Having a grab bag of algorithms did not make an idea. I suspect the author thought more of responding to user's intentions not software doing whatever it wills.

Having a grab bag of algorithms did not make an idea. I suspect the author thought more of responding to user's intentions not software doing whatever it wills. This being a multitude of hard AI unsolved problems layered one on another. Such as building a fitness function or goal based on natural language input combined with context and refinement. I know if two semi successful very limited attempts, one being Google DeepMind AutoML, the other being Bayou. They're really not too useful in practice anymore. (State of the art surpassed anything they are capable of.)