Mining is too centralized, and many of the major miners are clearly bad actors. But a PoW change looks like a net negative currently. Calling for a PoW change now is kind of like calling for a block-size increase: you see a problem, and you see a big red button labeled "solve problem", but you fail to look at what the button actually does.
First of all, it's important to recognize that Bitcoin is not ruled by miners. If it was, then Bitcoin would already be lost, since the majority of miners have long been incompetent and/or totally opposed to the ideals of Bitcoin. If miners try any attack, a response can be custom-tailored to it by the economy in order to stop the attack and undo any damage. In case of attack, it would absolutely not be necessary to accept whatever the miners are doing.
Secondly, the major miners (especially Bitmain and associates) are major miners because they have access to high-quality chip fabrication technology/infrastructure. If you changed the PoW to eg. SHA-3, this would be a massive boon to Bitmain, since they would be in a better position than anyone on Earth to create new hardware.
Thirdly, pure PoS has been totally discredited. A combination PoW+PoS might not be a terrible idea IMO, but the idea is fairly unpopular among experts for technical and philosophical reasons.
Finally, it's easy to say that you want to create an "ASIC-proof" PoW function, but doing so is much more difficult. Recall that "democratizing mining" was one of the goals of Litecoin's usage of scrypt, but Bitmain is now producing Litecoin ASICs. X11 was another attempt at an ASIC-proof PoW, and ASICs were produced for that one as soon as the altcoins which used it became valuable enough to make doing so profitable. You can think up whatever ultra-complex, apparently-CPU-unique thing you want, but:
Probably you will fail to create something that actually requires a true CPU rather than some more-efficient-than-CPUs ASIC, as doing so is very difficult. (To do so would require attacking the problem rigorously & mathematically, not just willy-nilly doing things that look hard for ASICs.)
If you somehow do create one, congrats: you've now given a monopoly on Bitcoin mining hardware to Intel and (to a much lesser extent) AMD, the only two companies with the technology for creating efficient x86 CPUs.There's been some research on creating a memory-hard PoW function rather than a CPU-hard function, but this is an experimental area. The best candidate function we have, cuckoo, is not proven to be memory-hard (ie. maybe more analysis could someday allow for eliminating the memory-hardness). Furthermore, even if we had a well-tested and proven memory-hard PoW, it is not clear that this would result in less mining centralization. Maybe the economies of scale would end up even more pronounced here, or maybe mining would be taken over by a couple of giant botnets. (I actually think that this is the most promising direction for solving mining centralization long-term, but there are serious questions and challenges which probably won't be well-addressed anytime soon.)
So my position has been and continues to be that:
The PoW should absolutely be changed if miners do any sort of attack (the most likely of which is censoring "dubious" transactions rather than anything more obviously evil). In that scenario, the miners must be fired or Bitcoin is dead, so a high-cost, risky PoW change is worthwhile. And if the first PoW change fails to keep those miners away,
you keep changing the PoW until they've created so many ASIC paperweights that they've run out of money.While I would like to replace the current miners, in the absence of any attack, a PoW change is a high-cost, high-risk move which is unlikely to actually improve things in the long-run anyway. It's not worth it.
https://www.reddit.com/r/Bitcoin/comments/7zvit8/cobra_miners_are_evil_we_need_to_get_rid_of_them/dus9tuq/
