The Professor's Challenge

The AI field needs major breakthroughs in its thinking to achieve continuous, sensory-gathered, machine learning from the environment on unlimited subjects. The way to motivate such dramatic progress is to articulate and endorse research goals for machine behavior so ambitious that limited-domain, problemsolving knowledge representation methods are disqualified at the outset, thus forcing ourselves to produce valuable new "thoughtware". After exploring why the tendency to associate intelligence with problem-solving may be a mental roadblock to further progress in AI science, some preliminary thinking tools are introduced more suitable for sensory learning machine rcscarch These include lifelong sensorimotor data streams, representation as a symbolic recording process, knowledge transmission, and the totality of knowledge. One day, Professor Nokemoff, a distinguished AI professor of robotics at a leading institution, called in a grad student to hand out a thesis project. "Build me a robot that can sort two different parts coming down a conveyor belt.