I’m curious what your impression is after reading the above, and whether there is other specific evidence that I should be paying more attention to
Sure.
because it involved major technical and conceptual/philosophical advances on the existing state of the art,
I agree Bitcoin is relatively audacious, novel, and technically sophisticated .
and these advances didn’t originate from nor was likely funded/supported by academia, government or industry
Couldn’t you describe many early-stage self-funded startups that way? Or do you mean you guess that Satoshi was not working in academia, government, or industry before developing Bitcoin?
Also, its social impact seems larger—if Craigslist or PayPal didn’t exist, something essentially identical would have been created very soon anyway, but if Bitcoin didn’t exist, another Bitcoin may not have been created for another decade, and/or may have been created with very different characteristics
What kind of impact? So far the volume of Bitcoin transactions is still relatively small, and presumably a large portion of the transactions conducted using Bitcoin would otherwise be undertaken using other payment systems. So getting a version of PayPal or Craigslist working somewhat earlier or better could easily affect more transactions and generate more consumer surplus if Bitcoin does not grow to larger scales.
ETA:
PayPal’s net Total Payment Volume for 2012, the total value of transactions, was $145 billion, up 22% year over year.
Couldn’t you describe many early-stage self-funded startups that way?
Self-funded startups that also involve major technical and conceptual advances that weren’t first developed in academia, government, or industry (and then spun off) seem rare. Can you give some examples that are similar to Bitcoin in this regard?
Or do you mean you guess that Satoshi was not working in academia, government, or industry before developing Bitcoin?
If Satoshi was working in academia, government, or industry, it seems very likely that he didn’t develop the ideas behind Bitcoin as part of his day job, otherwise he probably wouldn’t have been allowed to publish the ideas and software under a pseudonym.
So getting a version of PayPal or Craigslist working somewhat earlier or better could easily affect more transactions and generate more consumer surplus if Bitcoin does not grow to larger scales.
It’s hard to even say whether Bitcoin ultimately has a positive or negative impact at this point. For example one possible impact of Bitcoin might be that due to its deficient monetary policy and associated price volatility it can’t grow to very large scales, and by taking over the cryptocurrency niche, it has precluded a future where a cryptocurrency does grow to very large scales. If we take the expectation of the absolute value of its impact, it seems higher to me than the impact of a somewhat earlier or better PayPal or Craigslist.
due to its deficient monetary policy and associated price volatility it can’t grow to very large scales, and by taking over the cryptocurrency niche
I’m also quite worried about this, but on the other hand Bitcoin creates an obvious entry gateway into more advanced cryptographic currencies (i.e. once Bitcoin infrastructure is set up, other currencies can use Bitcoin infrastructure if there’s a way to exchange them with Bitcoins, lowering the bar to entry).
I’ve had all sorts of ideas along these lines, in fact. The main reason I haven’t published them is that I’m not sure that more advanced cryptocurrency advances FAI over AGI. In fact, you’d think it would be the reverse—the Great Stagnation may be all that’s keeping us alive right now.
I’m surprised that you are so interested in this area (i.e., monetary policy for cryptocurrency), given that the subject matter and required backgrounds to study it are not closely related to FAI. I don’t even have any strong opinions on what is the right policy, except that the one currently built into Bitcoin is pretty suboptimal (ETA: at least in the long run, in the short run it seems close to optimal for getting Bitcoin some initial scale).
The main reason I haven’t published them is that I’m not sure that more advanced cryptocurrency advances FAI over AGI.
Yeah, me either, or more generally whether cypherpunk-related technologies help or hinder a positive Singularity, which is part of the reason why I stopped pushing very hard on my cypherpunk ideas.
Econ relates to intelligence explosion dynamics, Scott Sumner appears to be a Correct Contrarian.
I don’t have any good ideas for how to do NGDP level targeting inside a cryptocurrency in a way that would automatically distinguish more widespread adoption from increased RGDP from somebody gaming the system.
Econ relates to intelligence explosion dynamics, Scott Sumner appears to be a Correct Contrarian.
No, he doesn’t. Edit: and I’ve found that his general reasoning to be poor in general. Some examples (which I can source later if anyone plans to update on this):
“Sumner, if what you’re saying is true, shouldn’t the Fed let anyone, including the average Joe, borrow from the Fed at 0%?” -- > “Yes.”
Sumner: “Income” is a meaningless concept. Critic: No, it’s obviously vital to know how much you can spend before becoming unable to buy anything. And how would you value an enterprise but by discounting its income streams? Sumner: If you want to know what a venture is worth, look at its stock price, not income.
I don’t have any good ideas for how to do NGDP level targeting inside a cryptocurrency in a way that would automatically distinguish more widespread adoption from increased RGDP from somebody gaming the system.
Non-crypto, “real world” currencies have exactly the same problem of gaming the numbers to make NGDP artificially high. They’re only worthwhile to pursue when NGDP specifically is targeted, rather than some “close enough” policy.
Right now, the best velocity measure seems to be coin days destroyed. But it is gameable. It is not being gamed in bitcoin because nothing is dependent on it.
The closest GDP measure in a cryptocurrency of the structure of bitcoin seems to be sum of transaction fees. It can be gamed by early adopters, but that is true of almost every measure
In fact, you’d think it would be the reverse—the Great Stagnation may be all that’s keeping us alive right now.
I don’t see why we should obviously expect funding for AGI to benefit from economic growth in a way that funding for FAI doesn’t. If Silicon Valley is booming, I’d expect MIRI to receive more donations and Google to put more money in to self-driving cars. If Silicon Valley is contracting, I’d expect MIRI to receive fewer donations and Google to put less money in to self-driving cars. Am I missing something here?
