At This Rate, The Last New Bitcoin Will Be Issued 15-55 Years Ahead Of Schedule | Hacker News Reader

At This Rate, The Last New Bitcoin Will Be Issued 15-55 Years Ahead Of Schedule

thegenesisblock.com

50 pointsmdelias51 comments

In 2013, a huge part of this is due to the flood of ASICs coming online. We've seen CPU -> GPU -> FPGAs -> ASICs in just a few years, but these improvements cannot keep coming at this rate. With ASICs, bitcoin mining has just about caught up with the rest of technology and the computation / $ that miners can bring to bear will stop jumping at such a huge rate.

I disagree with the statement in the article that "there’s strong reason to believe this trend [surges in network computing power] will continue."

Reminds me of this: http://xkcd.com/605/

There's still at least one order of magnitude of performance to be had with ASICs, and honestly that's conservative.

The danger is specialising such expensive silicon just on BTC, which is much more likely to cause a wall to be hit, rather than the raw performance that could be had if really serious resources were thrown into a very high end design on today's leading edge manufacturing processes.

It probably won't go on forever, but at this point there's three years of the trend and only a small fraction of the globe has even heard of bitcoin. Would be surprised if network hashrate didn't grow like it has been for the foreseeable future.

But they're building on established technology. This is like looking at a third world country just as they begin modernizing... holy cow, they've recapitulated the development of the developed world in just 20 years! They're on fire! They're going to take over the entire world! Except that they're not, because they've been buying their advances from the developed world. (This is not a criticism; tech transfer like this is an awesome, wonderful thing. It's just that one of its tiny side effects is that people can be fooled by the curves.)

Similarly, BitCoin isn't going to be able to continue advancing at this rate when the problem of accelerating BitCoin computation reaches the point where it is reducible to the problem of accelerating transistors in general.

But there's no need for it to continue forever here, since there's a finite target. The question is if current growth can be maintained until there are no more Bitcoins to be mined, not if the curve is sustainable.

I'm at a loss as to what you think you're disagreeing with me about. I don't see how observing that something won't go on forever is somehow a claim that it should.

There's still a lot of progress to be made on the ASIC front. The current generations of Bitcoin mining ASIC are brutes, several places mention that the Avalon miners use a 120nm process (think Pentium 4), which means that there's a lot of headway in terms of speed and power consumption that can be made in the future.

If there's an outrush of processing power, adjustment in the easing direction won't be as fast. (At least one smaller, post-Bitcoin cryptocurrency has faced a crisis when a large amount of hashpower withdrew... because the remaining hashing would take 'forever' to even advance to the next 'adjustment point'.)

So this question is a bit early to call.

It's remarkable testament to the power of greed that so much computing power is being poured into the digital equivalent of manufacturing Beanie Babies. Are there any reliable estimates on the volume of Bitcoins that are actually used as currency, compared to the volume of purely speculative trading?

It isn't greed, it is market opportunity. So long as they're cheaper to mine than they can be sold for, someone will exploit the opportunity. There's absolutely nothing wrong with that.

Once the ASICs replace the other mining systems, the correction estimates will be more accurate.

Many millions of dollars worth. Huge number if you include gambling related commerce. Even without gambling it is almost certainly an 8 figure amount.

edit: monthly

I realize it's sort of a funny thing to say, but I would argue that a bitcoin has more intrinsic value than a beanie baby. It's ease of transfer is a huge utility.

I would argue that a bitcoin has more intrinsic value than a beanie baby.

I would argue that it has less; kids have always loved stuffed animals, whereas no one found intrinsic value in a cryptographic hash until someone thought to brand and promote one.

A Bitcoin's value comes from the Bitcoin system, which does have some utility as a sort of distributed ledger. The problem is there are already plenty of e-cash systems already in place, virtually all of which are easier to use than Bitcoin.

As a currency, Bitcoin's volatility is a major disadvantage. No one will spend a Bitcoin if they think it will be worth more tomorrow. No one will take a Bitcoin if they think it will be worth less. Success as a speculative investment is generally antithetical to success as a currency.

