Artificial intelligence (AI) can read texts, recognize objects, and play chess – but the systems fail at questions which are trivial for people, such as "what is usually put in a trash can?" This is because AI lacks judgment and common sense. Teaching computers exactly that is the key focus of "Project Alexandria," which is being launched at the Allen Institute for Artificial Intelligence (AI2). The bulk of the $125 million pledged by institute founder Paul Allen will be invested in the project over the next three years, according to the Seattle Times . "Despite the recent achievements in the field of AI, many simple questions are remarkably difficult for artificial intelligence to solve," Institute Director Oren Etzioni commented. No system can currently provide a plausible answer to the question, "how do I know that a milk bottle is full?"
The researchers want to develop standardized methods to measure the common sense of an AI system as part of "Project Alexandria." People will help the machines develop this ability through crowdsourcing, but the systems are also supposed to learn from books and films. The goal is to significantly improve AI systems in many areas, such as robotics or machine reading.