> I don't entirely understand where that comes from. The Standard Model and General Relativity are both tested to extreme precision. Any experiment unifying them will have to involve insane energies. It's not as if there is some other model with easy tests that we've agreed to ignore.

You're absolutely right in everything you said, thank you.

Like another commenter posted, the planck scale is 10^19 GeV and we're about 10^15 short. Therefore it follows we won't be testing anything at planck scale for many generations, if ever. Therefore the argument of "I can't test it therefore the theory is useless" is just being defeatist. The fact that such theory isn't testable might be a feature of our Universe, not the theory. As in, these people don't normally make the distinction between something that "could be tested in principle, we just don't have the technology" (like string theory) vs something that "couldn't be tested even in principle" (like how many angels can dance on the head of a pin). They're basically playing with semantics when they say "it's not testable".

> As far as I can tell, it seems to come from the developers of Loop Quantum Gravity, who feel left out of funding. And maybe that's true. But their theory doesn't offer practical tests, either. It would be weird if it did.

Again, correct on all points. However, I'll add the following. Yes, LQG makes as many directly-testable predictions at the planck scale as string theory, which is to say none, because we don't have the technology to test anything directly at that scale.

I keep repeating these things on HNs, but people here fundamentally don't understand how research in theoretical physics is done. I'll try a little exposition:

Physics is: make experiments, and try to infer which laws/rules/formulas are common to all experiments or sets of similar experiments, and their domain of applicability. These are called theories.

Theoretical physics is: think about theories, and try to observe which laws/rules/formulas are common to all theories or sets of similar theories, and their domain of applicability. These are more general theories from which your directly-experimented theories can be derived. You can keep interacting constructing ever more general theories from an ever smaller set of principles.

So a lot of theoretical physics is about arguing which of the principles that you know are true because you've experimentally tested them will hold in circumstances where you can't directly test. As it turns out there's a lot that you can infer about things you've never seem because often times mathematics puts constraints on how different ideas work together.

The string theory/LQG thing is that LQG start from the guess that Lorentz invariance doesn't hold at planck scale. The reason why LQG is less appealing to a lot of physicists is that if you follow this through you can never quite make it mathematically self consistent. In string theory what happens in certain sub-domains is that you start with a lot of arbitrary possibilities, but then you demand certain types of mathematical self-consistency and magically it points out that there's only one or a small number possibilities. A classic example is: "how many dimensions does the universe has?" which no theory really gives as answer, but string theory at least points in a direction: "if you assume such and such, the the allowed answers are such and such". This happens a lot in string theory, and it's what drives people to keep digging. String theory on the other hand concludes that Lorentz invariance must hold at all scales "in some string-like theories" if you demand cancellation of divergences, which you must have for your theory to be renormalizable and therefore mathematically self-consistence. So in a sense this is a prediction of string theory. Not that LQG doesn't predict the opposite, that Lorentz invariance doesn't hold. Instead it assumes that it doesn't. String theory instead predicts that it does. The latter is much more impressive; anybody can start from an arbitrarily picked assumption that noone can prove wrong.