Elon Musk recently sent Bitcoin’s value plummeting after voicing concerns about the cryptocurrency’s “insane” energy use. But what if there were a way to channel its computing power towards something more productive?
Bitcoin currently consumes as much electricity as a small country thanks to the huge amounts of processing required to verify transactions. What’s worse, the bulk of this number crunching is dedicated to churning out strings of digits with no practical use.
As Bitcoin and other cryptocurrencies grow, there’s been growing criticism of the huge amounts of electricity being diverted into these wasteful calculations. That’s prompted some researchers to ask if we could replace them with something more productive.
“Bitcoin is now consuming too much electricity without creating any real benefit for the world,” says Taeho Jung, an assistant professor at the University of Notre Dame. “So we were wondering if we can do it better, if we can reinvest this energy consumption and computational power into something more useful.”
That’s easier said than done, says Jung. Most cryptocurrencies allow any computer connected to the network to update transaction records, so to prevent fraud they force users to solve fiendishly difficult numerical puzzles before they can make changes. In exchange these users, known as miners, are rewarded with cryptocurrency.
This approach is called Proof-of-Work and involves miners taking transaction data and feeding it into a cryptographic algorithm that generates a string of numbers called a hash. Each chunk of data can spawn more possible hashes than there are stars in the visible universe, but the goal is to find one with just the right numbers of zeros at the front. That means miners have to run the operation trillions of times and the processing power required makes tampering with records far costlier than any potential fraudulent gains.
The hashes are practically useless but the puzzle itself has several useful characteristics that makes it hard to replace, says Jung. Firstly, finding the correct hash is much harder than verifying it. The difficulty can also be predictably adjusted by changing the number of zeroes at the front, which helps control transaction processing speeds. And hashes are indelibly linked to transaction data so attackers can’t swap hashes from valid transactions onto fraudulent ones.
One potential alternative is an approach known as Proof-of-Stake. Rather than racing to generate hashes, users put up a chunk of their cryptocurrency holdings to gain the right to edit records. This keeps people honest while using a fraction of the processing power and the approach is gaining popularity. The number two cryptocurrency Ethereum is in the process of switching to this system. But Jung says there are still question marks over whether it can provide the same level of security as Proof-of-Work.
The other option is to find a more useful replacement for the hashing puzzle. PrimeCoin, which launched in 2013, made an early attempt at this by getting miners to produce chains of prime numbers of use to scientists and mathematicians. The approach ticked all the boxes, but quickly exhausted potentially useful prime numbers, pointing to another challenge. “The question is not whether we can find useful work, but whether we can find useful work that has impact,” says Jung.
To that end, he and colleagues devised a system called DLChain that uses miners’ processing power to train deep learning models. Much like the puzzles at the heart of bitcoin, training takes much longer than verifying the model’s accuracy. Their approach also ties the work to the transaction data by using it to generate random numbers used in training. And difficulty can be tuned by setting model accuracy milestones that miners have to hit to get rewards.
The approach substantially increases the amount of processing, but Jung points out it goes into something productive that would be done anyway. Nonetheless, he’s pessimistic about it replacing Proof-of-Work anytime soon. “The majority of people who invest in Bitcoin are investing without understanding how Bitcoin works,” he says. “Whether it's producing a hash or something useful, they don't care.”
Another problem, says Emin Gün Sirer, co-director of the Initiative for Cryptocurrencies and Smart Contracts at Cornell University, is that if you tie mining to a useful task with a known economic value it creates a price anchor. That will prevent the spectacular leaps in value currencies like Bitcoin have seen, which is likely to put off investors. “These people want to go to the moon,” says Sirer. “And if you have a price anchor then you can't really move.” You’re also at the mercy of dips in demand for your useful work, he adds, which could bring your currency grinding to a halt.
One way round these issues is to make sure you’re not tied to a particular kind of work. Ittay Eyal, an assistant professor at Technion in Israel, and colleagues have created an approach called REM that lets miners use almost any computer program to generate proof-of-work by running it in a “trusted execution environment”—a secure area of a processor.
“You do protein folding or simulations or machine learning, it doesn't matter,” says Eyal. “We don't need to pre-define what work is useful.” The group’s software translates programs into a form that runs on a trusted environment in Intel chips while keeping track of the number of instructions executed. This makes it impossible to fake how much computational work is done.
The catch is that you have to put your trust in the chip manufacturer. While a company like Intel is unlikely to abuse that, Eyal admits its would be a hard sell for purists who value the decentralized and trustless nature of most cryptocurrencies. More pertinently, trusted environment technology is still immature and researchers have found vulnerabilities in current versions.
“It will take some time until trusted execution environments are secure enough to secure large amounts of money in this manner,” says Eyal.