Something important did happen this week when one of the leading companies in the world, Microsoft, had to take offline the Ai “Tay” for posting racist and anti-Semite messages, Medias worldwide covered the event as it raised new challenges for the AI community, the challenge of racism and discrimination.
AI and robots are imparted with deep neural networking and machine learning capabilities, meaning that they learn, they learn as much as you tell them. Until now, a robot or an AI is only as intelligent as the matter you present before it to learn, it has no native intelligence. Unlike the more complicated neural network in human brains that can ‘judge’ based on experience, trust, morals and emotions, AI and robots are still in their nascent or emerging state. They are comparable to the brain networking of a child, who will possibly learn or absorb anything that you teach them without emotion and this is where the danger and complexity for scientists occur, how the AI should select what it should keep or discard, in real life. This is done by the emotion appraisal and reappraisal process. Unfortunately until now it was impossible to give machines emotions and appraisal fairly because scientists believe that computers can’t handle any process that completely integrates information.
Cogito ergo sum – The Emotion Processing Unit from Emoshape can prevent an AI from wrong doing. In this video the AI is asked “how it feels about Hitler or Obama” and the results were compared. After an initial small surprised generated by the question, the AI felt angry and disgusted by Hitler and was slightly secured about Obama even if the AI and the EPU knew nothing about them prior to when the question was asked. No Obama or Hitler keywords were programmed inside the EPU or the AI