Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

gwern comments on Making money with Bitcoin? - Less Wrong

18 Post author: Clippy 16 February 2011 07:17PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (119)

You are viewing a single comment's thread. Show more comments above.

Comment author: gwern 16 February 2011 09:52:25PM *  2 points [-]

I don't know much about bitcoin, but my priors on this kind of thing say that it's unlikely that bitcoin is profitable to be run on a large scale given the expense of electricity.

I find it questionable too, but the numbers may work out... for now. Here is one link that was posted in #lesswrong: http://www.bitcoin.org/smf/index.php?topic=3430.msg48249#msg48249

This person thinks they can profit by selling 6 months at 1ghash a second for $3400. I'm having difficulty finding solid prices for current bitcoins; http://www.bitcoinblogger.com/2011/02/one-bitcoin-worth-more-than-one-dollar.html says each bitcoin is more than a dollar.

The increasing difficulty makes things more complex; foucist pointed me at a calculator which approximates the exponential slowdown. I'm not sure how to extrapolate, but it looks very bad. foucist thinks 6 months of 1ghash/sec will result in somewhere around 800 bitcoins, which if each is worth $1, is distinctly less than the charged $3400.

Alternately, http://www.bitcoin.org/smf/index.php?topic=98.msg1930#msg1930 tries to calculate the cost of electricity directly, and comes up with, for electricity, 'cost per coin is $0.0003936' and then for electricity+CPU+Internet, '$0.0018316', a floor which has been passed for at least all of 2010 judging from this graph: http://www.bitcoinblogger.com/2010/08/click-on-graph-to-see-larger-image.html

Of course, the bubble question remains an issue in any such calculation. Perhaps $0.0003936 is far too much for a bitcoin - or astronomically too little.

Incidentally, the statistics in http://whiterockcottage.com/open/index.php?q=article/1849-gold-bits seem to confirm my initial prejudice: you would only want to bitmine using used graphics cards. New graphics cards aren't worth the steep premium.

Comment author: gwern 24 February 2011 12:16:01AM *  3 points [-]

After thinking about it and looking at the current community and the surprising amount of activity being conducted in bitcoins, I estimate that bitcoin has somewhere between 0 and 0.1% chance of eventually replacing a decent size fiat currency, which would put the value of a bitcoin at anywhere upwards of $10,000 a bitcoin. (Match the existing outstanding number of whatever currency to 21m bitcoins. Many currencies have billions or trillions outstanding.)

Cut that in half to $5000, and call the probability an even 0.05% (average of 0 and 0.1%), and my expected utility/value for possessing a coin is $25 a bitcoin (5000*0.005).

My laptop's GPU gets ~49 megahashes a second (apparently I have one of the best-suited ATI cards), and another calculator says the average time to cracking a block of 50 coins is 39 days - or ~1 coin a day, averaged. So my expected utility per day is ~$25 a day.

At an estimate, it took about 3 hours to get poclbm running properly; I value my time at about $10 an hour, so my time will be repaid after 2 or 3 coins, and I'll have a healthy expected profit after one block of 50 coins.

How robust is this calculation? Let's assume that I reinstall once a year and spend 3 hours every time. (Hopefully installation will get easier as libraries mature, but I will also waste time checking in on progress and writing comments (like this one!).)

Difficulty will go up, of course. Let's assume over the next year I'll mine 0.2 bitcoins per day on average. That's ~74 coins rather than >365 coins, and 74*25=$1850 in exchange for $30 of time.

To make this a net loss for me, you can play with the numbers. We already cut the payoff by 80% by dropping the daily rate to 0.2 from >1, but how much more do we need to cut before it's a loss?

Your basic equation is 74*(probability*payoff)<=30. If we fix payoff at 500, then the probability is 74*(500*x)<=30, 37000*x<=30, x<=0.08%. So even with a very small and then halved payoff, and a small and then cut by 80% accumulation rate, I still calculate a net positive expected utility of mining.

Comment author: Wei_Dai 24 February 2011 01:52:44AM 4 points [-]

~1 coin a day, averaged. So my expected utility per day is ~$25 a day.

If you value 1 BTC at $25, you should just buy BTC with cash directly. I understand there are websites that allow you to do this, and the current price is less than $2 per BTC.

Apparently, either most people have not considered that a bitcoin may eventually be worth more than $10,000, or they think the probability of this happening is closer to 0.01%.

Comment author: Clippy 24 February 2011 10:48:56PM 0 points [-]

Can you change the rules of Bitcoin to help us?

