He writes,

computers still are not meta-rational. They do not understand what they do not understand very well

But could one argue that Watson is Meta-rational, in that it assigns a probability to having the correct answer? Perhaps this feature should be added to other computer tasks, such as chess or trading stocks.

Note that Kasparov’s evaluation of Watson, to which Tyler links, explicitly says that Watson still was not capable of knowing when it was about to make a blunder.

I agree that this is a very important issue. Any decision-maker, human or computer, that does not realize when it is not well positioned to make a good judgmen, is dangerous. One could say that it is this lack of meta-rationality that I find most frightening about the Obama Administration’s policies. The policy makers assume that they know much more than they really do.

By the way, the flash crash happened on my birthday last year, in case you had forgotten about it (either the episode or my birthday).