The LSI chip is specifically designed to process AI inference workloads without the need to compress data or rely on cloud-based processing.
The Japanese giant touts the hardware as a low-power option for edge and power-constrained terminal deployments.
For example, NTT suggests the chip can detect individuals or objects from up to 150 metres (492 feet) above the ground when installed on a drone. The firm contends that conventional real-time AI video inferencing technology would limit a drone’s operations to about 30 metres (98 feet).
In edge and power-constrained terminals, meanwhile, the chip is reportedly capable of overcoming power consumption constraints caused by the GPUs.
Using a proprietary AI engine, the hardware “reduces computational complexity while ensuring detection accuracy, improving computing efficiency using interframe correlation and dynamic bit-precision control”.
NTT plans to commercialise the chips within the fiscal year 2025 through its Innovative Devices Corporation brand.
The company’s researchers are working with NTT Data to enhance the new hardware with its proprietary Attribute-Based Encryption (ABE) technologies, which could enable the chip to share data securely into existing applications and data stores.
Beyond the new chip’s unveiling, NTT’s Research division has unveiled a new group tasked with advancing the “physics of AI”.
The new group will take an interdisciplinary approach to understanding AI and will feature physics, neuroscience, and machine learning experts. It will be led by NTT Research Scientist Dr. Hidenori Tanaka, who recently led the PHI Lab’s Intelligence Systems Group.
RELATED STORIES
NTT Data expands agentic AI offerings with a new custom agent suite
Japan’s largest bank uses photons to prevent IT outages
NTT Docomo backs Ayar Labs' optical I/O tech for AI data transmission