(i.e. once Bitcoin infrastructure is set up, other currencies can use Bitcoin infrastructure if there’s a way to exchange them with Bitcoins, lowering the bar to entry)
Yes; it’s been pointed out that you can, and people have, set up competing currencies using the codebase but without the builtin caps. More interestingly, you apparently can build currencies directly on the existing Bitcoin codebase & main blockchain using colored coins (which wedrifid seems very interested in).
No, what I mean is that if anyone else sets up a cryptocurrency right now, they don’t have to worry about making it exchangeable with dollars, they just need a good way to make it exchangeable with Bitcoins, and that could easily be done using pure programming. Bitcoins is a horrible store of value and an even worse medium of account, but some of the underlying ideas have great potential as a medium of exchange, and Bitcoin can sneeze any previous development of real-world interfaces directly into a new, competing cryptocurrency.
I’m having trouble reconciling the “horrible store of value” part with the rest of your argument. A competing crypto-currency eventually comes into existence, better than dollars, everyone wants to use it instead, and you can’t get it without bitcoins… And we’re supposed to think bitcoins are not particularly likely to go up in value as a result?
Edit: I guess if you only mean “currently horrible” this makes plenty of sense. Also it could maybe lose value once it has played its part in getting everyone used to the new currency.
A competing crypto-currency eventually comes into existence, better than dollars, everyone wants to use it instead, and you can’t get it without bitcoins… And we’re supposed to think bitcoins are not particularly likely to go up in value as a result?
If such a crypto-currency comes out, everyone holding Bitcoin will want this Newcoin, everyone who would have held Bitcoin will instead be demanding Newcoin, and the only reason anyone will be holding Bitcoin will be as part of the float generated by people trading dollars for Newcoins via Bitcoin non-instantaneously. The exchange rate supported by the float could be far less, the same, or far higher than the current exchange rate.
(An example: suppose Bitcoin somehow lost popularity so that the sole use of the current n bitcoins was for Silk Road, and no buyer or seller let more than a day elapse between exchanging their $$$ for Bitcoin (buyers) and exchanging their Bitcoin for $$$ (sellers); if there’s $1m of total turnover on Silk Road per day, then buyers need to turn $1m into Bitcoins and seller need to turn X Bitcoins into dollars, and this need will be spread out over n bitcoins. IIRC, there’s like 5m bitcoins so in this scenario we would each day see 5m bitcoins traded from the sellers to the buyers in exchange for the buyers’ $1m, or an ‘exchange rate’ of $5/btc, which is approximately 1/28th the current exchange rate, and then the buyers move the bitcoins to SR and hand them over to the sellers, who move them back to the exchange to sell to the buyers...)
Are we talking about converting Bitcoins into Newcoin (say by sending them to a fake address as a precondition for minting X many Newcoins) or are we talking about trading them (“you send me X many Bitcoins, I send you Y many Newcoins” transactions)? The former strategy would drive scarcity of Bitcoin up as a direct result of demand for Newcoin.
I thought we were talking about the latter—people would convert dollars into Newcoin via an intermediary Bitcoin stage (perhaps because Newcoins are banned or something).
If we were talking about ‘converting’, such as by some of the suggested verifiable-destruction strategies… I’m not sure. Presumably each Bitcoin would always sell for at least as many dollars as its equivalent in Newcoin would fetch modulo the transaction fees and effort (since otherwise people who want Newcoins would buy up Bitcoins and convert them immediately), but if Newcoins were so much better why would any Bitcoin holder at all not immediately convert all their Bitcoins to Newcoin? Reminds me of the flows between coin and bullion in metallic regimes.
I now see that when I wrote the original post, I probably should not have used the term “trade” as a synonym for “convert”. Someone who did not read my post closely might have thought I was enthusing about incremental improvements that let you reassign ownership like a traditional marketplace, only more efficiently, with automatic bid processing or something. That might be nifty, but it is not the earth-shattering point that makes me want to buy lots of bitcoins. What makes me want to buy more bitcoins is that reassignment of ownership is not the easiest way to do it. Instead it’s easier to make a system that destroys so many bitcoins and creates so many newcoins. It is also something that the current owners of bitcoin have significant financial incentive to make sure happens. Since it’s easier and massively incentivised, I think it carries the bulk of the probability mass.
Is Bitcoin’s monetary policy really expected to be a problem? If it were to reach a steady state adoption and usage level, I imagine Bitcoin’s price would stabilize around some constant fraction of (the present value of) the market’s expectation of future world GDP. Is that not what you would predict?
Or is the problem that you don’t think Bitcoin can reach widespread adoption without users reasonably being able to price goods in BTC?
That’s a problem because then BTC is a perfect investment which always grows at exactly the same rate as the global economy. So it gives you the exactly average return on investment with zero volatility. So it seems like a near-perfect store of value and people will want to hold it rather than spend it. This decreases velocity which causes deflation and value that increases apparently even faster than the total global economy. This makes Bitcoin apparently an even better investment, until the volatility or expected volatility from the huge stores of unused Bitcoins outweighs its apparent returns on investment, and note that financial markets are apparently unusually bad at expecting future volatility to be greater than present volatility; people try to time bubbles instead. This is bad for Bitcoin because of the inevitable crash followed by hyperinflation. And it’s bad for the global economy because your currency is deflating and any given bank would rather hold Bitcoins, on average, than make loans; and then the inevitable crash is also bad. That’s a nutshell version of a longer story.
