I'm very late to this discussion. (I'll claim a delayed choice as the information took 4 years to make it to me even at subluminal speeds.)
To answer the original question, the answer is no. But then the video it comes from also says the answer is no, as does Ron Garret himself in this discussion. (I'm not sure what Emilio is talking about, given that the video itself does knock down the strawman and shows why the approach doesn't work, so it does stand on its own. To be fair, at the point in the video it is presented, Ron states it as if it is true, which is the strawman, but then later describes why what he said at that point isn't true. Now he doesn't specifically say that it re-validates Copenhagen as a possible interpretation, but that "invalidation" of CI came from the strawman so when Ron knocks down his own strawman he effectively knocks down the invidation of CI. But, then also talks at the end about potentially viable interpretations as his "zero world", and the many worlds, and doesn't seem to note that CI is still formally viable. At best he's ambiguous about whether he's suggesting it is viable or not, but the video taken as a whole with the math seems to accept that CI is viable. You just have to be careful to note which statements said earlier are tentative based on the strawman vs those left standing once the strawman falls.
But, I have an expansion of the question based on the answers here, using the famous "delayed choice quantum eraser" experiment. While I assume most people here know it, or can look it up (e.g., Wikipedia on that exact phrase), I'll summarize what I understand:
Particles pass through standard double slit then go into a beta barium borate crystal, creating the entangled pair, which are further diverged via a Glan–Thompson prism. One pair of paths go "up" to detector D0 (let's say in Alice's lab), and the one pair of paths go down to the delayed choice mechanism (let's say in Bob's lab).
The two paths going to Alice's lab are combined such that no "which way" slit information survives at detector D0.
The two paths going to Bob's lab come in contact with beamsplitters BSa (path for slit A) and BSb (path for slit B), each which allows a 50% chance of the photon passing through or being reflected. If reflected, path A ends at detector D3. If triggered, we know the photon went through slit A. If a photon is reflected at BSb (path B), path B ends at detector D4, and we know the photo went through slit B.
If a photon of either path A or B passes through the beamsplitter (BSa or BSb), it continues on a path with mirrors Ma and Mb, respectively, followed by beamsplitter BSc where path A goes through from one side and path B goes through from the other side and both have a 50% chance of the photon passing through or reflecting, ending up at either detectors D1 or D2. That is, D1 detections could equally come from path A reflecting from BSc or from path B passing through BSc. Similarly D2 detections could be path A passing through BSc or path B reflecting off BSc. Hence D1 and D2 have no "which way" path information.
The result of the experiment notes that, when photons are detected at D3 and D4, where path information is known (D3 = slit A, D4 = slit B), the corresponding collection of detections of the entangle photons at D0 form a normal random distribution with no interference pattern, R03 and R04. When photons are detected at D1 and D2, where path information is not known, the corresponding collection of detections of the entagled photons at D0 form interference patterns, R01 and R02, where the patterns of R01 and R02 are 180 degrees out of phase. I believe R01 + R02 combines the patterns to look like a normal distribution.
My question then, which I'll propose the answer below, is why you can't use this for FTL comms.
Here are the steps I'd imagine. Move Ron's lab even further away, say 1 lightyear, to make the FTL case clear.
Now, replace BSa and BSb with 100% reflective mirrors but are movable in and out of paths A and B, respectively. When Ron puts them in the path, all photons go to D3 and D4, so Alice gets patterns R03 and R04 on her detector. When Ron takes the mirrors out of the path, all photons go to D1 and D2, so Alice gets patterns R01 and R02 on her detector.
So, in principle, Ron could modulate the positions of the mirrors such that in-path = 1 and out-of-path = 0, and Alice could read patterns R03 and R04, which are identical normal curves, as a 1, and patterns R01 and R02, which are interference patterns, as 0, and hence Bob can send a signal at any point that Alice can read based on the patterns on her D0 detector.
Why this can't work, from Ron's video and the answers above, I think, is because all Alice sees is a normal distribution on her detector all the time, either R03 + R04, which is a normal distribution, or R01 + R02 which is an identical normal distribution. R01 and R02 are separate interference patterns, but the only way to separate them on Alice's detector is to know which individual detections hit D1 or D2 in Bob's lab. And, the only way to know that is to send the D1 or D2 detection notification from Bob's lab to Alice's lab using conventional subluminal speeds. (This is the "coincidence counter" in the actual experiment which matches up D0 detections with D1, D2, D3, and D4.)
If I follow Ron's video, e.g., the slide at ~41:41, and the answers here that seem to reinforce the same thing, the combined interference patterns are what removes the ability to separate the signals to create FTL communications.
Is that correct, or have I misinterpreted the reason you can't do this?