I don't have a precise definition, but regardless which of the many candidate definitions you choose, none are "mini-computers". A computer, if that term is to mean anything at all, is Turing Complete. It's a machine that can become any other type of machine. And unless your object (or actor) happens to be emulating a Turing Complete machine, it's not a "mini-computer".
The only self-similar architecture we have in computing is the virtual machine, not the "object", whatever that word means.
What exactly do you think that an Actor cannot do? Trivially, an Actor can implement a Turing Machine. However, there are some Actors that cannot be implemented using a Turing Machine, e.g., an Actor with unbounded nondeterminism.
Difficult to know since you don't specify exactly what an "object" is. Seems more reasonable that "Actors are a recursion on concept of interacting computers."
As I said, pick your definition of object (Java objects, Smalltalk objects, actor, etc.). Unless it's "Turing Complete machine", which isn't anyone's definition of object, you can't equate objects with computers. Saying objects are computers is like saying objects are bicycles. Some of them are, yes, but that's not the point.
Of course, a computer is an Actor ;-) This would be a joke except for the fact that Actors have been defined up to a unique isomorphism. See the following:
That's an interesting result, but you'll have to connect more dots for me. I don't see how the existence of a categorical theory of Actors relates to what we're discussing.
The Smalltalk-72 model was extremely limited. Code in the language was viewed by the interpreter as simply a stream of tokens. According to [Ingalls 1983]:
"The first (token) encountered (in a program) was looked
up in the dynamic context, to determine the receiver of
the subsequent message. The name lookup began with the
class dictionary of the current activation. Failing
there, it moved to the sender of that activation and so
on up the sender chain. When a binding was finally found
for the token, its value became the receiver of a new
message, and the interpreter activated the code for that
object's class."
SENDER and stack of previous senders was retained as part of the message-passing protocol, which is problematical for concurrent systems.
Also remember that what most people think of Smalltalk (ie Smalltalk-80) doesnt seem to be what AK thinks of Smalltalk (I get the feeling that the Smalltalk-72 approach is closer to the ideals he had in mind).
See the following: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3418003