So, exactly which fiat currency do you use that does not fluctuate in value?

Fluctuating is one thing, gyrating wildly is another.

How will transactions get confirmed in the block chain if there's no longer any blocks worth mining? Won't most people stop mining blocks to save money?

Or is the hope that enough people will continue mining out of goodwill to keep the economy flowing?

Miners get bitcoins two ways:

1) A reward (currently 25 BTC, cut in half at regular intervals)

2) Transaction fees: a percentage of each transaction in the block, typically with a 0.01BTC minimum currently.

Once the rewards are gone, miners will be competing over who can include the most transactions in their blocks in order to capture most of the fees.

at the point, the miners will collect fees for processing transactions.

For large transactions, you typically 'donate' a small bit to the miners, which is earned by the person who mines the next block. It is expected that as the number of transactions keep going up, the reward from transaction costs will eventually outstrip the block reward.

P.S 1 might have messed up terminology P.S 2 too small transactions (typically by Sastoshi Dice) even today are not being accepted by miners unless accompanied by a donation.

I think the progress will become more predictable now that they've basically reached the maximum amount of compute possible from a single chip. The speed breakthroughs going forward will be optimized ASIC designs and smaller gates. But going from CPU -> GPU -> FPGA -> ASIC we saw huge gains in performance. That's no longer the case and hence we should see more linear scaling of mining operations.

But I do tend to agree with the assessment that if BTC remains popular and continues to gain credibility then we will see more and more mining operations come on line as the mainstream starts to invest. I mean, can you imagine if a large bank decided one day to hedge its bets and do it's own dark pool of miners? They could throw so much money at it that it could really disrupt current trends.

Kinda interesting, but really not a big deal. The distribution mechanism is working as intended, just faster.

A better guess would be that the last new bitcoin will never be mined, because exponential growth in hardware dedicated to it will not last, given that the cost/benefit ratio (in terms of power, hardware, and effort) gets worse the fewer bitcoins you can mine per minute. And when people stop being able to mine them, their value will drop because people aren't making them anymore, which leads to it being even less worthwhile to mine them, until they are being mined too slowly even to hit the adjustment point - in other words, when the bubble pops, the currency will collapse completely.

So from that point of view, the last new bitcoin is likely to be mined earlier than expected - but it won't be the 21 millionth.

If there were no transaction fees built into the bitcoin protocol, then that might be true. But the system is designed to slowly and smoothly transition away from funding the mining activities by issuing bitcoins to funding the mining activities by paying transaction fees.

How do you write something like

"The mining rate has been faster than expected over the last four years. This trend will probably continue as ASICS enter the market."

and then proceed to draw a chart covering the next 130 years!?! This trend might last five.

I would like to see "cost to mine[1]" vs values in dollars over time. Go lazy web, go!

[1] In a unit expressible in dollars. Like the cost to run a Linode machine to do it for the time needed to mine it, or something.

http://www.bitcoinx.com/charts/

"That's a lot!" Nah, if anything, it's surprising Nakamoto would even be that close. Hopes are one thing, but nobody predicts successes of this magnitude.

How is a naive back-ward looking algorithm a "success of this magnitude".

Some thinking about it and any experience in modelling time-series would lend someone to put in something that predicts the difficultly of the next 2016 blocks (with some damping so that it doesn't oscillate on a constant signal) rather than using a value derived from the previous 2016.

He didn't mean that bit of the algorithm was a success, he meant that the 2016-block algorithm was failing because adoption has been so strong.

"Adoption" in this case being ASICs coming online...

But adoption doesn't affect this, in fact it was more strongly negatively affected in 2010 when adoption was overall lower.

Is it really success when a monetary policy decision turns out to have been wildly inaccurate?

Wildly inaccurate?

How happy would central banker be if they could get outcomes within 10% of their planned goals?

A central banker could easily achieve a perfectly accurate prediction of the amount of currency available, which is the type of prediction we're talking about. They print the money, so they can just print according to their predictions.

