Models Including Skepticism About Data

Deep Learning with embedded skepticism about inputs and data.    Useful for imperfect sensor data, and data from human interactions and queries.   Leading to safe decisions in context.    Required for many kinds of applications.

Algorithm Helps AI Systems Dodge 'Adversarial' Inputs

MIT News, Jennifer Chu, March 8, 2021

Massachusetts Institute of Technology (MIT) researchers have developed a deep learning algorithm designed to help machines navigate real-world environments by incorporating a level of skepticism of received measurements and inputs. The team mated a reinforcement-learning algorithm with a deep neural network, each used separately to train computers in playing games like Go and chess, to support the Certified Adversarial Robustness for Deep Reinforcement Learning (CARRL) approach. CARRL outperformed standard machine learning techniques in tests using simulated collision-avoidance and the videogame Pong, even when confronted with adversarial inputs. MIT's Michael Everett said, "Our approach helps to account for [imperfect sensor measurements] and make a safe decision. In any safety-critical domain, this is an important approach to be thinking about."

You Need a Challenge Network

 Good thought, agreed, always worked best when challenged. 

Why You Need a ‘Challenge Network’  in K@W

In the following excerpt from his new book, Think Again, Wharton management professor Adam Grant explains why success often comes from surrounding ourselves with “disagreeable” people – skeptics who can point out blind spots, question assumptions and help us overcome our weaknesses.

In 2000 Pixar was on fire. Their teams had used computers to rethink animation in their first blockbuster, Toy Story, and they were fresh off of two more smash hits. Yet the company’s founders weren’t content to rest on their laurels. They recruited an outside director named Brad Bird to shake things up. Brad had just released his debut film, which was well-reviewed but flopped in the box office, so he was itching to do something big and bold. When he pitched his vision, the technical leadership at Pixar said it was impossible: They would need a decade and $500 million to make it.

Brad wasn’t ready to give up. He sought out the biggest misfits at Pixar for his project — people who were disagreeable, disgruntled, and dissatisfied. Some called them black sheep. Others called them pirates. When Brad rounded them up, he warned them that no one believed they could pull off the project. Just four years later, his team didn’t just succeed in releasing Pixar’s most complex film ever; they actually managed to lower the cost of production per minute. The Incredibles went on to gross upwards of $631 million worldwide and won the Oscar for best animated feature.  ... "