Epicenter - Learn about Crypto, Blockchain, Ethereum, Bitcoin and Distributed Technologies - Dave Hudson: Insights from the Data Mine and Other Adventures Around the Block

Episode Date: August 3, 2015

Bitcoin mining and the intricate game theory that surround it is a topic that we’ve visited many times. Difficult to grasp in its complexity, it is these dynamics that determine the security of Bitcoin in the short and in the longer term. This time we were joined by none other than Dave Hudson. He is the author of the leading mining blog hashingit.com, VP of Software at PeerNova and had a long career in chip manufacturing including at Qualcomm. We talked about dynamics in the mining market, security issues with Bitcoin’s design and his data-informed views on the blocksize debate. Topics covered in this episode: How he became interested in mining and what metrics he finds most interesting How miners would be affected by a blocksize increase The flaws in Satoshi’s design of the reward scheme Why transaction fees will not make up for the block reward decrease in 2016 and why that will lessen the security of Bitcoin The long-term role of Bitcoin as a settlement network The problem with accepting 0-conf transactions The future of mining and why 21 Inc’s plans seem nonsensical Episode links: Dave's Personal Blog Waiting for blocks The Myth of the Megabyte Bitcoin Block The Future of Transactions Fees Bitcoin SF Devs Seminar: A deep dive into understanding Bitcoin mining hashrates Dave's company PeerNova This episode is hosted by Brian Fabian Crain and Sébastien Couture. Show notes and listening options: epicenter.tv/090

This episode of Epicenter Bitcoin is brought you by Voltauro, the gold to Bitcoin Exchange. Trade gold to Bitcoin instantly and securely, starting at just one milligram. Go to Voltoro.com to deposits in Bitcoin and start trading today. And by Shapeshift, with no account or sign-up required, it's the easiest way to buy and sell gems, dash, new bits, Minero, and other leading cryptocurrencies. Go to ShapeShift.I.O. to instantly convert your altcoins and to discover the future of cryptocurrency exchanges. Hi, welcome to Epicenter Bitcoin, the show which talks about the technologies, projects, and startups driving decentralization and the global cryptocurrency revolution. My name is Sebastian Kutu.

And my name is Brian Fabin Krain. We're here today with Dave Hudson. He's someone we wanted to have on for a long time. Some of you may know him from the block, hashing it, which is the leading or one of the leading blocks about Bitcoin mining. He's written a lot of really interesting stuff there. And he's also the VP of Software at Pironova, which used to be a mining company.

They've sort of pivoted to doing other things in the blockchain Bitcoin space at the moment. And previously he was also working for Qualcomm, which is a big chip manufacturer. And then they've also been sort of involved in the Bitcoin space recently, I think we've did 21 thing, which we can talk about too later. So Dave, thanks so much for coming on. It's a pleasure to be here. So to get started, tell us how did you get into Bitcoin and especially how did you become interested in writing about mining?

Okay, so I got interested in Bitcoin by one of the guys I currently work with at Pianova. He and I had worked at another company called Ubikon for quite a number of years. And he had been off and done some other stuff and then he'd actually set up a company which is called High Bitcoin. And they were looking at building A6 for mining. And so he was really interested in that whole hardware design. And he called me one day when I was over in California and said, have you actually looked at this stuff?

What do you make of the technology? Are you interested in doing something with this? And so I looked at it and I read everything I could find. And this is about the beginning of last year, but beginning of 2014. And it was just sort of fascinating. I looked at this thing and just thought this is huge. This network's fast. Some really interesting ideas.

But I wanted to understand a bit more about it. So before I was even going to think about actually going and doing anything in the space, I wanted to understand what the space was about and how it worked. And so I started looking for information about how the core network worked. So you could read the source code. That's fine. And you could sort of see postings from various people about how mining worked. But I started trying to find any in-depth analysis.

of what was really going on in the mining network, what really worked in terms of how things were put together and how they really behaved. And I just couldn't find anything. So I'm sort of a bit of an analyst in sort of most of what I do. I like to have good data. And I like to be able to see numbers and actually compare things. So because I couldn't find them,

I sat down and started to try and work things up for myself. Yeah, that's true. A lot of when you come into the space, you quickly realize that a lot of the, the information that's circling around is either opinionated or highly anecdotal and having actual statistical analysis, other than just the data that's provided by just the raw blockchain is really valuable. Can you tell us what are some of the metrics that you find the most interesting or some of the core metrics that you're specifically looking at when looking at mining?