The types of central banking predictions that are usually criticized are predictions of prices in the economy. Bitcoin fares no better there.

> A central banker could easily achieve a perfectly accurate prediction of the amount of currency available

Not really. Banks and corporations hold reserves, lend "money" that's actually debt between banks not backed by hard currency, pay debt directly to the central bank... it's not that easy. At a given time, there are many "amounts of currency" coexisting at the same time.

Right. I was trying to choose straightforward terms that would make it clear that I was referring to the monetary base, not any other measure of the money supply. I guess that didn't work.

The prediction was 131 years -- I'm not sure 15 years early would be considered wildly inaccurate. I'm also not sure why we should believe that the assumptions the article makes will hold true over the next 72-112 years.

Maybe what I mean is wildly imprecise. The figure of 131 suggests a precision in years, but based on these projections it should clearly be in decades. Since there will likely be dramatic effects to running low on coins to mine, it's a pretty important thing to predict, and if that's not possible it's not a great sign.

I guess I don't know what sort of precision the market requires to deal with it, though.

The prediction was given as ~2140 rather than ~131 years, which perhaps implies less precision, and anyway was presented as an estimate.

Fair enough. I don't mean to belabor the point.

Not long from now, a quantum computer will solve each new block within a femtosecond before getting back to its main task of high freq trading and playing out millions of tactical nuclear war simulations.

OK, now I understand mining a bit better. Its a race, with the difficulty tuned to try and make each race last 10 minutes.

Not just to make each race last 10 minutes, but to also make sure it scales with the processing power available. This keeps the transaction history secure because you would need to control immensely more computing power to try and counterfeit the transactions for every 10 minutes back in the past you try to change.

So we have temporarily inflation in a deflationary currency. That is interesting phenomenon.

You're mixing two definitions of inflation. Bitcoin undergoes monetary inflation as the supply increases until the late 21st century. Since the supply increases more slowly than the demand for the currency increases, the bitcoin economy sees price deflation. This isn't a particularly interesting phenomenon because it's common for the value of a currency to increase (price deflation) even as the supply of the currency increases (monetary inflation). This is what happens with gold when it is used as a currency, for instance.

When the terms "inflation" and "deflation" are used without qualifiers, they usually refer to prices, not money supply.

Can one of our resident cryptography geniuses (e.g. cperciva, tptacek) weigh in on how Peer2PeerCoin (PPCoin) does/doesn't address Bitcoin 51% vulnerability?

Not a genius, but here is my take.

The ability to mint blocks without mining makes a long term attack harder (since even if you leveraged coin-age in your attack you have finite coin-age to work with).

This is because if you have 51% of the power you can beat the rest of the network when it comes to generating N blocks. However if someone mints a coin-age block you can't make that guarantee anymore. Note that this is especially painful because attempted takeovers like this require giving up valid blocks. (So if you tried to do a 6 block rollback you would need to give up 6 blocks in the case of failure)

So I would say that long term a complete shutdown of the protocol due to a 51% attack is unlikely, as coin-age would make it require more than just a lot of hashing power, especially as the protocol gets older.

However I think it may provide a weakness in some edge cases. I wonder what happens if someone tries to use a significant amount of coin-age to do a quick double spend (say spend 1,000 peercoin that are about 3 years old then use the peercoin age to create a parallel block chain to undo the spend). You would have to give up your coin-age in such an attack, but that could be a good trade depending on the rate that coin-age pays out at.

However to be fair, this would require a lot of faith in peercoin, since it appears the protocol calls for a significant timelapse to acquire coin-age. So the likelihood of it happening is lowered.

It would help if you could point us at a formal description of the protocol.

http://www.ppcoin.org/static/ppcoin-paper.pdf

That is pretty lacking in detail. Is there something a bit more complete -- something that might actually be useful for even a cursory analysis?

[Edit: for example, it is not clear how the timestamp is computed/validated by the network. Do all nodes have to have synchronized clocks?)

Reply on news.ycombinator.com