First a bit of background on QM vs GR, then I'll return to your point about horizons.
It's not that they disagree, it's just that when you do perturbative quantization of General Relativity you find out you can't use the renormalization (by power-set counting) that is done on other perturbatively quantized field theories. This is only a problem in strong gravity -- where in this context, "strong" means more than one loop of gravitons on a Feynman diagram. With multi-loop Feynman diagrams involving other massless propagators we can use a number of techniques to reduce the integrals; these techniques do not work with (massless) gravitons.
One could describe this form of renormalization as a enabling a reduction of infinite modes by making a finite set of measurements at high energy to fix a parameter; this works well for light, for example. Einstein gravity, on the other hand, is perturbatively non-renormalizable, at least using the same techniques, because you need to make an ever increasing set of these types of parameter-fixing measurements at higher energies, and ultimately in strong gravity (as defined above) you need an infinite number of them.
Since there is an overwhelming amount of evidence for Einstein gravity everywhere we have been able to look so far, this poses a problem: how do you quantize gravity in a way that makes useful predictions about systems in strong gravity? There are lots of research programmes looking at this problem.
(The opposite side of the coin is that programmes to geometrize the fields of the Standard Model exist, but they run into difficulty too. Taken together, this is the underlying situation supporting the claim that QM and GR disagree about QM scale physics where local curvature distorts lengths and times at the scale of SM scatterings.)
Outside of strong gravity, however, perturbatively quantized gravity on the one hand, and semiclassical gravity on the other, are perfectly fine effective field theories (effective field theories in the Kenneth G Wilson sense; General Relativity as an effective field theory in this sense I'll call "the EFT" below).
One of the features of black holes generically is that strong curvature is found only very near the (unremovable by change of coordinates in General Relativity, gravitational) singularity. The event horizon or horizons, depending on metric the black hole sources, are found further from the singularity as the black hole's mass increases. Around an arbitrarily high mass black hole, one will find the horizon is in a region of arbitrarily flat spacetime. If the compact massive objects we observe to date are astrophysical black holes, they do not have strong gravity at the horizon.
This is one of the critical parts of the AMPS firewalls "paradox", wherein one of the set of properties of a black hole which cannot all be true is the no-drama conjecture: an infaller at a sufficiently massive black hole will experience no tidal stresses.
Afshordi, one of the authors of the paper that's the topic of the article has been doing productive phenomenology in his various collaborations, and his argument that there are classes of quantum gravity theory that can be excluded if the observations do not disappear under analysis (e.g., they're not systematic errors, and aren't "lucky" low-sigma correlations).
It's more than a bit provocative to suggest that the observations suggest that the EFT fails outside the horizon (one of the possibilities that has been discussed by Polchinski and others in the wake of AMPS (he's the "P")), which is fine, since the point will stand or fall on the basis of evidence rather than the consequences for various QG research programmes. :-)
For now it's pretty safe to work on the assumption that the EFT is fine at least everywhere outside the event horizon and the extremely hot dense early universe, and that a UV completion to GR will be completely compatible with the EFT outside strong gravity and inside a region at least a little bigger than the observable universe.
Finally, all this means that although I won't exactly endorse your wording in your third last sentence, I do certainly agree with the sentiment.
Any reading suggestions for someone who wants to understand this and has a very strong math background but no physics background beyond basic second-year QM?
It's hard to gauge (pardon the pun) where to start you based on that.
I'll guess that you're keen on understanding the General Relativity part in detail.
Carroll and ‘t Hooft have kindly put up lecture notes that might be a good starting place. Stefan Waner has made available good lecture notes on differential geometry in the GR context.
If you can wrap your head around those you could proceed to any of the standard grad texts on GR (MTW, Wald, Weinberg mainly). Weinberg is popular with people who like concise maths.
If it's all too novel, then Hartle, Schutz and Carroll all have excellent introductory texts aimed at grad students.
Once you understand how General Relativity works as a general background to any field theory -- classical or quantum -- then you'd be ready for semiclassical gravity or various quantizations of GR.
An alternative approach might be to aim you instead towards QFTs via group theory, Lie groups, Yang-Mills theory, renormalization, renormalization group flow, and so forth.
Eventually you hit on gauge/group correspondence arguments in general, which will equip you to understand the attractions of AdS/CFT in moving the tedious calculations from one setting to another setting in which they're a lot less tedious, and hopefully not fall too hard for the idea that AdS/CFT automatically helps us with gravity and matter theories in our universe.