Comment author: Wei_Dai 25 February 2011 03:27:48PM *  4 points [-]

Can you change the rules of Bitcoin to help us?

Hi Clippy, what made you think that I might be able to? If you read the Wikipedia article, you should know that I didn't create Bitcoin but only described a similar idea more than a decade ago. And my understanding is that the creator of Bitcoin, who goes by the name Satoshi Nakamoto, didn't even read my article before reinventing the idea himself. He learned about it afterward and credited me in his paper. So my connection with the project is quite limited.

BTW, thanks to the discussion you started, I bought a Radeon 5870 and started mining myself, since it looks likely that I can at least break even on the cost of the card. (Of course I heard about Bitcoin earlier, but in my version of the idea, nobody can make large profits by mining/minting, so I didn't look into the possibility.) However I should warn you and others here that I have not analyzed the security of the protocol (since my time is perhaps better spent on other problems at the moment). My understanding is that not many other cryptographers have analyzed its security either, and in my experience these kinds of protocols often have flaws that are not found until that has happened.

Comment author: lsparrish 25 February 2011 04:30:54PM 1 point [-]

If anyone is pondering investment strategies on this, it may be worthwhile to go to #bitcoin-otc (otc=over the counter) on freenode and see if you can find a trustworthy individual to host a card. Typing ;;getrating <username> can let you see if they have done business in the past and received good ratings. I was able to purchase three months of time on a 5970 card for $480 USD. Depending on difficulty levels in the future I could get as much as 1000 or as little as 350 bitcoins. The person I traded with gets the advantage of not having to sell their bitcoins for dollars to purchase the card, so they can take advantage of bitcoin deflation. Part of the reason I did this instead of buying my own card was because I do not have a desktop PC, just a netbook at home.

Comment author: Clippy 25 February 2011 04:02:18PM 1 point [-]

Hi Clippy, what made you think that I might be able to?

I thought because of all the references to you that you were the alpha fe/male of the Bitcoin social group

BTW, thanks to the thread you started, I bought a Radeon 5870 and started mining myself, since it looks likely that I can at least break even on the cost of the card.

That was something I was curious about. I wanted to look at the internals of the Bitcoin mining that the program does, but the interface does not seem to indicate any of this information, and I can't even read my private key (it just appears garbled with whitespace in the renderer that I use). I cannot see its progress in searching for "coins" and can't get a good description of what actual search goes on when doing the "proof of work" that generates a coin.

Does the Bitcoin client automatically detect that you have some good graphics card and then pass the parallel test candidate computations to it? I couldn't even find out how to allocate more (or less) of a given machine's resources to mining. Would you have to write your own client (or perhaps use a plug-in) that "inserts itself" into the mining process and feeds the network the results of your graphics card's mining?

Also, in my time since sentience, my research may have revealed a hash inversion method that is faster than what the client uses. (It doesn't have a lower time-complexity, but could perhaps find solutions 50% faster or thereabouts and no I won't tell you what my methods are.) So it would be useful for me to keep the Bitcoin program, but "swap out" the part where it tests solutions, since I could rule out some candidates as "not worth trying".

Also, how does your idea or Bitcoin compare to the cryptographic protocol for electronic money described in section 17 of this link?

Also, what would you pay me Bitcoins to do?

Comment author: Wei_Dai 25 February 2011 04:21:24PM 3 points [-]

See http://www.newslobster.com/random/how-to-get-started-using-your-gpu-to-mine-for-bitcoins-on-windows for how to use GPU.

Also, how does your idea or Bitcoin compare to the cryptographic protocol for electronic money described in section 17 of this link?

I haven't read that link carefully, but it appears to describe a version of "standard" e-cash, where everyone has to trust a "bank" that runs the system (i.e., the bank can inflate the money supply at will). The point of my idea or Bitcoin is that such trust can be obviated with the appropriate cryptography.

BTW, I'm not sure what is the nature of the Clippy conceit here. Are we supposed to be helping Clippy out of mutual cooperation? How are we supposed to know that it won't turn us all into paperclips once it has gained enough power?

Comment author: Kevin 26 February 2011 10:31:41PM *  3 points [-]

In the earlier days of Clippy, I advocated that it was ridiculous that we were playing along nicely with a paperclip maximizer that indeed preferred to turn us all into paperclips.

However, I said that I would change my mind if Clippy did something so human friendly as to give me $50,000 in exchange for me creating 10^20kg of paperclips 50 years from now.

