Machine learning revolutionizes software development

Revolutions in software development are pretty rare. Developing for the web probably qualifies. Mainstream acceptance of garbage collection as well, probably. This seems overblown.

How is this different from tools like Acovea? http://www.coyotegulch.com/products/acovea/

More of a search problem than a learning problem if that's what's really going on.

Much better explanation here: http://ctuning.org/wiki/index.php/CTools:MilepostGCC

The basic idea is to use machine learning to pick the best set of gcc compiler options for a program. A better optimizer for a compiler is useful but it's hardly a revolution.

Edit: to clarify, this isn't just choosing command-line flags, but using machine learning in the compiler's internals.

little bit of sensationalism here...

"compiler optimisation has become a bottleneck in the development process"

Certainly not true for arbitrary development.

"an automatic way to optimise compilers for re-configurable embedded processors"

and they are working on a particular special case...

How is this better than JITting?

It seems like most recent performance gains come from runtime/dynamic optimizations, not static/compile time optimizations.

Ten times faster? Ten times? I find that completely implausible.