I'm not arguing here for its societal value, just that dan appears to have misunderstood some fundamental bits about how it works.
> Robot 1 offers Book A for $10. It probably doesn't have it in stock, but it knows where it can buy it for, say, $9, drop-ship it to the customer without ever seeing the acutal product, and make a buck on the arbitrage. Robot 2 notices this and lists the same book for $11, hoping to do the same thing. Robot 1 notices Robot 2's listing, doesn't know or care that Robot 2 is another arbitrage trader, and reprices its listing for the book to $12. Robot 2 notices this and goes to $13, and away we go.
This makes no sense; the fundamental rule of amazon bots is that consumers buying used books almost always buy the lowest-priced used book available. So the bot war would send the price down, not up.
Why in the world would bot 2 offer the book for $11? It would know that nobody would buy it when there was one already listed for $10.
The silly-high-priced books that I saw when I was working for the book company were rare long-tail books. One of the discoveries that many people have made is that for rare books, whether or not you understand what the heck it is, some % of them will get bought even if they're listed for silly high prices. Enough people will pay thousands of dollars (not an exaggeration) for a used copy of a book that they can't find elsewhere that stocking the long tail of books can in fact be very valuable.
That means that, when you find lots of rare books, your incentive is to throw out some crazy-high price for each, because some % of them will, in fact, get paid, making the whole enterprise worthwhile.
edit: Arbitrage on the same market Does Not Work, in my experience, and to just claim that it does with no evidence doesn't make it any more likely to do so.