Actually, that is wrong.
First, some perspective.
Quantum mechanics is not a theory, it is a set of construction rules for theories that reflect reality. Whatever quantum field theory you construct of whatever operators you care to name (the rules of QM/QFD being most easily phrased in terms of operators), some of those operators do not commute [A,B] != 0, and some of those operators do not anti-commute {A,B} != 0.
We happen to call those bosons and fermions respectively.
String theory, being a quantum field theory including gravity (hence, a superset of general relativity), is also a set of construction rules for theories of reality.
As it so happens, some of those theories produce large extra dimensions of > than the millimeter scale. In general, we characterize those theories by a parameter space, and theories with a parameter space producing large extra dimensions of a particular type *are* disallowed by observation.
If observation disallowed the entire parameter space of string theory, it would be entirely falsified.
Instead, it is something dubbed "The Landscape".
Secondly, it is interesting, but not informative, to listen to opinions of people who have never studied string theory remark on its structure/usefulness.
In fact, string theories do provide some elegant solutions to things like supersymmetric (quantum) gravity (SUGRA), supersymmetry (SUSY), and a good number of other subtle points that are only understood with sufficient background in QFT, GR, etc.
Everyone laughs at "the internet is a series of tubes", but then we get long expostulations about string theory from people that haven't studied it.
Third, remember that the authors of the article (e.g. Lee Smolin) have a somewhat vested interest in denying the usefulness of String Theory, as it allows their own theories (i.e. Loop Quantum Gravity) to be more prominent.
As a matter of my own opinion, LQG has even more difficulties than String Theory, for example, it assumes the spacetime metric is smooth even at quantum scales, and that a classical theory (GR) can be "quantized". As GR is not renormalizable, this is violently at odds with known, tested QFT, which is accurate to 11 decimal places.
