You're confused about who has what burdens here. The only fact established in our scenario is that an AI computer has stopped functioning in the manner in which it is intended to function--we know nothing else for sure about the nature of the failure. The two basic possibilities are (A) the failure has compromised only the AI function, or (B) the failure has compromised more than just the AI function...you're assuming that because generally failures only compromise the AI function, that is always true. That may be a fairly safe assumption given how infrequently your experience shows failures of AI to mean failures of anything else like deco calculations, but it's still an assumption one need not make if there's no AI functionality to fail in the first place.
I have tried to go beyond assumptions in considering this, as it is important to me and a lot of other people. However, several posters have said that all I have is an "absence" of evidence, but I don't think that is exactly right.
1. It is a fact that tens of thousands of AI computers are in use and have been in use for many years, and (an assumption here?) hundreds of thousands of dives. Even if the AI is not used, the extra software exists in all of the computers, for all of these dives.
2. There is no documented instance or evidence that AI computers, as deco computers, fail more often or in different ways, than computers generally, merely due to the software.
3. There is no documented instance of a mechanical failure--the loss of AI signal connection between transmitter and computer-- that triggers deco algorithm failure.
4. There is no documented instance that this mechanical failure also causes a computer crash or lock-out. Victor pointed to a failure of a DC-03 that occurred simultaneously with AI failure, but even he cannot say that there was another, unrelated to AI reason that caused all computer functions to fail. Several DG03s with a particular firmware failed, this is widely reported, but no reports ever indicated that is was related to the AI software or hardware, and it was corrected with a firmware update. Perhaps Hollis can tell us more, and explain how the firmware fixed this crash problem, and whether the new firmware had to correct the AI to fix the crash problem.
Of course, I can't know everything that has happened, everywhere, all the time. But, I have looked at this on the net, in reports of lawsuits, in magazines, and just have not found any examples.
Now, the lack of any confirmed instance of AI causing deco algorithm failure over hundreds of thousands of dives, across multiple brands of computers, over many years, is far more than "absence" of evidence, it is, actually, positive evidence indicating that the theoretical concern about AI leading to deco unreliability in computers is simply not based in experience or observable phenomena.
Also, no one addressed the plain fact that Shearwater has found a technology that it deems reliable enough to do not just AI but AI
over very long distances, from multiple computers, all at the same time. What is the reason it could not be implemented over a few feet from first stage to wrist for a single computer?
Whether people want it, whether Shearwater wants to make it, those are different things.
I posted before an explanation of why AI (and a compass) is a tangible benefit in my personal diving, and a Shearwater with AI would cause me to dump my other computers immediately. I have to admit I am thinking about it even now (that color display . . .) but am not in a hurry because when I switch I would like it to be my final and last computer that has the features valuable to me.
I am not alone, so I just encourage Shearwater to think about it . . . . that, really, was my only intent in posting. I do not want to change tech divers' minds about how they dive, only to raise the point that AI, if they do not use it, should not cause them to trust their computers any less, based on what is known thus far over a pretty big data set.