RT @shinboson: researching AI safety on the side of AI