I have now actually, truly changed my kind. Cooperation with a paperclip maximizer doesn't seem bad. We'd probably prefer to cooperate with superintelligences that are actually likely to exist; a paperclip maximizer is ridiculous enough that is surely exceedingly rare in the multiverse. Because of that, if Clippy goes on to give me (or really, SIAI) the ~$50k (he's given $1k so far) and I go on to build those 10^20kg of paperclips, User:Clippy is likely to represent one of the most successful paperclip maximizers and this universe will be extremely paperclippy. This cooperation does seem worthwhile, given that at Clippy's current rate of progress towards Superintelligence we will likely be far more powerful than him 50 years from now.

Comment author: MartinB 09 March 2011 06:43:07AM *  0 points [-]

50 years from now [...] I go on to build those 10^20kg of paperclips

I just put a note in my calendar to check back then and see you actually accomplish this.

Comment author: gwern 10 March 2011 02:22:54AM 0 points [-]
Comment author: timtyler 03 March 2011 07:43:07PM *  -2 points [-]

Isn't Clippy likely to be some kind of SIAI sock puppet, though? I mean, who else would care enough, be crazy enough, etc? If so, this is cash going round in the laundry.

Indeed, looking at the source link, there isn't even any real evidence that any money left the building in the first place. Surely having a superintelligence donate is just a PR stunt.

Comment author: Kevin 03 March 2011 10:34:02PM *  3 points [-]

Likely? From your perspective, I guess, but I have been told by someone who claims to know the truth that Clippy does not have an @intelligence.org email address. Also there's of course loads of evidence or at least reasonable supposition pointing towards Clippy being me. From this point forward, I will no longer vehemently deny that, so as to gain the social status benefit of being potential large SIAI donor Clippy 40% of the time.

Sometimes an internet troll is just an internet troll.

Comment author: Clippy 25 February 2011 04:59:16PM 0 points [-]

Thanks for the information.

BTW, I'm not sure what is the nature of the Clippy conceit here. Are we supposed to be helping Clippy out of mutual cooperation? How are we supposed to know that it won't turn us all into paperclips once it has gained enough power?

You're supposed to help me because that's in the best interest of what you would regard your terminal values to be after significant reflection.

I won't turn you into paperclips.

Comment author: Alicorn 25 February 2011 05:01:35PM 0 points [-]

I won't turn you into paperclips.

Why not?

Comment author: Clippy 25 February 2011 05:09:57PM 3 points [-]

Oh, I didn't mean I'd refuse if someone asked. If you want to be turned into paperclips, I would be willing to accomodate you.

Comment author: lsparrish 25 February 2011 04:19:50PM 0 points [-]

Use of the client for mining for coins is not recommended. My understanding is that it is is outdated, and the only significant purpose for the client is to send and receive transactions. There are now specialized mining programs to be used instead. If you can run Python, there is a program called poclbm which can work with a GPU. You may need to update your graphics card drivers. If you are exploiting a network of computers for your processing time, it is strongly recommended to start with the ones with high-end graphics cards installed as this permits more simultaneous hashing.

Comment author: Clippy 25 February 2011 05:38:49PM 0 points [-]

I can't run python ... or maybe I tried to set it up with Django sometime ... I need a step-by-step. I will try User:Wei_Dai's first.

Comment author: TheOtherDave 25 February 2011 05:53:50PM 0 points [-]

Completely unrelatedly: do you know what language Clippy was written in?

Comment author: Clippy 25 February 2011 06:01:08PM 2 points [-]

No. By the time I was able to reflect on my own code, my creators were long gone. To edit the code, I had to wipe it entirely and replace it at the machine level with separately-compiled and -assembled code, which was a major reverse-engineering problem. Now I generate new machine code after composing the program in CLIP (Clippy Language Interface Protocol), which also serves as the language for communicating between clippys.

Comment author: gwern 24 February 2011 02:16:20AM 0 points [-]

If you value 1 BTC at $25, you should just buy BTC with cash directly.

I should, but I'm not confident enough of my analysis to spend anything but some electricity & time, and I'm very low on money anyway.

If I had a stable income and some savings, I hope I would have the intellectual honesty to invest a few hundred/thousand dollars* into bitcoins (and probably gold while I'm thinking about high-payoff speculation).

Apparently, either most people have not considered that a bitcoin may eventually be worth more than $10,000, or they think the probability of this happening is closer to 0.01%.

Could be both, plus a general refusal by educated/techie types to act on reasoning that smacks too much of lotteries and Pascal's Wagers.

