I'm not keen on the quote either, but I think you did miss one way in which it (or something like it) could be right.

Consider statements like these: "X would do better if it had more colour-blind people working there", "Y would do better if it had more deaf people working there", "Z would do better if it had more depressed people working there". At first sight, these are obviously wrong: being colour-blind, or deaf, or depressed, is a disadvantage, right? But that's waaaay too simple. A colour-blind person has the ability to see what something looks like to other colour-blind people, and someone with normal vision doesn't. They may also have better perception of some kinds of detail than normal trichromats. Similarly (but more so) for deaf or depressed people. The point here: "person with disability X" does not, by a long way, equal "'normal' person with such-and-such capabilities simply removed".

So, rationality. Let's stipulate that perfect rationality, if it could be achieved, would bring with it the ability to do any cognitive task at least as well as a less-rational person could; and that rationality doesn't imply autism, social awkwardness, lack of emotions, and all the other tiresome tropes you find in movies and books and TV. Even so: if you take a sample of "rational people" and a sample of "irrational people" by any realistic means, you will find other differences between them other than their level of rationality -- and it is not at all obvious that those differences will all be to the advantage of the "rational people".

Obvious example: you want to know how people will react to an advertisement, a web page, a dialog box, or whatever. In principle, someone sufficiently smart could work that out, given enough information about the population that's going to see the ad/page/dbox. But if you just get a sample of people representative of that population -- and, let's face it, many of those people will not be very rational -- you can just show them whatever-it-is and see how they react.

So, a company that's entirely filled with rational people is in practice liable to be missing some viewpoints that might be useful to know about. Sure, if those were infinitely intelligent perfectly rational people with complete information about everything, they could work it all out. But back in the real world, that's not going to happen.

I think that's probably what he was getting at. Only he said it in one sentence instead of four paragraphs. Draw from that whatever conclusions about "rational people" (like me) versus "irrational people" (like Edwards?) you find appropriate :-).

Doing something like entrepreneurship is, on the face of it, irrational. Not to say that there aren't plausible sets of values which would rationally take on that much work for such a low chance of large reward--but most real people who do it overestimate their chances of success.

The net effect seems to be positive for the world, but you'd need everyone to be extremely rational (basically TDT level) before they'd do something like flipping a coin five times and doing a startup if it comes up all tails.