So I think there's lots and lots of really interesting things. And one of the things I've found fascinating is that every time you look at a different set of data, you can find something new in there that you haven't seen before. And I've seen lots and lots of different patterns. But the stuff that started me at being interested was looking at things like hash rates. I had a background of doing semiconductor development as well.

And so seeing a technology that erased through every generation of all the way from software through to three generations of ASICs via GPUs in the space of five years was quite amazing. I've never seen anything like that before. So I wanted to really understand what was driving that and understand where those numbers came from and understand what they actually really meant. So the starting stuff I was looking at was hash rates and trying to understand the relationship between hash rates and the technology changes and the amount of money that was involved in exchanges. Some of it was pretty good. Some of it early on was pretty wrong. But I was trying to understand what actually drove those hash rate numbers and where would that ultimately lead the network.

So that was one aspect to it. The second one was to try and understand what it means in terms of who is actually managing and controlling network. And also to understand how the network's being used. So what are the nature of the transactions? What sort of things are being transacted? Are the real transactions? Are they? What's actually causing these various things to happen?

and then to try and understand where that can actually go in the future. So the thing I'm really interested to understand was, or it was very interested to understand when I started doing it, was to see what happens when the network starts to get full, what happens when block rewards change, what happens to mining incentives and what happens to the nature of mining and see how that evolves over time. So really, I just wanted to understand how the network can evolve

and which ways it could move. And the starting point is to understand the basics of what the network does right now. So I agree that those are sort of the torque two core things related to mining is the hash rate and who's controlling the hashing power. Now, you mentioned basics. So, I mean, I think most of the listeners will be aware of what hash rates are. But just for the sake of introducing this topic, can you define what is the hash rate and how it's calculated? Okay.

So the hash rate isn't directly measured. It's not something you can actually really measure. The best you can do for the hash rate is you can sort of indirectly measure it. And this is something I looked up last year. The only metric you actually have within the main Bitcoin network is to see the number of actual completed blocks. So you can measure the number of blocks that were actually calculated. You can see the difficulty that was being applied.

You can sort of see what's happening to the difficulty changes. And from that you can get a gauge as to what the real hash rate is. is. But of course, the nature of Bitcoin blocks is that they're found relatively infrequently. It's nominally once every 10 minutes. But that has a huge amount of statistical variability in it. Something that's that random when you're dealing with a nominal 144 blocks in a day can have a huge amount of variance. So on a day-to-day basis, the difference in the number of blocks you can find could be plus or minus 20 percent quite easily. So it's only when you look at things over the longer term, you start to get a real picture for what's going.

on. And so I wanted to understand, you know, what's the variability here and what's, what can you and can't you safely say about this? And then, you know, look at whether you can actually make some predictions about how things will change over time. So fundamentally, with the Bitcoin network you have what's actually classed as a non-homogeneous or in-homogeneous process. So a process on process is basically modeling random events where each new events, in a in the case of Bitcoin, each block is statistically uncorrelated with the previous block. So it doesn't matter what the previous one was or how long it took. The next one, the amount of time it takes to find it is pretty much independent of the previous one.

And so this leads you to the situation where you can get blocks that are found in very rapid succession, or you can have blocks that take well over an hour to find. But the statistical mean is going to be 10 minutes. Is there, are there some phenomena, or patterns outside of the Bitcoin space in our life that people are familiar with that are ruled by a similar or the same statistical process? Yeah, in fact,

process is a very, very common. The classic ones that you tend to find people describe are things like radioactive decay. That's another example about a pass on process. Things like the number of phone calls arriving at a call center. There's lots of these sorts of things that actually look like these pass-on processes. And so there's a lot of mathematical

modeling that's been done for various different domains, and you use some of those to actually understand what's happening with Bitcoin. I mean, it actually turns out there are some subtleties to Bitcoin that don't make it quite as clean as some of these others. There are a few subtle interactions, but I don't think anybody's done a great deal of analysis on those yet. I think I recall learning about

plus on process and statistics in school, but that was probably one of my worst subjects. So it's fair to say I've forgotten about that. So we mentioned that, you know, and we know that nominally blocks take 10 minutes, sorry, on average to find, but that there is quite a bit of variance in that amount of time. Can you talk about that?