That’s a problem because then BTC is a perfect investment which always grows at exactly the same rate as the global economy. So it gives you the exactly average return on investment with zero volatility. So it seems like a near-perfect store of value and people will want to hold it rather than spend it.
Assuming BTC gives exactly the average return on investment with zero volatility we shouldn’t expect all people to hold it rather than spend it. Neither an economy of actual humans nor an economy of ideal agents would act that way.
With respect to consumption: Use as a transactional currency for spending would track convenience factors. People buying stuff with one fungible asset is much the same as buying stuff with another asset then transferring between their two accounts. For spherical cow in a vacuum purposes we can ignore this. Investment spending is the issue here.
In the counterfactual BTC currency which perfectly tracks the global economy the incentive is for anyone who believes they know of any investment that they expect to have higher returns than the average of the global economy to spend their bitcoins and invest in that opportunity. Those who don’t believe they have any knowledge of anything that will produce better than average returns or who are risk averse will instead purchase bitcoins either directly or indirectly from those that do have that knowledge.
In that idealised scenario the BTC currency is essentially operating as a vehicle to efficiently transfer real-world capital to places those who people with value expect will provide better return in investment than the average growth of the economy. Note that I am emphatically not claiming that this is an ideal system, it would be bizarre if something so arbitrary happened to be optimal. Just that it doesn’t seem to quite have the degree of problem that is described. People would certainly want to spend it.
There are plausible reasons why predictable inflation of the above currency could be more desirable than precisely zero inflation. Let’s say Satoshi had arbitrarily decided that BTC mining should go on indefinitely, with the bitcoins produced per year exactly equalling 2% of the number of bitcoins already mined. Then the incentives to the the participants change slightly. Rather than people who expect an investment to grow at more than the average for the global economy to be the only ones to so invest, it is any (risk neutral) person who expects an investment to grow at not less than 98% of the rate of the global economy. That has (well known) advantages.
The unfortunate problem with the above monetary policy is that we just effectively dedicated 2% of the of the value stored in the bitcoin currency each year to the computation of irrelevant hashes (in addition to irrelevant computation that is proportional to transaction fees). This problem applies to any cryptocurrency based on cryptographic mining. There may not be a good solution to that problem that potentially prohibitive degree of waste that does not rely on something external to the cryptocurrency as basis. (And the latter is not necessarily a problem. The currency having value in itself isn’t the most potentially useful feature of bitcoin.)
Let me rephrase: The problem is that Bitcoins will have an advantage over the average productive investment, e.g. stocks (sort of), as a store of value, since Bitcoin has all their average expected growth with none of their added (local) volatility. This is what presents the starting problem in an economy that starts out with a steady velocity of Bitcoins, and then increased holding makes the velocity go down (and the value go up, and the bubble effect hit even harder). This is why we don’t get an equilibrium with steady Bitcoin velocities. Even if we did have that equilibrium, people would have a much greater incentive to just “invest” in Bitcoins instead of being forced to try to invest in something productive. You don’t want an economy to have a perfect non-inflating store of value which is intrinsically unproductive!
Let me rephrase: The problem is that Bitcoins will have an advantage over the average productive investment, e.g. stocks (sort of), as a store of value, since Bitcoin has all their average expected growth with none of their added volatility.
I like the rephrasing. To expand on what seems to be a generalisation of this problem: Any cryptocurrency sibling of bitcoin that relies on cryptographic mining as a basis will either have this problem or will result in (value of currency * inflation rate) additional resources wasted on computation each year.
I believe (tentatively) that the above is an unavoidable result of the cryptographic and micro-economic principles that such currencies rely on.
Note, I wrote this in reply to the original version of the grandparent, which is as quoted in the parent. This is confusing since it is a bug/feature of the lesswrong system that Eliezer’s edits to his own comments do not get marked with an asterisk like others.
I do not endorse the current version of the grandparent, in as much as it overstates the position and seems to verge on encouraging magical thinking about how a currency can extract value from a system.
To expand on what seems to be a generalisation of this problem: Any cryptocurrency sibling of bitcoin that relies on cryptographic mining as a basis will either have this problem or will result in (value of currency * inflation rate) additional resources wasted on computation each year.
I believe (tentatively) that the above is an unavoidable result of the cryptographic and micro-economic principles that such currencies rely on.
This is not limited to cryptocurrencies, e.g., gold-based currencies cause people to “waste resources” mining.
This is not limited to cryptocurrencies, e.g., gold-based currencies cause people to “waste resources” mining.
Yes, the ‘mining’ metaphor was well chosen.
In terms of that gold analogy, what we are talking about in the context would be if gold spontaneously generated itself in proportion to the amount of existing gold and automatically buried itself at whatever depth makes it barely worthwhile to dig up. That waste is the unavoidable cost of making bitcoin-style cryptocurrency have ongoing inflation.
Okay, I think I understand the argument that Bitcoin will likely be permanently volatile, because growth at exactly the rate of the global economy is not a stable equilibrium for the reasons you describe (esp. ‘people like to time bubbles’).
Thinking about this a bit more though, it seems like the same argument would apply to any asset we might otherwise expect to grow in sync with global wealth. In particular, the apparently-perfect-store-of-wealth-attracting-investment-and-appearing-to-be-an-even-better-store-of-wealth phenomenon seems like a straightforward explanation of what’s been happening with the price of gold in the last decade.
But also, it seems like this argument could even apply to the stock market as a whole—would we expect a global ETF (like Vanguard’s VT) to grow at the rate of the world economy? Is that stable?
