Redpoll and Rutgers University have entered into a 42-month collaborative agreement with DARPA to develop OTACON, an introspective AI system that thrives in dynamic, uncertain worlds. The OTACON project, lead by Dr. Patrick Shafto at Rutgers and Dr. Baxter Eaves at Redpoll, falls under the Science of Artificial Intelligence and Learning for Open-world Novelty (SAIL-ON) program, which aims to move AI systems out of closed-world problems with static constraints, into open worlds with dynamic constraints.
Current artificial intelligence (AI) systems excel at tasks defined by rigid rules – such as mastering the board games Go and chess with proficiency surpassing world-class human players. However, AI systems aren't very good at adapting to constantly changing conditions commonly faced by troops in the real world – from reacting to an adversary's surprise actions, to fluctuating weather, to operating in unfamiliar terrain. (SAIL-ON Website)
OTACON seeks to solve this problem by making AI introspective and malleable. According to Co-PI, Dr. Baxter Eaves:
OTACON is introspective and malleable. Whenever it sees something new it asks itself, "does this make sense given what I currently know?" If "no", the machine will adapt. It determines where its knowledge has broken and why, and rearranges its structure to override the old outdated knowledge while preserving the still useful knowledge.
Eaves says OTACON is just one set of capabilities Redpoll is developing toward building machines that are partners to humans, not just tools.
Novelty is in the eye of the beholder. And because we want our systems to be interpretable and intuitive to humans — so they can work with humans — we have used our research in human cognition to ensure that OTACON can learn and store its knowledge similar to the ways people do.
Redpoll's mission, says Eaves, is "...to make machines the are safe and reliable partners to human decision makers, not just tools". To learn more about Redpoll's AI philosophy and platform capabilities, visit www.redpoll.ai.