I personally have had the experience that you can train your ability to hold a mental image of what you see.
I have been drawing from life as a hobby off and on for more than twenty years, and one of the exercises that help you with learning to draw in a realistic manner is to look at your subject for several seconds and then turn to your canvas and draw what you have seen without looking back at it. With time you become better at this until you can hold a rather detailed representation of what you saw in your mind. This image is limited to what you focussed on, obviously, so parts of the background or anything you found uninteresting will be blurred or missing, and the detail is not very fine, so I can't count matches. And it most certainly does not work for random dots.
But I draw only as a hobby, spend little time on it even at my most intense phases and have long stretches of time where I don't draw at all. I believe I could get very much better at this, and don't need autims to accomplish it.
Minsky (or you) will probably counter that I'm "using mnemonics and other, non-eidetic memory enhancing exercises", but looking at your example of "quickly memoriz[ing] ... a page of text in a few seconds", I see a problem that would make eidetic memory – in the sense of storing the raw data of what you see without processing it – pretty much impossible anyway:
Only the central 2° of your field of vision are focussed and sharp. If you look at a page of text without letting your eyes roam over it, you will have a thumbnail-sized area of sharp and readable text, and a whole page of blurred text. If you want to know how blurred, focus on the dot at the end of this sentence and try to read the lines below it. So obviously, if you want to store a whole page (or any detailed image), you'll have to let your eyes "scan" it. But what do you do, when you do that? You send a row of images to your brain, and your brain has to process them to build the whole page from that. That's processing, if you want to put a name to it. So, to say this clearly: It is impossible to view a whole page without processing it.
What this means is that everybody who looks at anything through their eyes has to process the information. And what happens during this processing? No-one knows, not even Marvin Minsky. Maybe it remains raw and unchanged apart from the assembly. But maybe – or rather, very likely – there's a whole lot of top-down processing going on, making sense of what is seen, relating it to other stored information, preparing it for easy storage. Perhaps there are individuals whose brains are exceptionally quick at this, and who have a larger than average storage capacity in their working memories. From the outside this would look a lot like eidetic memory.