MightyFP - an AI Chip Accelerator to Enhance Boardroom and Battlefield AI Programs

AI technologies in the boardroom and the battlefield are hampered by existing microchips that are unfit for AI applications. Current technologies struggle to keep up with the exponential growth of AI/ML dataset and models. To address this critical Defense Need, the DAF needs efficient AI chip solutions that deliver significantly faster training speeds, more accurate inference and drastically lower energy consumption. Additionally, despite the export controls on AI chips, the US must maintain a competitive edge over China in the global “Chip Battle” for better machine-based reasoning, predictive data analytics and cognitive integration of human-machine teaming. AI chip acceleration could speed up AI computations, reduce the sizes of required models and data structures, and increase computation throughput and frequency for real USAF boardroom and battlefield applications, such as algorithmic warfare using generative AI. Arith Inc. is addressing these challenges with mightyFP™, a groundbreaking AI chip accelerator that sets new benchmarks for AI applications. mightyFP™ is the industry's first Interoperable Lifelong Learning AI Chip Accelerator, designed to comply with the forthcoming IEEE 3109 Standard for Floating Point Formats for Machine Learning. mightyFP™ aims to revolutionize AI training and inference across a wide range of Artificial Neural Networks (ANNs), effectively addressing the DAF’s challenges with prolonged training durations, expensive inference processes, tedious model conversion for inference, and loss of accuracy from post-training quantization. With mightyFP™, the DAF can enhance its AI-enabled applications under all operational imperatives, ensuring superior decision-making and operational capability in both strategic and tactical scenarios.