Spintronics in the age of ChatGPT

The launch of ChatGPT has been a remarkable milestone in the development of artificial intelligence (AI) and has captured widespread attention. However, current computing hardware presents limitations for AI development, particularly for edge computing. Neuromorphic computing systems, utilizing in-memory analogue computation, are a promising solution that is attracting significant interest. Among them, a spintronic realization based on a magnetic tunnel junction (MTJ) crossbar is gaining momentum. Unfortunately, the current MTJs are constrained to only two resistance states, posing significant limitations. In this talk, I will present our research efforts towards developing a new type of MTJ that supports multiple resistance states, facilitating the realization of a significantly more efficient spintronic crossbar to be used in future AI applications.