[Retric]

The issue isn’t the useful product of light or a functional crypto system, the issue is the amount of waste heat produced by the light bulb or data center.

This should be obvious when you consider proof of work systems that generate less waste heat by using a memory bound rather than a computationally bound algorithm. They both operate on the principle of wealth destruction as all proof of work algorithms do, but you get different externalities.

Interestingly, if you found something really useful to do with the waste heat of crypto then the systems would become roughly equivalent.

PS: The idea of a memory bound algorithm is 1 million dollars of RAM consumes less energy than 1 million dollars of ASIC’s.

[Karunamon]

the issue is the amount of waste heat produced by the light bulb or data center.

..which in the case of the average crypto, is overshadowed many times over. One block mined equates to $333K worth BTC generated alone, and that's not counting the value of every transaction included in that block. A block happens every 10 minutes give or take, so in an hour, that's about 2 million worth of coin generated and 13K transactions.

There's no way on earth every bitcoin miner, combined, is using $2M+ worth of electricity every hour, even factoring in negative environmental externalities (which in the case of crypto mining, will be lower than most people think given their proximity to renewable, hence cheap, energy.)

It is, objectively speaking, not a waste.

[Retric]

Coal power plants generate more value in the from of electricity than they lose value as waste heat. Otherwise they wouldn’t operate. But, clearly the creation of value is independent of something being wasted as you can have more or less efficient power plants that generate the same value using more or less fuel and thus more or less waste heat.

Similarly, as I just pointed out different algorithms would use more or less energy and therefore be more or less efficient. Thus the heat produced is a waste byproduct mining, and electricity is the fuel. A more efficient process simply uses less fuel though it may have higher capital costs.

The same is true of say cars, you’re burning gas because driving is useful. But, increasing efficiency means less fuel while creating identical value etc etc.

[Karunamon]

But that heat you speak of applies to any kind of long-running computation. Whether that be render farms, ML model training, distributed computing like BOINC, or anything else with similar properties.

It seems to me that crypto is being unfairly singled out. Things that are digital have meaningful existence.

[Retric]

Increasing efficiency is a huge aspect of every other field in computing with companies literally spend 10’s of billions of dollars increasing efficiency of software and hardware every year.

It’s directly offensive to many people in the field that an inefficient first implementation is maintained because people’s investments in hardware are considered more valuable than anything else. Not latency, not transactions per second, not environmental costs, just maximize the value of their hardware investments.

Completely understandable, yet still horrific.