Yeah, so the idea is that you have an exponential decay function. So initially, you know, the most likely time to find, find something is basically near time zero, so it's when you start off. And then progressively you find fewer and fewer sort of time intervals. As you get a longer time interval, you get less and less probability that that particular time will be at the next event. So the actual function is fairly straightforward to plot because it's just a basic exponential decay.

In terms of what that actually means, though, there's a few subtleties for Bitcoin itself. So that model works if you assume that everybody has the same information at the start time, and so everybody starts from the same point. But in fact, in Bitcoin, there are some subtleties that mean that the initial few seconds aren't actually quite as evenly distributed as they should be. So there's a propagation delay time in the network where it takes time for blocks to propagate, and so you have a period of time where not everybody is actually working on the same problem. So even assuming everybody starts from the previous block, it takes a while for them to all see that previous block.

So the actual curve isn't quite that ideal process curve. And there's a scenario, I think, that needs a little bit more analysis. I haven't had the time, unfortunately, but it's something's on my list of things to do. So many will have heard about the idea of propagation, how fast blocks propagate in one context in which this has come again and has been. discussed is the block size debate. So the question whether the block size should be increased, we've talked about that quite a few times on the show. You come from this very, I would say, analytical background or analytical approach to looking at Bitcoin, right, looking at the data. So from that perspective, what is your view about the block size question?

So I think the biggest problem with the block size debate right now is there's a lack of really good data driving it. And I think there's a few things that need, if you could actually sort of plan this as an engineering exercise, there's a few things you'd want to do first. I think the first thing you'd really want to understand is what is the real behavior of the network. So I have seen some work where people have been trying to measure propagation times because that has a, that propagation characteristic has a very profound impact in terms of how the network's secured and how fast the miners are actually operating. There's a, there's a, there's a, you know, there's a a lot of the data seems to be very unreliable in the sense that nobody's really taking into

account where the majority of the hashing is right now. And so I think that there's a lot of analysis needs to be done in terms of measuring real propagation delays. So, you know, if you can get a statistical mean, that's fine, but you actually need to know what the worst cases in certain circumstances. You need to know how well connected all the various miners are and actually try and model what the network looks like and with the bandwidth that they all have

between them, what will actually happen under various circumstances. And it's not a hugely complicated model because there's only really maybe a dozen large miners at the moment. So if somebody can actually model that, I think that would go a huge way to really understanding what changing

various characteristics and network would really do to it. Yeah, I mean, I agree, right? It often seems that in the discussion there's sort of a lot of opinion and not a lot of analysis, but given that we don't have that data at the moment, what is what is sort of your personal opinion and your personal view, you know, given the information we currently have, what would happen if the block size was increased to 20 megabyte? So I think 20 megabytes would be a vast increase and I think it would be probably very unwise

to do. I think that there would be problems exposed from doing that and it's such a huge step increase. I mean, it's more than an order of magnitude increase, that I think that would be very, very problematic. And I don't think the analysis has been done to sort of size that. The truth is that in practice, you know, we wouldn't actually go to 20 megabyte blocks straight away,

but what would happen is that if somebody sort of runs a stress test like the ones that have been run recently, that would start to get, it would be, it certainly be more interesting. I think you'd see different effects on the network under those circumstances. So I think 20 megabytes is a huge change. I think the other thing that's unclear to me at the moment is just how much the blocks that we have right now are really actually full.

So I mean, I did some analysis on trying to work out just how full blocks were and historically how full it'd actually been based on what miners were really trying to mine. So we did some statistical analysis and looked at, you know, how full those blocks are and what does that mean for transaction propagations. But the big unknown is actually just how many of those transactions that were mined into blocks were actually really useful and interesting. And if you started to push up against the limit where the fees went up, how many of those transactions would just simply go away?

There were people shuffling coins around or just moving things between wallets and not really taking much regard for how they did it because there was plenty of space available. I think there's a lot of unknowns there. So what do you mean by some of these transactions are not of any? Well, there's a lot of, if you actually look at sort of blockchain.info, they actually have a statistic for some of these now where you have these sort of so-called long chain transactions

where you see things basically moving around in circles almost. So the same amount of coins just moving from an address to another. Yeah, and so obviously it's very difficult to do a full analysis of it, But there seems to be a reasonably high proportion of the transactions that have been made to date aren't actually commercial transactions. They're actually simply people moving things around and then moving them back to where they came from in some cases. Yeah, actually, one thing related to that I was sometimes thinking about, you know, one of the metrics people always talk about when they say, oh, Bitcoin is growing and stuff is the transaction volume. Of course, let's say you're a company like Coinbase, sir, right? So you care quite a lot about what Bitcoin is looked at.

