An Argument Against The Memory Objection?

Last time, I discussed an articulation of the memory objection against the absolute certainty thesis (the view that we can know propositions like "God exists" with absolute certainty).  These are propositions that are the conclusion of arguments and so their assessment depends significantly on the faculty of memory to be reliable. Here's is the most prominent response by my interlocutors.  Note, this is my reconstruction and extrapolation of the argument rather than a verbatim recounting of it.

1) The memory objection is true (i.e., we cannot know with absolute certainty that our memory is reliable either in general or on any given instance). 

2) If 1), then we must doubt the reliability of all instances of memory (to be consistent).   

3) But assessing the argument (and related considerations) for the memory objection itself depends on memory (i.e., to see if the memory objection is a good one, I need to employ my memory). 

4) If so, then I cannot be absolutely certain that my memory is reliable in assessing the goodness of the memory objection. 

5)This is self-referentially absurd and thus by reductio ad absurdum, the memory objection is not true.   

This seems to be a kind of transcendental argument.  I think it doesn't help at all and only causes more problems for its advocates.  I discuss some issues below. 

First, a person of my bent is not opposed to the view that absolute certainty is not possible.  This alone seriously takes the bite out of this argument.  In fact, unless absolute certainty is possible, the conclusion simply doesn't follow from the premises.  Why not?  Because the above argument depends on showing an absurdity. Premise 4) is where this is made explicit.  The thinking seems to be, that if we cannot trust our memory to be reliable (i.e., memory objection is sound), then we cannot be absolutely certain that the memory objection itself is sound.  But notice, this presents a problem only for those after absolute certainty (or arguing for the thesis that absolute certainty is possible) to begin with. But I am trying to argue that we cannot have absolute certainty, so if my argument does just that, I've succeeded at my aim.  Importantly, I am not arguing that we can be absolutely certain that we cannot be absolutely certain.  In this way, the above response fails to make any headway for my opponent.  

Detour: one objection to what has just been discussed is to respond with: "well then thought and talk are not possible" or "there is no point in talking or thinking if we cannot be absolutely certain about anything".  But how does that follow?  I don't see it. The person wielding this approach presupposes that thought and talk require absolute certainty, but where does this rule come from?  How does one know that this is absolutely true?  How can one justify it?  From what I have seen one must appeal to intuitions here.  But I certainly welcome an argument. So my first point in summation is that if my memory objection logically entails that the absolute certainty thesis is false, then this only furthers my aim.  Further, the persons pointing out that the memory objection implies that we cannot have absolute epistemic certainty may voice this as a problem with the memory objection but they first need to show why we should want or need absolute certainty for this to be an actual problem.

A second issue with the purportedly reductio argument above is that one can avoid it simply by affirming that memory is reliable when (and only when) the memory objection is being considered.  Admittedly, this is a strange result.  So if my employing the memory objection undermines the memory objection, then we can just affirm that memory is reliable but only when we assess the memory objection.  This means there is no self-undercutting going on since I am saying I can know that my memory is reliable if and only if I am presenting/assessing the memory objection.  By doing this we no longer have a self-referential absurdity which again is to tame my opponent's response.  In other words, all we can gather from the above argument is that whenever I think about or assess the memory objection, my memory is reliable (or I can know that my memory is reliable) as it pertains to my appraisal of the argument. (Note: it doesn't follow from this that memory is generally reliable or reliable on any other occasions). If so, then (granting my interlocutor a great deal) perhaps what the argument under question shows is that we can know (perhaps with certainty) the soundness or unsoundness of the memory objection (but nothing else).  

Third, we should proceed with caution here in trying to employ a transcendental or psuedo-transcendental argument.  When one argues that necessarily, 'a is a', we seem to be dealing with a necessary truth.  Further, we cannot even conceive of the denial of 'a is a'.  In fact, it is thought that 'a is a' is something of a precondition of philosophical inquiry.  If one is thinking, one cannot help but instantiate that 'a is a' (so this line of thinking goes).  But notice a significant difference with regards to the objection to the memory objection.  We should hope not to get a conclusion via this transcendental argument that necessarily, memory is reliable.  After all, we can conceive of our memories as being unreliable (indeed with think it a fact that our memories are not impeccable). If we somehow get the result that our memory is necessarily reliable, we've done something terribly wrong. After all, how can the reliability of memory be a necessary truth, when memory is a contingently reliable faculty?

If it can't be that our memory must (necessarily) be reliable, then what is the objection to the memory objection supposed to show?  Again transcendental arguments seem to give us some strong necessary  precondition to the phenomena we are dealing with.  So somewhere there must be a claim about what must be the case.  I actually think that the argument is a confused one.  This is because I think what the proponents of the objection to the memory objection are arguing for is something like this: "we must presume memory is reliable if absolute certainty is to be possible".  So the desired conclusion formulated using 'necessity' seems to be: we necessarily, presume memory reliability, if absolute certainty is to be possible.  But it's important to note the scope of the 'necessity' operator here.  As we've already seen it would be ridiculous to state that necessarily, memory is reliable (this would mean memory could not possibly fail, ever).  Instead, on the view considered, what seems to be the desired conclusion is that our presumption that memory is reliable is what is necessary.  But now, how is this supposed to assure us that we can know things on the basis of memory with absolute certainty?  How does the necessity of a epistemic presumption do this? Consider the following coherent state of affairs:

a) Bob's memory is actually fooling him right now (so that what he thinks he recalls, did not actually happen).
b) Bob has no indication that his memory is failing him right now.
c) Bob must necessarily presume his memory is reliable if he can know anything with certainty.
d) Thus, Bob is fooled by his memory, but does not know it.

I think that a)-d) describe an undesirable epistemic situation, at least if Bob is after absolute certainty.  He must, out of necessity, presuppose that his memory is working, but because he has no reasons to question the reliability, he goes on believing a falsehood.  How do we rule this possibility out in our own lives with the presumption?  Again we can't say that memory is necessarily reliable.  All we can say is that we must presume that it is.  This seems to be no achievement, but rather a sad state to be in (again if one is after absolute certainty). I don't see how this supports the absolute certainty thesis (as it regards those propositions known in virtue of memory). Once again, this seems only to lend support to the memory objection that I have suggested.