This is an old, old alleged refutation of evolution, and it unfortunately rests on a gross misunderstanding of how "chance" works in this process. What would be a better analogy?
Well, suppose that tornado blew through a junk yard and just by accident, happened to put a couple pieces together that worked. What if, the next time the tornado blew through, we could preserve those pieces? And only replace them if we accidentally came up with a configuration that worked better? Now that would be more like what evolution does.
Will this take a long time? You bet. And remember, that we aren't just building a 747, we are building all kinds of things in our tornado. A watch, and a light bulb, and a chair, and all the independent little things that make up the 747, and all kinds of other stuff. We may find that eventually we get that watch to work, and that it can now be incorporated into the 747 we are building, as well as work independently. And we aren't just doing this for just a year, or even in a thousand junkyards. We are doing this for billions of years and in trillions of junkyards. When you start talking about a number with 21 zeros after it, the chances of life ending up in our current configuration don't need to be very large, even incredibly improbable, to still end up with what we've got. And when you further figure that we aren't really aiming for a Boeing 747, but only something with equivalent capabilities (i.e. an Airbus model would be an acceptable alternative), it's even more likely to end up with something acceptable.
More than this, though, this "proof" has a similar problem to it that the previous several "proofs" did as well: and that's this idea of starting out with a goal, and then trying to arrange the universe to achieve that goal. But evolution has no goal. Evolution is only going with what works. Nothing more.
For more information on the numbers game being played with the argument, read Irreligion by John Allen Paulos. As a mathematician spends a considerable amount of time with it.