Relativity 'improved'

[Argos]

http://www.pparc.ac.uk/Nw/EinsteinTheory.asp

A Chinese astronomer from the University of St Andrews has fine-tuned Einstein's groundbreaking theory of gravity, creating a 'simple' theory which could solve a dark mystery that has baffled astrophysicists for three-quarters of a century.

[Nereid]

ToSeeked.

An interesting pedigree (Milgrom, Berkenstein, et al.), and a good example of how science works (so I'm moving this to the Astronomy section).

[TaoTeFarrell]

wow, that is fascinating, i never realized that it would have been einstein's general theory of relativity that was incorrect. but it kind of makes sense, saying that dark matter exists is kind of like just throwing in the towel because you can't actually define the true inner workings that are at play here.

[Ken G]

But I feel I must note that just getting an empirical formula for gravity that works in galaxies is not by itself enough to rule out dark matter or disprove Einstein until you can show that the formula works over a range of mass scales that is too broad to believe dark matter could reproduce a similar agreement. In other words, if the galaxies tested have a similar ratio of matter to dark matter in all of them, it would certainly be possible to replace the presence of the dark matter by an effective modification of gravity, even though it was really the dark matter that was doing it. The real test is how the two theories scale between big and little galaxies where you might expect the dark matter to be doing something different than a modification to the spatial dependence of Newton's gravity. So I'd be looking for that type of comparison.
Also note that both Newton's and Einstein's theories "made sense" in terms of fundamental ideas that would lead naturally to their theories, whereas a purely empirical modification to gravity would not be complete until it could be derived from deeper principles that seemed natural in some way. It would be exciting to try and do that, if the scale comparisons work out. But until I see that, I'd be worried that they might have just given themselves enough free parameters to fit the degrees of freedom that are actually in the limited observations they have confronted so far.