this primitive suggests a concrete design of a new L1 base-layer protocol, which nearly eliminates the energy-waste of Bitcoin mining – allowing GPU consumers to reduce their AI training and inference costs by “re-using" it for blockchain consensus, in exchange for block rewards (2-for-1)
Merge mining (trying to reuse proof of work for multiple chains) has been around for a long time. I think this paper is claiming to have found a way to base proof of work on solving the same kind of problem that most GPUs are working on when being used for LLM purposes, rather than random hashing (which is what happens in Bitcoin currently).
From the delving post about it:
An immediate objection is that there’s no inherent waste.Wastefulness (in CAPEX or OPEX) isn’t as foundational to secure & decentralized PoW consensus as non-repurposeability. Non-repurposeability means attacking the coin would reduce the coin’s value so much that the future income from the majority’s CAPEX investment is harmed more than an attack would be a reward. The chain must be the miners’ primary source of future potential income. If something else paid more, they wouldn’t mine as much and it wouldn’t be as secure because those external sha256d operations could periodically visit Bitcoin to double-spend.The innovation in this PoUW that bypasses the need for “imprisoned” (dedicated) work is that it’s like merge mining. It enables miners to get their primary profit from other sources. It’s secure if the exchange value of double-spending is less than the cost of the work provided by the portion of the world’s matrix multiplication that’s “merge mining” the coin.There has to be enough “merge mining” to raise the cost of solving the difficulty above the value of an attack.
There's a good post on X about the paper as well.
I feel like I've seen schemes like these before and largely they don't work. This is possibly just because Bitcoin has such strong memetic defenses that things like the 21 million cap and Proof-of-Work will simply never changed.
Curious, though: do you think it is important that the work expended to mine bitcoin is not used for any other purpose (is wasted beyond trying to produce a new block)?
The authors of the paper note that
This blockchain is currently under construction