So I’m curious, do you agree that the no-stable-equilibrium argument applies to the price of these other assets as well, and if so, does the existence of Bitcoin still seem like it would be a problem for the global economy?
It partially explains the price of gold, yes. Gold’s situation isn’t really the same for three reasons: First, gold can be mined if the price goes too high, and higher prices would imply larger amounts of recoverable gold. Second, a lot of the gold on the market is paper gold, theoretical gold that two parties are trading rather than sending large gold bars around, which also adds to the supply. But most of all, unlike the supposed use-case of Bitcoin, gold is not being used as a medium of account or medium of exchange any more, just one store of value among many, so its real competition is not paper gold or mined gold but other stores of value such as platinum, silver, real estate, and many other things being added to the competition for ‘stores of value’ as the economy grows. If the same fraction of the population tried to store the same fraction of their net assets in gold today as in the 1600s then the price of gold would be vastly higher—or so I would think, I haven’t run the numbers. But this in turn means that the share of the economy represented by gold can easily drop further, making it less than a perfect store of value etcetera, although gold has still tended to be a better store of value than fiat currency.
Of course fiat currency is really supposed to be a medium of exchange and account, not a long-term store of value, though dumb people like me tend to use it as a store of value too because it’s convenient and we haven’t gotten around to setting up anything different and we don’t have that much value to store. And then using your medium of exchange and account as a store of value causes recessions and depressions due to the paradox of thrift; when people want to consume in the future instead of the present they try to hold paper money instead of demanding equity in projects with long-term payoffs. On the plus side, central banks can, in principle, easily rectify some part of this problem by printing more money to meet demand for currency when fear rises, and thus make up for velocity slowdowns, keeping NGDP on a level growth path. On the minus side, central banks are stuck in 30-year-old economic thinking and don’t keep NGDP on a level growth path. Bitcoin has the potential to make things much, much worse though.
New stocks on the other hand are constantly being created as the economy grows—no particular stock, or set of stocks starting at a fixed time, are guaranteed to grow at the same rate as the global economy.
To paraphrase, you’re pointing out that stocks and precious metals come with built-in demand shock absorbers, whereas Bitcoin has none. I’m not totally sure that I accept this point, because I could see alternative cryptocurrencies playing the role of marginal new stocks or newly mined gold. However, even if Bitcoin were unique in having no demand shock absorbers, I’m not sure this matters, because it seems empirically to be the case that these shock absorbers are not always up to the task, and that both stocks and precious metals do experience a great deal of price volatility, even over the medium to long term.
In other words, even if Bitcoin is especially sensitive to changes in demand, it is neither novel nor unique in being susceptible to bubbles.
This would seem to me to imply that Bitcoin’s existence and use as a store of value is no threat to the economy. (And its use as medium of exchange seems harmless as well.)
It would seem that problems would only arise for those who try to use Bitcoin as a unit of account. This is in line with Wei’s comment where he suggests that with a currency in fixed supply, fluctuating velocity of money implies that either prices or GDP must be unstable.
So my conclusion is that using Bitcoin as a medium of exchange or store of value is not detrimental to the economy, but one should continue to price goods or services in some other fiat, ideally NGDP-targeted, currency. Does that sound about right?
Anyone bidding on a Bitcoin is not bidding on a productive project.
It seems that the same goes for gold, real estate, and so forth when they are used as a store of value. The difference is that unlike bitcoin, these things have other productive uses that they could be put to, less expensively, if they weren’t being used as a wealth-counting mechanism.
If it were to reach a steady state adoption and usage level, I imagine Bitcoin’s price would stabilize around some constant fraction of (the present value of) the market’s expectation of future world GDP.
According to Wikipedia, “velocity of money” is not a constant but tends to fluctuate. The article lacks citations, but I think this is the current mainstream view among monetary theorists. My understanding of the implication of this is that if you have a fixed supply of money then either prices or GDP would have to be unstable. In other words either you end up with an economy with very flexible prices that change constantly, or you end up with an economy that goes through constant boom and bust cycles (or some combination of each), and both of these outcomes are costly.
I feel like this isn’t nearly the issue it is made out to be when you separate real growth from nominal growth. Say you have real growth but prices fluctuate a lot, why should you care? Frictional costs should decrease over time as people figure out how to hedge properly in this environment.
So is the solution just to use Bitcoin as a medium of exchange and a store of value, but not as a unit of account? Then prices are free to fluctuate in BTC terms, while they can remain relatively stable in fiat terms, and GDP will be unaffected.
That would make it a terrible at being a medium of exchange or a store of value, though, wouldn’t it? No one knows how much it’s worth, and you have to acquire some, pass it off, and then (on their side) turn it into currency every time you use it.
That depends on how volatile it is. On the timescale of a single transaction, a certain level of volatility might not matter very much even if the same level of volatility would prevent you from wanting to set prices in BTC.
I wonder to what degree FAI/CEV engineering considerations overlap with cryptocurrency/efficient-market engineering considerations. If they are a close match, encouraging the development of the latter would have benefits for the former, and could even be essential to overcoming scaling problems (since FAI is harder to sell investors on than cryptocurrency).
This isn’t just a random idea; markets are how humans in the absence of superintelligence actually do try (with some, not-unlimited success) to implement their values. Prediction markets are a possible extension of this concept, but even your run-of-the-mill securities markets are reliant on various kinds of predictive logic that responds somewhat to human desires and needs.
Not trying to trivialize the AGI field since it is outside my specialization, but is there some not-terribly-unlikely way in which a really good cryptocurrency could basically be/evolve into the same thing as an FAI? If so, are there any particular properties that would be likely to nudge it in that direction / away from uFAI?
