Creating autonomous intelligent robots with perception, cognition, and action, able to coexist with humans can be viewed as the ultimate challenging goal of artificial intelligence. Approaches to achieve such a goal that depend on rigid task and world models that drive precise mathematical algorithms, even if probabilistic, are doomed to be too restrictive, as heuristics are clearly needed to handle the uncertainty that inevitably surrounds autonomy within human environments. Instead we need to investigate rich approaches capable of using heuristics and flexible experience-built webs of knowledge to continuously question and revise models while acting in an environment. Significant progress depends upon a seamless integration of perception, cognition, and action, to provide AI creatures with purposeful perception and action, combined with the ability to handle surprise, to recognize and adapt past similar experience, and to learn from observation. Hence and interestingly, I find that the achievement of the ultimate goal of the field requires us, researchers, to accept that AI creatures are evolving artifacts most probably always with limitations, similarly to humans. Equipped with an initial perceptual, cognitive, and execution architecture, robots will accumulate experience, refine their knowledge, and adapt the parameters of their algorithms as a function of their interactions with humans, other robots, and their environments.