Human/Machine Teaming: Dancing with the Bear
Discipline: Human Factors
With the advent of Foundation Models (e.g., large language models like ChatGPT) and stunning successes such as AlphaGo’s creation of a new and surprising winning strategy, dire predictions of human obsolescence have reemerged. However, current developments are in line with past AI trends which suggest that AI will be amazing but with regard to real-world operations (i.e., not games), there will still be a non-negligible (10-20%) portion of operations where AI will perform poorly or simply fail. These situations will continue to require a human to make the overall system work successfully. In a previous academy talk, I discussed how best to use a human not only in these situations but many others as well. In this talk I will expand on certain aspects of human/machine teaming including how to make the most of your machine.
About the Presenter:
Paul Schutte is a Principal researcher in Human-Machine Teaming in Applied Cognitive Science at Sandia National Laboratories in NM. He has worked at Sandia for 5 years. He is currently leading research efforts in Human Machine Teaming with regard to Foundation Models (e.g., GPT) and Machine Learning, Function Allocation, Trust, and Transparency. Prior to Sandia, he worked 35 years for NASA LaRC developing AI decision aids and cockpit interfaces for commercial aviation. He has expertise in Human-Machine Teaming, Naturalistic Decision Making, Function Allocation, and Aviation. Schutte has a MS in Experimental Psychology and an MS in Computer Science.