I’m kind of concerned because I see bitcoin (and/or anything sufficiently similar) funding competition to purchase obscenely large amounts of hardware—which could possibly even extend to the point of satellite arrays that harvest solar energy, and space based fabrication of new ones. If it gets to that point, we might end up with a Dyson sphere that basically does nothing but compute bitcoin hashes. Extraordinarily wasteful, but not necessarily catastrophic for existing humans if the network continues to recognize them as owners/controllers of the resources in question.
If it gets to that point, we might end up with a Dyson sphere that basically does nothing but compute bitcoin hashes.
This seems unlikely. If you’re going to invest that much capital, why waste it on bitcoin hashes when you could instead provide a product and sell it? This would be analogous to worrying that everyone will go into the financial sector because it pays so well, and we’ll have no one left producing actual goods.
Hmm. I think you’re probably right, now that I think about it. The maximum size of the bitcoin reward falls towards transaction fees, which are themselves a small fraction of any given transaction. So there should tend to be significant money out there to reward manufacture of other kinds of space based goods more highly than bithashes.
If Satoshi was working in academia, government, or industry, it seems very likely that he didn’t develop the ideas behind Bitcoin as part of his day job, otherwise he probably wouldn’t have been allowed to publish the ideas and software under a pseudonym.
Sure.
I agree Bitcoin is relatively audacious, novel, and technically sophisticated .
Couldn’t you describe many early-stage self-funded startups that way? Or do you mean you guess that Satoshi was not working in academia, government, or industry before developing Bitcoin?
What kind of impact? So far the volume of Bitcoin transactions is still relatively small, and presumably a large portion of the transactions conducted using Bitcoin would otherwise be undertaken using other payment systems. So getting a version of PayPal or Craigslist working somewhat earlier or better could easily affect more transactions and generate more consumer surplus if Bitcoin does not grow to larger scales.
ETA:
Cf Bitcoin transaction volumes.
Self-funded startups that also involve major technical and conceptual advances that weren’t first developed in academia, government, or industry (and then spun off) seem rare. Can you give some examples that are similar to Bitcoin in this regard?
If Satoshi was working in academia, government, or industry, it seems very likely that he didn’t develop the ideas behind Bitcoin as part of his day job, otherwise he probably wouldn’t have been allowed to publish the ideas and software under a pseudonym.
It’s hard to even say whether Bitcoin ultimately has a positive or negative impact at this point. For example one possible impact of Bitcoin might be that due to its deficient monetary policy and associated price volatility it can’t grow to very large scales, and by taking over the cryptocurrency niche, it has precluded a future where a cryptocurrency does grow to very large scales. If we take the expectation of the absolute value of its impact, it seems higher to me than the impact of a somewhat earlier or better PayPal or Craigslist.
I’m also quite worried about this, but on the other hand Bitcoin creates an obvious entry gateway into more advanced cryptographic currencies (i.e. once Bitcoin infrastructure is set up, other currencies can use Bitcoin infrastructure if there’s a way to exchange them with Bitcoins, lowering the bar to entry).
I’ve had all sorts of ideas along these lines, in fact. The main reason I haven’t published them is that I’m not sure that more advanced cryptocurrency advances FAI over AGI. In fact, you’d think it would be the reverse—the Great Stagnation may be all that’s keeping us alive right now.
I’m surprised that you are so interested in this area (i.e., monetary policy for cryptocurrency), given that the subject matter and required backgrounds to study it are not closely related to FAI. I don’t even have any strong opinions on what is the right policy, except that the one currently built into Bitcoin is pretty suboptimal (ETA: at least in the long run, in the short run it seems close to optimal for getting Bitcoin some initial scale).
Yeah, me either, or more generally whether cypherpunk-related technologies help or hinder a positive Singularity, which is part of the reason why I stopped pushing very hard on my cypherpunk ideas.
Econ relates to intelligence explosion dynamics, Scott Sumner appears to be a Correct Contrarian.
I don’t have any good ideas for how to do NGDP level targeting inside a cryptocurrency in a way that would automatically distinguish more widespread adoption from increased RGDP from somebody gaming the system.
No, he doesn’t. Edit: and I’ve found that his general reasoning to be poor in general. Some examples (which I can source later if anyone plans to update on this):
“Sumner, if what you’re saying is true, shouldn’t the Fed let anyone, including the average Joe, borrow from the Fed at 0%?” -- > “Yes.”
Sumner: “Income” is a meaningless concept.
Critic: No, it’s obviously vital to know how much you can spend before becoming unable to buy anything. And how would you value an enterprise but by discounting its income streams?
Sumner: If you want to know what a venture is worth, look at its stock price, not income.
Non-crypto, “real world” currencies have exactly the same problem of gaming the numbers to make NGDP artificially high. They’re only worthwhile to pursue when NGDP specifically is targeted, rather than some “close enough” policy.
Right now, the best velocity measure seems to be coin days destroyed. But it is gameable. It is not being gamed in bitcoin because nothing is dependent on it.
The closest GDP measure in a cryptocurrency of the structure of bitcoin seems to be sum of transaction fees. It can be gamed by early adopters, but that is true of almost every measure
I don’t see why we should obviously expect funding for AGI to benefit from economic growth in a way that funding for FAI doesn’t. If Silicon Valley is booming, I’d expect MIRI to receive more donations and Google to put more money in to self-driving cars. If Silicon Valley is contracting, I’d expect MIRI to receive fewer donations and Google to put less money in to self-driving cars. Am I missing something here?
