Let us accept that science has an answer. The self then would be a representation of one's own body in the context of the environment. However if that is true wouldn't it imply that something is aware of this self representation. Then the burden of definition of the self falls upon that aware entity. That entity couldn't possibly be yet another representation if it were to be fundamental. Maybe what I'm trying to do here is to reduce the problem.
Consider an infant just born. The infant has no information about the physical world besides a few reflex reactions genetically encoded in its brain. As it grows, it will form mental relations between objects and the sensory input it receives to get a feel for the basic laws of physics, throwing a ball, touching something hot etc. Soon enough it will be smart enough to know the physical premises of the source of those sensory signals to get an idea about the extent of his body. And from here onwards he begins to see the world in relation to this self. But somewhere along the way he forgets what was initially there, and that is the set of experiences that gave shape to his current understanding of his physical body. Could we therefore argue that what he calls his 'self' is nothing but these sets of experiences and their mental deductions that led him to form a boundary between him and the world?
Can we then not say that there is no way for a computer AI to become self aware unless we provide it with a body with sensors with which it could interact with the environment? In other words the self only exists in relation to the environment and not in isolation? How does pinpoint the 'self'?