I was once involved in a project where this sort of thing was going on, and those that had the better looking GUI got the nod.
I'll settle for the one that HAS A GUI.
This guy's advice is good for his target services and audience. In particular: he's giving advice with no existing relationship with a programmer, who are going to jump feet-first into elance, guru, odesk, and vworker. These first-time users of the services will be lucky to see a project through to a remotely satisfactory conclusion if they only hire one programmer.
Having been in that position once, I can vouch for what he's saying: you'll be ignorant enough of the criteria you should be using, that it's going to be very useful to hire 2-3 programmers for a small milestone up front. You'll get the guy who completely ignores the spec, and sends you something he cobbled together in 10 minutes. You'll get the guy who eats up the full time allotted, then at the very end cancels the project and refunds the money. And you'll get the guy who's actually a solid, communicative programmer, and gets the job done. Then you go forward with that guy.
It's going to be rare that, on your first use of these services, you can make a decision based on a close call, evaluating code quality, UI design, and so on. No, you'll be evaluating "did the project get completed AT ALL?". And you'll learn a lot about how to find people using these services, and how to write better specs yourself, so that you don't waste everyone's time in the future.