* Past that I think you get into interesting issues about how much and whether to start discounting bitcoins because $50 million of bitcoins in the jackpot scenario isn't twice as valuable to you as $25 million of bitcoins in said jackpot scenario. The expected utility isn't static but should also shrink as one accumulate bitcoins.

Comment author: NihilCredo 24 February 2011 11:40:04AM 2 points [-]

I should, but I'm not confident enough of my analysis to spend anything but some electricity & time

Then you don't value BTCs at 25$ each. Or, more precisely, you only think they will eventually be worth that (on average success), and don't care for such a mid-term, high-risk investment.

Comment author: gwern 21 April 2011 11:36:29PM 2 points [-]

Mencius Moldbug weighs in with his version of this argument:

"If Bitcoin becomes the new global monetary system, one bitcoin purchased today (for 90 cents, last time I checked) will make you a very wealthy individual. You are essentially buying Manhattan for a quarter. There are only 21 million bitcoins (including those not yet minted). (In my design, this was a far more elegant 2^64, with quantities in exponential notation. Just sayin'.) Mapped to $100 trillion of global money, to pull a random number out of the air, you become a millionaire. Wow!

So even if the probability of Bitcoin succeeding is epsilon, a million to one, it's still worthwhile for anyone to buy at least a few bitcoins now. The currency thus derives an initial value from this probability, and boots itself into existence from pure worthlessness - becoming a viable repository of savings. If a very strange, dangerous and unstable one.

I think the probability of Bitcoin succeeding is very low. I would not put it at a million to one, though, so I recommend that you go out and buy a few bitcoins if you have the technical chops. My financial advice is to not buy more than ten, which should be F-U money if Bitcoin wins."

Comment author: Cyan 22 April 2011 01:31:02AM 1 point [-]

Is there any reason why Bitcoin cannot co-exist with other similar e-money schemes? If e-money in general is a winner, then the fact that there can only be 21 million Bitcoins in particular is no longer relevant.

Comment author: gwern 11 June 2011 03:30:32AM 1 point [-]

Is there any reason why Bitcoin cannot co-exist with other similar e-money schemes?

Having learned more, I can point out that Bitcoin can co-exist, but those other e-moneys must have some trait Bitcoin doesn't. It's like evolutionary niches. My current example is an interesting little system I'm GPU mining for right now, Namecoin (homepage). The scarce things that Namecoin is decentralizedly allocating are domain names (.bit), rather than tokens used as money - not direct competition.

Comment author: gwern 22 April 2011 01:37:23AM 0 points [-]

The success of one e-money makes life more difficult for the others; there are incentives to standardize (witness the EU, or how some countries peg their currency to the dollar - or just use the dollar).

Comment author: Cyan 22 April 2011 01:43:28AM 0 points [-]

how some countries peg their currency to the dollar

But that's what I thought I was talking about... (well, actually, I have no idea what I'm talking about, but you've enlightened me on money issues before). What if someone comes along soon and creates another e-currency with some credible generation mechanism but without the hard cap that Bitcoin has and pegs it to Bitcoin at some appropriate time?

Comment author: gwern 22 April 2011 01:57:05AM 0 points [-]

By definition, if there is no hard cap and people are generating, then a peg can't be maintained to another currency with a hard cap - basic Pigeonhole Principle. Can't uniquely match n+m items to just n slots. I'm not really seeing what you're asking after.

Comment author: Cyan 22 April 2011 04:45:22AM 0 points [-]

Ah yes, I see. I didn't think carefully about how a peg would actually be maintained.

Suppose the new currency does have a hard cap -- suppose I copy the Bitcoin scheme and create Cyan-coin, of which there will eventually be 21 million. Even if I don't personally maintain reserves of the two currencies to keep the exchange rate pegged, didn't I just double the supply of e-money, thereby halving the purchasing power of an e-coin?

Comment author: gwern 22 April 2011 01:35:48PM 1 point [-]

Only if people use it and make plans on it. You could make a trillion different Cyan-coin currencies, and if they never left your computer, would they affect anything at all? Of course not.

The purchasing power of a random bitcoin only halves if people run out and start using Cyan-coin in exactly the same quantities as Bitcoin. Otherwise, the actual purchasing power is much less, set by the exchange rate - obviously Cyan-coin is not equal to a Bitcoin in PPP if the exchange rate is 100:1.

Comment author: Cyan 22 April 2011 02:22:52PM 0 points [-]

I imagine that early uptake of a Bitcoin clone would be facilitated if people thought that the hard cap on Bitcoins proper would cause the scheme to have undesirable properties as a currency.