Yes; it’s been pointed out that you can, and people have, set up competing currencies using the codebase but without the builtin caps. More interestingly, you apparently can build currencies directly on the existing Bitcoin codebase & main blockchain using colored coins (which wedrifid seems very interested in).
No, what I mean is that if anyone else sets up a cryptocurrency right now, they don’t have to worry about making it exchangeable with dollars, they just need a good way to make it exchangeable with Bitcoins, and that could easily be done using pure programming. Bitcoins is a horrible store of value and an even worse medium of account, but some of the underlying ideas have great potential as a medium of exchange, and Bitcoin can sneeze any previous development of real-world interfaces directly into a new, competing cryptocurrency.
I’m having trouble reconciling the “horrible store of value” part with the rest of your argument. A competing crypto-currency eventually comes into existence, better than dollars, everyone wants to use it instead, and you can’t get it without bitcoins… And we’re supposed to think bitcoins are not particularly likely to go up in value as a result?
Edit: I guess if you only mean “currently horrible” this makes plenty of sense. Also it could maybe lose value once it has played its part in getting everyone used to the new currency.
If such a crypto-currency comes out, everyone holding Bitcoin will want this Newcoin, everyone who would have held Bitcoin will instead be demanding Newcoin, and the only reason anyone will be holding Bitcoin will be as part of the float generated by people trading dollars for Newcoins via Bitcoin non-instantaneously. The exchange rate supported by the float could be far less, the same, or far higher than the current exchange rate.
(An example: suppose Bitcoin somehow lost popularity so that the sole use of the current n bitcoins was for Silk Road, and no buyer or seller let more than a day elapse between exchanging their $$$ for Bitcoin (buyers) and exchanging their Bitcoin for $$$ (sellers); if there’s $1m of total turnover on Silk Road per day, then buyers need to turn $1m into Bitcoins and seller need to turn X Bitcoins into dollars, and this need will be spread out over n bitcoins. IIRC, there’s like 5m bitcoins so in this scenario we would each day see 5m bitcoins traded from the sellers to the buyers in exchange for the buyers’ $1m, or an ‘exchange rate’ of $5/btc, which is approximately 1/28th the current exchange rate, and then the buyers move the bitcoins to SR and hand them over to the sellers, who move them back to the exchange to sell to the buyers...)
Are we talking about converting Bitcoins into Newcoin (say by sending them to a fake address as a precondition for minting X many Newcoins) or are we talking about trading them (“you send me X many Bitcoins, I send you Y many Newcoins” transactions)? The former strategy would drive scarcity of Bitcoin up as a direct result of demand for Newcoin.
I thought we were talking about the latter—people would convert dollars into Newcoin via an intermediary Bitcoin stage (perhaps because Newcoins are banned or something).
If we were talking about ‘converting’, such as by some of the suggested verifiable-destruction strategies… I’m not sure. Presumably each Bitcoin would always sell for at least as many dollars as its equivalent in Newcoin would fetch modulo the transaction fees and effort (since otherwise people who want Newcoins would buy up Bitcoins and convert them immediately), but if Newcoins were so much better why would any Bitcoin holder at all not immediately convert all their Bitcoins to Newcoin? Reminds me of the flows between coin and bullion in metallic regimes.
I now see that when I wrote the original post, I probably should not have used the term “trade” as a synonym for “convert”. Someone who did not read my post closely might have thought I was enthusing about incremental improvements that let you reassign ownership like a traditional marketplace, only more efficiently, with automatic bid processing or something. That might be nifty, but it is not the earth-shattering point that makes me want to buy lots of bitcoins. What makes me want to buy more bitcoins is that reassignment of ownership is not the easiest way to do it. Instead it’s easier to make a system that destroys so many bitcoins and creates so many newcoins. It is also something that the current owners of bitcoin have significant financial incentive to make sure happens. Since it’s easier and massively incentivised, I think it carries the bulk of the probability mass.
Is Bitcoin’s monetary policy really expected to be a problem? If it were to reach a steady state adoption and usage level, I imagine Bitcoin’s price would stabilize around some constant fraction of (the present value of) the market’s expectation of future world GDP. Is that not what you would predict?
Or is the problem that you don’t think Bitcoin can reach widespread adoption without users reasonably being able to price goods in BTC?
That’s a problem because then BTC is a perfect investment which always grows at exactly the same rate as the global economy. So it gives you the exactly average return on investment with zero volatility. So it seems like a near-perfect store of value and people will want to hold it rather than spend it. This decreases velocity which causes deflation and value that increases apparently even faster than the total global economy. This makes Bitcoin apparently an even better investment, until the volatility or expected volatility from the huge stores of unused Bitcoins outweighs its apparent returns on investment, and note that financial markets are apparently unusually bad at expecting future volatility to be greater than present volatility; people try to time bubbles instead. This is bad for Bitcoin because of the inevitable crash followed by hyperinflation. And it’s bad for the global economy because your currency is deflating and any given bank would rather hold Bitcoins, on average, than make loans; and then the inevitable crash is also bad. That’s a nutshell version of a longer story.
“No one wants bitcoins anymore, they’re too valuable.”
Assuming BTC gives exactly the average return on investment with zero volatility we shouldn’t expect all people to hold it rather than spend it. Neither an economy of actual humans nor an economy of ideal agents would act that way.
With respect to consumption: Use as a transactional currency for spending would track convenience factors. People buying stuff with one fungible asset is much the same as buying stuff with another asset then transferring between their two accounts. For spherical cow in a vacuum purposes we can ignore this. Investment spending is the issue here.
