Uncovering Design Principles for Lifelong Learning AI Accelerators by Dhireesha Kudithipudi, Anurag Daram, Abdullah M. Zyarah, Fatima Tuz Zohora, James B. Aimone, Angel Yanguas-Gil, Nicholas Soures, Emre Neftci, Matthew Mattina, Vincenzo Lomonaco, Clare D. Thiem, and Benjamin Epstein.
Lifelong learning, an agent’s ability to learn throughout its lifetime, is a hallmark feature of biological learning systems and a central challenge for artificial intelligence (AI). Recent progress in lifelong learning algorithms holds promise for enabling a new generation of applications with such capabilities. As these models continue to mature, hardware requirements become paramount, driving the demand for a paradigm shift in AI accelerator design. We offer a high-level overview of features for lifelong learning accelerators and outline a program for designing custom lifelong learning systems for deployment in untethered environments.
Read the full white paper here.
Each week Tenstorrent will be highlighting a paper that has inspired our product development.