And if Bitcoin is perceived to have traction, you know, when you try to raise your next round of venture capital, like that's a chart you're going to have up there. So how much does it cost to actually, you know, just keep on generating pointless transactions to show that kind of gradual increase? Probably not very much. So I think it would be a profitable.

Yeah. Even the stress test didn't cost very much. And that was actually one of the interesting things about the recent stress test is that the cost for doing that was actually pretty low. And when you're talking about people have an incentive, a lot of Bitcoin is about incentives. But when you have people have an incentive to see a higher transaction volume, it's actually very easy to create a much, much larger transaction volume. It's time for a word from our sponsors, voluntorial.com, the gold to Bitcoin Exchange.

Look, if you've ever traded gold with Fiat, you know how much of a hassle that can be. Of course, trading gold with Bitcoin makes it super easy and super fast to send money into the exchange. And with Voltoro, you can start trading gold as little as one milligram and their trading fees start at 0.2%. And with the world leading security and transparency that Voltauro are providing, you can rest assured that your bitcoins are safe. And by the way, because you're trading commodities, you don't need to provide any KYC documentations for deposits less than $5,000 worth of Bitcoins per day, which means that there's barely any, there are no barriers to entry. You can get started today. You might be asking yourself, what do those guys do all day?

Do they lie in their bathtub swimming gold? Well, you might think that, but actually they may be doing that part of the time, but the other part of time, they keep doing the same thing over and over again, and that's improving Voltaroor. And they've done it again. So this time they've added instant confirmations. They've partnered with BlockCyfer, and you can now deposit your Bitcoins and start trading straight away.

So keep improving your service. So go to walteroro.com and start trading gold today. We would like to thank Voltaura for their support of Repsend of Bitcoin. So let's talk about these stress tests. Can you give a brief context? First, like, what exactly happened? Okay, so, I mean, I certainly wasn't involved for those, so I've only seen some of the analysis that's been subsequently.

Although this was a thought experiment, I came up with the end of last year, new this year when I was trying to work out what would happen to transaction confirmations if the network starts to congest. And so the idea of the stress test is basically, rather than, say, we'll take the normal volume of Bitcoin transactions, what you'd actually do is you would run some period of time where you'd dramatically increase the number of transactions and see what happens to the behavior of the network. So the interesting thing about the stress test was that the network actually responded exactly like you'd hope it would respond.

everything kept working. Blocks were absolutely full. Nothing failed. Some transactions got delayed, especially the ones with lower fees, they would be delayed because it's in the minor's incentive to take the highest fees that are available. So from that perspective, the stress tests were actually extraordinarily positive from the network. They actually demonstrated the network worked really well.

And a few people found a few things that broke and they've been able to go and fix those. So it's actually been a good thing. And in fact, stress testing is usually a good thing for any engineering system. It's just tricky to do in the live system. But the idea is that basically you create a large number of transactions, flood the network, and see if that will affect the behavior. And in fact, one of the things that actually did come out of the biggest one of the stress tests was that you saw fees dramatically increase.

So it got to the stage where in order to see transactions really confirmed quickly. and still not taking many, many tens of minutes or hours to actually confirm. People were actually putting higher fees associated with transactions. So you actually got, for the first time, you've got a viable fee network and fees actually being paid to minors. So in this block size debate, one of the vocal points of view has been Mike Kern and he's been on this show as well to talk about that. And he has argued and written some blog posts that have gotten a lot of it.

attention that if the block size is not increased and if we do have this consistently, you know, because the stress test was only short duration, but if this keeps going on and on and that this would lead to basically a disaster with nodes crashing and particularly a bad user experience, I think that's where a lot of it comes in, right? Because all of a sudden users send transactions, they don't get confirmed. You know, they don't know what fee. choose, they might have to like resend it, but how and wallets aren't set up to do that? What are your thoughts on that? Do you think these concerns are valid? Certainly some of them are valid. I think there's no question that bad user experience is

definitely a valid concern. And I think this is something that there's been some discussion over the last few weeks about exactly this issue. The problem of not knowing in advance what fee you need to use in order to get your transaction confirmed would certainly be at least a little character intuitive to most people. I mean, normally when you go to send something to somebody else, you pretty much know what it's going to cost to send it. If you suddenly discover that actually what you thought it was going to cost is actually not that. It's actually three, four times that.