In the counterfactual BTC currency which perfectly tracks the global economy the incentive is for anyone who believes they know of any investment that they expect to have higher returns than the average of the global economy to spend their bitcoins and invest in that opportunity. Those who don’t believe they have any knowledge of anything that will produce better than average returns or who are risk averse will instead purchase bitcoins either directly or indirectly from those that do have that knowledge.
In that idealised scenario the BTC currency is essentially operating as a vehicle to efficiently transfer real-world capital to places those who people with value expect will provide better return in investment than the average growth of the economy. Note that I am emphatically not claiming that this is an ideal system, it would be bizarre if something so arbitrary happened to be optimal. Just that it doesn’t seem to quite have the degree of problem that is described. People would certainly want to spend it.
There are plausible reasons why predictable inflation of the above currency could be more desirable than precisely zero inflation. Let’s say Satoshi had arbitrarily decided that BTC mining should go on indefinitely, with the bitcoins produced per year exactly equalling 2% of the number of bitcoins already mined. Then the incentives to the the participants change slightly. Rather than people who expect an investment to grow at more than the average for the global economy to be the only ones to so invest, it is any (risk neutral) person who expects an investment to grow at not less than 98% of the rate of the global economy. That has (well known) advantages.
The unfortunate problem with the above monetary policy is that we just effectively dedicated 2% of the of the value stored in the bitcoin currency each year to the computation of irrelevant hashes (in addition to irrelevant computation that is proportional to transaction fees). This problem applies to any cryptocurrency based on cryptographic mining. There may not be a good solution to that problem that potentially prohibitive degree of waste that does not rely on something external to the cryptocurrency as basis. (And the latter is not necessarily a problem. The currency having value in itself isn’t the most potentially useful feature of bitcoin.)
Let me rephrase: The problem is that Bitcoins will have an advantage over the average productive investment, e.g. stocks (sort of), as a store of value, since Bitcoin has all their average expected growth with none of their added (local) volatility. This is what presents the starting problem in an economy that starts out with a steady velocity of Bitcoins, and then increased holding makes the velocity go down (and the value go up, and the bubble effect hit even harder). This is why we don’t get an equilibrium with steady Bitcoin velocities. Even if we did have that equilibrium, people would have a much greater incentive to just “invest” in Bitcoins instead of being forced to try to invest in something productive. You don’t want an economy to have a perfect non-inflating store of value which is intrinsically unproductive!
I like the rephrasing. To expand on what seems to be a generalisation of this problem: Any cryptocurrency sibling of bitcoin that relies on cryptographic mining as a basis will either have this problem or will result in (value of currency * inflation rate) additional resources wasted on computation each year.
I believe (tentatively) that the above is an unavoidable result of the cryptographic and micro-economic principles that such currencies rely on.
Note, I wrote this in reply to the original version of the grandparent, which is as quoted in the parent. This is confusing since it is a bug/feature of the lesswrong system that Eliezer’s edits to his own comments do not get marked with an asterisk like others.
I do not endorse the current version of the grandparent, in as much as it overstates the position and seems to verge on encouraging magical thinking about how a currency can extract value from a system.
Reclarified?
This is not limited to cryptocurrencies, e.g., gold-based currencies cause people to “waste resources” mining.
Yes, the ‘mining’ metaphor was well chosen.
In terms of that gold analogy, what we are talking about in the context would be if gold spontaneously generated itself in proportion to the amount of existing gold and automatically buried itself at whatever depth makes it barely worthwhile to dig up. That waste is the unavoidable cost of making bitcoin-style cryptocurrency have ongoing inflation.
Okay, I think I understand the argument that Bitcoin will likely be permanently volatile, because growth at exactly the rate of the global economy is not a stable equilibrium for the reasons you describe (esp. ‘people like to time bubbles’).
Thinking about this a bit more though, it seems like the same argument would apply to any asset we might otherwise expect to grow in sync with global wealth. In particular, the apparently-perfect-store-of-wealth-attracting-investment-and-appearing-to-be-an-even-better-store-of-wealth phenomenon seems like a straightforward explanation of what’s been happening with the price of gold in the last decade.
But also, it seems like this argument could even apply to the stock market as a whole—would we expect a global ETF (like Vanguard’s VT) to grow at the rate of the world economy? Is that stable?
So I’m curious, do you agree that the no-stable-equilibrium argument applies to the price of these other assets as well, and if so, does the existence of Bitcoin still seem like it would be a problem for the global economy?
It partially explains the price of gold, yes. Gold’s situation isn’t really the same for three reasons: First, gold can be mined if the price goes too high, and higher prices would imply larger amounts of recoverable gold. Second, a lot of the gold on the market is paper gold, theoretical gold that two parties are trading rather than sending large gold bars around, which also adds to the supply. But most of all, unlike the supposed use-case of Bitcoin, gold is not being used as a medium of account or medium of exchange any more, just one store of value among many, so its real competition is not paper gold or mined gold but other stores of value such as platinum, silver, real estate, and many other things being added to the competition for ‘stores of value’ as the economy grows. If the same fraction of the population tried to store the same fraction of their net assets in gold today as in the 1600s then the price of gold would be vastly higher—or so I would think, I haven’t run the numbers. But this in turn means that the share of the economy represented by gold can easily drop further, making it less than a perfect store of value etcetera, although gold has still tended to be a better store of value than fiat currency.
