Written by 3:02 am AI, Discussions, Uncategorized

### Introducing AI to Industry Using Photonics: NTT’s Innovative Approach

Light-based photonics rather than silicon & electricity, photonics can give us 125 times more t…

Silicon is gradually aging as time progresses. Ever since the inception of Moore’s Law in 1965, the computational capabilities of machine microprocessor (CPU) cards utilizing silicon wafer technologies have advanced significantly over the past six decades. While Moore’s Law has largely held true, it has been acknowledged that its relevance has a finite lifespan. The evolution of technology has witnessed a phase where CPUs were amalgamated in parallel, leading to the emergence of Graphical Processing Units (GPUs), a trend that persists even today, especially in the realm of cloud computing. It was inevitable that there would come a point where cramming more transistors onto a chip would no longer be feasible.

Nevertheless, in response to this challenge, alternative solutions have been devised.

The All Photonics Network (APN)

NTT, driven by a vision to revolutionize the utilization of computing devices through a series of advancements rooted in foundational infrastructure but materializing in practical real-world applications, has introduced The All Photonics Network (APN) as part of its Innovative Optical & Wireless Network (IOWN) framework, alongside its cognitive foundation and digital twin computing initiatives.

The transformation unfolding in this domain is a product of the intersection of data science, software engineering, and technology, all under the influence of optics. Photonics, a field that shares commonalities with particle electronics and leverages light as its medium, is an area that may be unfamiliar to those who overlooked the photonics module in their academic pursuits. Photonics encompasses operations such as emission, transmission, manipulation, signal processing, switching, replication, and detection of photons to govern the particles and waves constituting light.

In essence, photonics has the potential to offer 125 times greater transmission capacity than conventional silicon technology while being 100 times more energy-efficient. This nascent technology, still in its developmental stages, has the capacity to redefine the architecture of machine networks in the next decade and even the subsequent fifty years.

NTT Unveils its Array

NTT frequently showcases its latest innovations through various “showcase” configurations at its diverse R&D centers, including one in Sunnyvale that caters to a global audience spanning locations like Tokyo. This concerted effort aims to demonstrate and validate the technological propositions put forth by the company. NTT underscores its substantial annual investment of $3.6 billion in scientific and technological research and development. Central to its ongoing endeavors is the Innovative Optical and Wireless Network (IOWN) initiative, a global communications infrastructure poised to facilitate high-speed, high-capacity Internet services through optical technologies.

One such demonstration by NTT pertains to the IOWN All-Photonics Network (APN), which facilitates the connection of on-site construction equipment with remote-control systems, particularly in the domain of remotely monitoring construction equipment. This technology seeks to address prevalent challenges in the construction industry, including employee safety, workforce shortages, and extended work hours. By virtue of its high capacity and minimal latency, NTT aims to enhance safety and efficiency, with plans underway to integrate this technology across diverse operational settings.

By enabling operators in remote areas to gain real-time insights into on-site conditions with minimal latency, the APN empowers them to grasp the nuances of the working environment akin to being physically present at the site. The adoption of remote control mechanisms for construction equipment is anticipated to drive operational enhancements in the future, such as reducing operator downtime and bolstering safety standards by minimizing hazardous on-site tasks.

NTT recently published a white paper delineating the evolution of its “EquitableCore” structures tailored for the 6G/IOWN era. This architectural framework aims to reduce terminal processing burdens and fortify service privacy over the network by integrating concepts like In-Net Computing and Self-Sovereign Identity (SSI) within the core network. NTT asserts that it has effectively validated this architecture by leveraging real-world scenarios, including the utilization of a smart, resilient “typhoon-tracker” raft-canoe-buoy system engineered to venture into stormy seas to monitor tidal patterns and weather conditions. Such hardware and software innovations exemplify NTT’s commitment to pioneering advancements that resonate both in the Japanese market and globally. Concurrently, the company is also exploring novel approaches in the evolving landscape of enterprise software, aligning with trends in generative AI (gen-AI) and Large Language Model (LLM) technologies.

Enter the “tsuzumi” LLM, named after the traditional Chinese percussion instrument.

Unveiling the “tsuzumi” LLM

The “tsuzumi” LLM, as elucidated by Akira Shimada, NTT’s official director, president, and CEO, distinguishes itself through its bilingual capabilities in Asian and English languages, adaptability across diverse data modes, and cost-efficient performance metrics. Noteworthy is its superior processing speed compared to OpenAI’s counterparts, rivaling Meta in English LLM capabilities, all while operating at a mere fraction of the Graphical Processing Unit’s (GPU) capacity. Beyond text and graphics comprehension, the “tsuzumi” LLM transcends conventional boundaries by interpreting voice commands. NTT envisions leveraging this technology to address global labor shortages and societal challenges effectively.

Emphasizing the pivotal role of photonics in augmenting computational loads, Shimada underscores the significance of optics in mitigating latency issues inherent in current cloud-based infrastructures. The overarching objective of the IOWN initiative is to cultivate a data-centric ecosystem that is environmentally sustainable, given that APNs harness light and demand lower energy consumption. NTT has devised an innovative cloud environment to train the LLM model, exemplified by the “tsuzumi,” with a focus on agility to operate in controlled remote settings where data security is paramount. This lightweight approach, characterized by judicious utilization of compact technology frameworks, holds promise for sectors like healthcare, where safeguarding electronic patient records (EPR) is critical. NTT’s research and development efforts underscore a commitment to fostering a sustainable future, especially as digital twin technologies and the Internet of Things (IoT) edge closer to mainstream adoption.

Democratizing AI with a Human Touch

Shingo Kinoshita, senior vice president and head of the R&D planning office at NTT Corporation, articulated in November 2023 to media and analysts that the “tsuzumi” LLM’s operational efficiency stems from its minimal power consumption per hour. The primary objective of the “tsuzumi” LLM isn’t to serve as an all-knowing expert but rather to exhibit heightened awareness of language nuances in critical usage scenarios.

This approach resonates with the concept of private AI or small language models (SLM) advocated by various vendors. Leveraging its adaptable design models, the LLM can leverage “adapter” technology to fine-tune intelligence for specific business applications, including mission-critical functions prevalent in financial markets, thereby enhancing operational efficiency and performance.

Kinoshita is optimistic about the potential applications of this system, underpinned by the company’s meticulous curation of high-quality educational data from inception. By discerning emotional cues like rage or frustration in verbal interactions, the LLM can aid in identifying stress points in workplaces, fostering a conducive environment that benefits both organizations and employees alike.

Having evolved from its origins as a telegraph and telephone entity post-1950, NTT’s metamorphosis into a cloud-data and AI evangelist underscores the dynamic shifts in its scope, mandate, and market orientation. While NTT aspires to project a tech-savvy image akin to Silicon Valley stalwarts and global tech innovators, it remains grounded in its historical legacy and industry traditions, which, in the current tumultuous era, serve as a testament to its enduring relevance. NTT’s foray into “new tech tomorrow” underscores its commitment to innovation while preserving its brand identity, hinting at a balanced approach that doesn’t necessitate a complete reinvention.

Visited 1 times, 1 visit(s) today
Last modified: February 25, 2024
Close Search Window