M-theory suffers from the same flaws that string theories did. First is the problem of empirical accessibility. Membranes, like strings, are supposedly very, very tiny—as small compared with a proton as a proton is compared with the solar system. This is the so-called Planck scale, 10^–33 centimeters. Gaining the kind of experimental confirmation of membranes or strings that we have for, say, quarks would require a particle accelerator 1,000 light-years around, scaling up from our current technology. Our entire solar system is only one light-day around, and the Large Hadron Collider, the world's most powerful accelerator, is 27 kilometers in circumference.
Hawking recognized long ago that a final theory—because it would probably involve particles at the Planck scale—might never be experimentally confirmable. "It is not likely that we shall have accelerators powerful enough" to test a unified theory "within the foreseeable future—or indeed, ever," he said in his 1980 speech at Cambridge. He nonetheless hoped that in lieu of empirical evidence physicists would discover a theory so logically inevitable that it excluded all alternatives.
Quite the opposite has happened. M-theory, theorists now realize, comes in an almost infinite number of versions, which "predict" an almost infinite number of possible universes. Critics call this the "Alice's restaurant problem," a reference to the refrain of the old Arlo Guthrie folk song: "You can get anything you want at Alice's restaurant." Of course, a theory that predicts everything really doesn't predict anything, and hence isn't a theory at all. Proponents, including Hawking, have tried to turn this bug into a feature, proclaiming that all the universes "predicted" by M-theory actually exist. "Our universe seems to be one of many," Hawking and Mlodinow assert