There was some discussion about the role of modal or counterfactuals in a proper definition. Istvan Berkeley drew attention to the case of a computer which comes off the assembly line but is never switched on:
[Berkeley:] Under such conditions, the notion of ``changes in pattern'' would not be relevant as there would never be any patterns present to change! Such a device would be a pretty much prototypical example of a ``computer,'' even though it would never perform a computation. What I think this shows is that the notions of ``computation'' and ``being a computer'' are to some degree independent of one another. Perhaps the best way of describing the relationship between the two would be to say that something X is a computer just in case it is possible that X may perform a computation (understood in Hayes' sense).(This is what I meant to suggest by saying that patterns can be put into it.) Bringsjord (1995), in a previous paper on this topic, develops this point in more detail. As he emphasises there, it is important to choose the right modality. Logical possibility would be too weak, for example. Physical possibility seems about right; but given this, Bringsjord claims that we would not need to add Berkeley's qualifier ``in Hayes' sense,'' but simply understand ``perform a computation'' in the traditional sense, referring to Turing machines.
This does not seem to avoid the Putnam construction, however, since there seems to be nothing physically impossible in interpreting successive moments in a pebble's history as states of a finite-state machine, no matter how ridiculous this may seem. Valerie Hartcastle also had worries about how to choose the right kind of possibility:
[Hartcastle:] I worry though that (Bringsjord's) definition isn't too broad as well. That is, I don't think that it would block the very examples that Searle uses (at least in his public talks on this matter) of a set of molecules instantiating Wordstar. (Some of this might turn on how we define the notions of agent or observer.) Okay, so we move to the most restrictive notion of possible: humanly possible. I see this as too strong, for exactly the sort of reasons McGinn gives for us not being able to understand consciousness: we are too stupid. I would think that we would want to count as a computer some things that we in fact can't manage to analyze (because of some inherent physical limitations: we are too stupid, or too slow, or we don't live long enough, or whatever). So, I am looking for a notion of possibility somewhere between physically possible and humanly possible.The case of the machine which is never switched on seems to arise for any definition of computer, rather like an automobile engine which never runs, or a hat which is never worn. I confess that this seems to me to be a rather unrewarding issue. My inclination would be to seek to define the notion of ``working computer,'' and let philosophical logicians quarrel about such cases.