There is certainly ample scope for talented mathematicians to test the correspondence argument (and especially whether AdS/CFT specifically or gauge/gravity generally really is a duality) rigourously.
I think that'd cover all the ideas touched on in comment you replied to.
PS: Sorry I meant to list off some QFT resources for you but I have run out of time today. :(
I think that as you have some rough exposure to relativity already, you could first absorb the idea that Minkowski (flat) spacetime is a theory where at every point the Poincaré group is the isometry group. That's a good way to hit on representation theory.
However, Lancaster and Blundell's book has some reviews suggesting that someone good at math should be able to work through it without the background needed by textbooks like Srednicki's https://www.dur.ac.uk/physics/qftgabook/ (I have not read it though).
Thanks again. Your suggestions led me to scan the QC174.45 shelves at a nearby university library. I settled on Maggiore's A Modern Introduction to Quantum Field Theory, which seems to be almost all Math.
If a Markov chain could generate something that long and coherent, it would be a much greater find than a simple explanation of gravity in General Relativity.
If it's any help, I can, and it's not. I just learned a ton from that post, in terms of clarifying the framework for things I already know a little about.
It's not that they disagree, it's just that when you do perturbative quantization of General Relativity you find out you can't use the renormalization (by power-set counting) that is done on other perturbatively quantized field theories. This is only a problem in strong gravity -- where in this context, "strong" means more than one loop of gravitons on a Feynman diagram. With multi-loop Feynman diagrams involving other massless propagators we can use a number of techniques to reduce the integrals; these techniques do not work with (massless) gravitons.
One could describe this form of renormalization as a enabling a reduction of infinite modes by making a finite set of measurements at high energy to fix a parameter; this works well for light, for example. Einstein gravity, on the other hand, is perturbatively non-renormalizable, at least using the same techniques, because you need to make an ever increasing set of these types of parameter-fixing measurements at higher energies, and ultimately in strong gravity (as defined above) you need an infinite number of them.
Since there is an overwhelming amount of evidence for Einstein gravity everywhere we have been able to look so far, this poses a problem: how do you quantize gravity in a way that makes useful predictions about systems in strong gravity? There are lots of research programmes looking at this problem.
(The opposite side of the coin is that programmes to geometrize the fields of the Standard Model exist, but they run into difficulty too. Taken together, this is the underlying situation supporting the claim that QM and GR disagree about QM scale physics where local curvature distorts lengths and times at the scale of SM scatterings.)
Outside of strong gravity, however, perturbatively quantized gravity on the one hand, and semiclassical gravity on the other, are perfectly fine effective field theories (effective field theories in the Kenneth G Wilson sense; General Relativity as an effective field theory in this sense I'll call "the EFT" below).
One of the features of black holes generically is that strong curvature is found only very near the (unremovable by change of coordinates in General Relativity, gravitational) singularity. The event horizon or horizons, depending on metric the black hole sources, are found further from the singularity as the black hole's mass increases. Around an arbitrarily high mass black hole, one will find the horizon is in a region of arbitrarily flat spacetime. If the compact massive objects we observe to date are astrophysical black holes, they do not have strong gravity at the horizon.
This is one of the critical parts of the AMPS firewalls "paradox", wherein one of the set of properties of a black hole which cannot all be true is the no-drama conjecture: an infaller at a sufficiently massive black hole will experience no tidal stresses.
Afshordi, one of the authors of the paper that's the topic of the article has been doing productive phenomenology in his various collaborations, and his argument that there are classes of quantum gravity theory that can be excluded if the observations do not disappear under analysis (e.g., they're not systematic errors, and aren't "lucky" low-sigma correlations).
It's more than a bit provocative to suggest that the observations suggest that the EFT fails outside the horizon (one of the possibilities that has been discussed by Polchinski and others in the wake of AMPS (he's the "P")), which is fine, since the point will stand or fall on the basis of evidence rather than the consequences for various QG research programmes. :-)
For now it's pretty safe to work on the assumption that the EFT is fine at least everywhere outside the event horizon and the extremely hot dense early universe, and that a UV completion to GR will be completely compatible with the EFT outside strong gravity and inside a region at least a little bigger than the observable universe.
Finally, all this means that although I won't exactly endorse your wording in your third last sentence, I do certainly agree with the sentiment.