligence, it is necessary to explore how artificial intelligence feels and how this progress can be in new solutions. The new group aims to remove concerns and vitality about AI solutions forward forward for the compatible path.

The artificial intelligence group physics adopts a multidisciplinary approach to artificial intelligence, with physics, neuroscience and psychology. This approach seems to go beyond traditional standards, while realizing the need to support goals such as fairness and safety that leads to the adoption of sustainable artificial intelligence. Regarding energy efficiency, other groups of PHI are already involved in efforts to reduce energy consumption of artificial intelligence computing platforms through visual computing and delicate TFLN technology. Moreover, inspired by the vast difference between WhatsApp consumed by LLMS and the human brain or animal, the new group will also explore ways to take advantage of the similarities between biological minds and artificial nerve networks.

Tanaka said in a statement, “The key to Amnesty International to reach the side of humanity is worthy of confidence and how we deal with the design and implementation of artificial intelligence solutions,” Tanaka said in a statement. “With the emergence of this group, we have a way forward to understand the calculations of the brain and how it is linked to deep learning models. It looks at the future, and our research hopes to achieve more algorithms and natural smart devices through our understanding of physics, neuroscience and machine learning.”

Since 2019, PHI LAB has led the search for new ways to compute systems by taking advantage of the technologies based on photonic. TFLN -based devices are explored through this effort, while the coherent ISING machine provides new views on complex improvement problems that are difficult to solve historically on classic computers.

In addition to a joint research agreement (JRA) with Harvard University, the Phi Laboratory has worked over the years with the California Institute of Technology (Caltech), the University of Cornell, Harvard University, the Massachusetts University of Technology (MIT), the University of Notre Dame, the University of Stanford, the University of Swinburn Technology, the University of Michigan and the University of NASA Research. In general, the PHI laboratory presented more than 150 sheets, five appears in nature, one in science and twenty in Nature Sister magazines.

NTT announces the inference of artificial intelligence to process 4K video in actual time

NTT Corp has also announced a new broad integration (LSI) to treat infection in the actual time of a high -resolution video of 4 kg and 30 frames per second (FPS). This low -power technology is designed for energy -restricted peripheral publication in which traditional artificial intelligence concludes highly high video pressure for actual time processing.

For example, when LSI is installed on a drone, the drone can discover individuals or things from up to 150 meters (492 feet) above the ground, the maximum legal aircraft flights in Japan, while the traditional AI video inference technology in the actual time will limit drone operations to about 30 meters (98 feet). One of the applicants for inspection on the infrastructure -based infrastructure includes operations that exceed the visual visual line of the operator, and reduce employment and costs.

“A mixture of low -energy artificial intelligence with a high -resolution video carries a huge
Jumi said in a statement: “The amount of capabilities, from examining the infrastructure to public safety to live sporting events.” In a statement.

At power -resting stations, artificial intelligence devices are limited to energy consumption, which is less than the size of the graphics processing units used in artificial intelligence servers; Dozens of watts by the first compared to hundreds of watts by the latter. LSI overcomes these restrictions by implementing the engineering engine of Amnesty International. This motor reduces arithmetic complexity while ensuring accuracy of detection, improving computing efficiency using interface and dynamic control of accuracy. The implementation of the algorithm detecting objects that only once (Yolov3) using LSI is possible with energy consumption less than 20 watts.

NTT is planning to market this LSI during the 2025 fiscal year through NTT Intriftative Device Corporation. NTT announced and this LSI in Upgrade has shown the company’s annual research and innovation. The upgrade 2025 was kept in San Francisco 9-10 April 2025.

Looking at the future, the researchers study the LSI application on the data infrastructure initiative (DCI) for the Innovative Optical and Wireless Network initiative (IONN) led by NTT and the IOWN Global Forum. DCI enhances the high-speed and low-technology capabilities of the IWN All-Photonss to meet the infrastructure challenges of modern networks, including obstacles that prevent expansion and restrictions in performance and high energy consumption.

In addition, NTT researchers cooperate with NTT Data, Inc. On the progress of this LSI with regard to the ABE coding techniques (ABE). ABE allows accurate control to access granules and prepare a flexible policy in the data layer, as common secrecy encryption techniques allows the participation of safe data that can be combined into current applications and data stores.

IDN ID

Yesterday, NTT announced that Akira Shimada, President and CEO of NTT, and Katsuhiko Kawazoe, First CEO and CTO in NTT published a book, IDN IDIn which they discuss the IWN (the innovative optical and wireless network) that NTT is universal
Technology Commander.

The newly translated book is exploring the NTT vision of IWN and how it will enable it to be a more sustainable society in a world -based world.

“IDN ID” is now available on Amazon after publishing during the NTT and upgrade annual research and innovation summit. The upgrade 2025 was kept in San Francisco 9-10 April 2025.