Of course fiat currency is really supposed to be a medium of exchange and account, not a long-term store of value, though dumb people like me tend to use it as a store of value too because it’s convenient and we haven’t gotten around to setting up anything different and we don’t have that much value to store. And then using your medium of exchange and account as a store of value causes recessions and depressions due to the paradox of thrift; when people want to consume in the future instead of the present they try to hold paper money instead of demanding equity in projects with long-term payoffs. On the plus side, central banks can, in principle, easily rectify some part of this problem by printing more money to meet demand for currency when fear rises, and thus make up for velocity slowdowns, keeping NGDP on a level growth path. On the minus side, central banks are stuck in 30-year-old economic thinking and don’t keep NGDP on a level growth path. Bitcoin has the potential to make things much, much worse though.
New stocks on the other hand are constantly being created as the economy grows—no particular stock, or set of stocks starting at a fixed time, are guaranteed to grow at the same rate as the global economy.
To paraphrase, you’re pointing out that stocks and precious metals come with built-in demand shock absorbers, whereas Bitcoin has none. I’m not totally sure that I accept this point, because I could see alternative cryptocurrencies playing the role of marginal new stocks or newly mined gold. However, even if Bitcoin were unique in having no demand shock absorbers, I’m not sure this matters, because it seems empirically to be the case that these shock absorbers are not always up to the task, and that both stocks and precious metals do experience a great deal of price volatility, even over the medium to long term.
In other words, even if Bitcoin is especially sensitive to changes in demand, it is neither novel nor unique in being susceptible to bubbles.
This would seem to me to imply that Bitcoin’s existence and use as a store of value is no threat to the economy. (And its use as medium of exchange seems harmless as well.)
It would seem that problems would only arise for those who try to use Bitcoin as a unit of account. This is in line with Wei’s comment where he suggests that with a currency in fixed supply, fluctuating velocity of money implies that either prices or GDP must be unstable.
So my conclusion is that using Bitcoin as a medium of exchange or store of value is not detrimental to the economy, but one should continue to price goods or services in some other fiat, ideally NGDP-targeted, currency. Does that sound about right?
Using it as a store of value is detrimental. Anyone bidding on a Bitcoin is not bidding on a productive project.
It seems that the same goes for gold, real estate, and so forth when they are used as a store of value. The difference is that unlike bitcoin, these things have other productive uses that they could be put to, less expensively, if they weren’t being used as a wealth-counting mechanism.
That seems to be saying that all of the BTC would always be able to buy exactly all of the things. I can’t imagine how that could be the case.
According to Wikipedia, “velocity of money” is not a constant but tends to fluctuate. The article lacks citations, but I think this is the current mainstream view among monetary theorists. My understanding of the implication of this is that if you have a fixed supply of money then either prices or GDP would have to be unstable. In other words either you end up with an economy with very flexible prices that change constantly, or you end up with an economy that goes through constant boom and bust cycles (or some combination of each), and both of these outcomes are costly.
I feel like this isn’t nearly the issue it is made out to be when you separate real growth from nominal growth. Say you have real growth but prices fluctuate a lot, why should you care? Frictional costs should decrease over time as people figure out how to hedge properly in this environment.
So is the solution just to use Bitcoin as a medium of exchange and a store of value, but not as a unit of account? Then prices are free to fluctuate in BTC terms, while they can remain relatively stable in fiat terms, and GDP will be unaffected.
That would make it a terrible at being a medium of exchange or a store of value, though, wouldn’t it? No one knows how much it’s worth, and you have to acquire some, pass it off, and then (on their side) turn it into currency every time you use it.
That depends on how volatile it is. On the timescale of a single transaction, a certain level of volatility might not matter very much even if the same level of volatility would prevent you from wanting to set prices in BTC.
I wonder to what degree FAI/CEV engineering considerations overlap with cryptocurrency/efficient-market engineering considerations. If they are a close match, encouraging the development of the latter would have benefits for the former, and could even be essential to overcoming scaling problems (since FAI is harder to sell investors on than cryptocurrency).
This isn’t just a random idea; markets are how humans in the absence of superintelligence actually do try (with some, not-unlimited success) to implement their values. Prediction markets are a possible extension of this concept, but even your run-of-the-mill securities markets are reliant on various kinds of predictive logic that responds somewhat to human desires and needs.
Not trying to trivialize the AGI field since it is outside my specialization, but is there some not-terribly-unlikely way in which a really good cryptocurrency could basically be/evolve into the same thing as an FAI? If so, are there any particular properties that would be likely to nudge it in that direction / away from uFAI?
I’m kind of concerned because I see bitcoin (and/or anything sufficiently similar) funding competition to purchase obscenely large amounts of hardware—which could possibly even extend to the point of satellite arrays that harvest solar energy, and space based fabrication of new ones. If it gets to that point, we might end up with a Dyson sphere that basically does nothing but compute bitcoin hashes. Extraordinarily wasteful, but not necessarily catastrophic for existing humans if the network continues to recognize them as owners/controllers of the resources in question.
This seems unlikely. If you’re going to invest that much capital, why waste it on bitcoin hashes when you could instead provide a product and sell it? This would be analogous to worrying that everyone will go into the financial sector because it pays so well, and we’ll have no one left producing actual goods.
Hmm. I think you’re probably right, now that I think about it. The maximum size of the bitcoin reward falls towards transaction fees, which are themselves a small fraction of any given transaction. So there should tend to be significant money out there to reward manufacture of other kinds of space based goods more highly than bithashes.
Basically zero on a technical level. Philosophy of caution overlaps with cryptography. Some econ knowledge overlaps with hard takeoff theory.
Well if Satoshi is in academia and has tenure.