Developmental learning machine and method

A machine having developmental learning capability, said machine comprising:

• one or more sensors for sensing an environment of the machine and generating one or more sensed signals in response thereto;
• one or more effectors for acting on one or more objects in the environment;
• a sensor based level builder having one or more level building elements, said sensor based level builder receiving as an input successive frames of said sensed signals one at a time, and generating action signals each having a relative probability, the sensor based level builder autonomously generating representations of tasks to be learned from said one or more sensed signals;
• and
• a confidence accumulator for receiving said action signals and accumulating confidence of said action signals based on priority to determine most probable action signals, said confidence accumulator producing action control signals to control said one or more effectors in response to said determined most probable action signals, wherein said machine learns directly from continuous unsegmented sensory streams on-line while performing an operation and learns new tasks of unconstrained domains without a need for reprogramming, and wherein said learned new tasks include tasks that are not predetermined at the time of machine programming, said learned new tasks comprising at least two of autonomous navigation, speech recognition, and object manipulation.