Then it's not the most obvious of models. So that one's sort of a little weird. Part of the reason I think that that concern exists, though, is because the network isn't always busy. And so there are times when you can get away with them. much lower fees, then nobody's used to paying those higher fees. So the volatility is the bigger problem more than anything else. I think if you actually had a network where the steady state was that the fees had to be reasonably large, then it would be unlikely that it would change

too much because the cost of running an even bigger stress test would start to go up. I mean, that's actually one of the problems is that for somebody who wants to run those sorts of stress environments is that as they push the fees for transactions higher and higher, they're also pushing the fees higher that they need to use in order to stress the network even more. So they're having to pay more to run that test. So certainly there is an aspect where if users aren't sure what the fees need to be, that's not a good situation. And that really comes down to the Bitcoin wallets. Are the wallets actually able to predict what the fees need to be and actually do that effectively. I mean, just one thing, one brief point on that, to me, it seems kind of obvious

that users should never have to think about the fee, right? I think the idea that users need to think like, oh, what fee do I need to pay? That's not a good idea. I mean, if we talk about like mainstream adoptions, people will, you know, people will just struggle with that. They won't know what decision to take. Exactly. And if you look at mainstream financial transactions like credit cards, it's invariably the recipient pays that fee, not the sender. I mean, it sometimes is hidden and sometimes it's an explicit charge. But generally speaking, the recipient's responsible for dealing with those fees. Now, I think the stress test to me demonstrated that there was a clear risk that Bitcoin could be attacked fairly simply and easily and perhaps, I mean, most certainly,

pretty cheaply, what did you take away from this? What did you learn? Or what do you think that we learned as the space? What do you think to learn? What do you think the space learned from this? So I think that the good things were that, you know, the network stayed up, it stayed reliable. That was actually, you know, in some respects it's surprising because quite often with, you know, engineering systems, they don't behave as well as you'd like them to. So that was actually a very positive thing. I think that the fact that the track, that it could be run for such an inexpensive cost was predictable. I think that that becomes harder as the fee market goes higher and fees go up. But the thing that I'm concerned about is that if you actually look at this and say, well,

you know, if we went from, say, a one megabyte block to a two megabyte or four megabyte, eight megabyte, whatever, you are just sort of kicking the can down the road a little bit. You're not actually really making it dramatically harder. You can still cause stress to the system. If everybody's expectations, the fees are very, very low, the cost. the stress test of the 20 megabyte block sizes are actually that high, but the potential consequences of having 20 megabyte blocks are sort of greater to the network. The risks are much higher because you have a much greater race of orphaning. You end up with potentially causing minors to change their incentives. And I think this is one of the areas that's worrying and one of the reasons

I think that more modeling is actually required to actually make some of these changes. And then, of course, there's the other question of hard forks are something that are problematic at best. And really, do you want to start the whole ball rolling in terms of saying, well, we're going to accept making hard fork changes to the network for unquantified reasons sometimes? Today's magic word is hash, H-A-S-H. Go to let's talk bitcoin.com to sign in, enter the magic word, and claim your part of the listener reward. So next year, 2016, the block halving is going to take place. How does that come in here?

So I think the block reward halving is going to be a very interesting time. If you look at the last one, so the one that's three years ago, very quickly after the block reward halving, the price of Bitcoins basically doubled against the US dollar. And so the effect was pretty much lost. So yes, there's anecdotal evidence that some miners decided that it was no longer cost effective to mine and they shut down their mining. But the price of the coins went up and so the miners were still getting a significant amount of money.

And so we saw that hash rate sort of race away and you saw the whole development of ASICs. I think the next one is more tricky. The probability that the coin price will double seems unlikely. And so then you get down to a question of what's actually incentivizing the miners. And I think that's unclear at the moment. I think beginning of last year, it was probably easier to predict that most mining was being done by sort of large, or relatively large cloud mining infrastructures. So you had, you know, tens of thousands of people who are paying in to actually do mining activity. I think it's less clear how much of the network is currently mining that way now.

I think you have some big pools that it's not absolutely certain quite what they're doing or quite how they're how they're going to operate. So the big question really comes down to costs. Mining ultimately is about, you know, how do you convert electricity into hashers and gain the largest share of the total network? And the question you have to ask is, well, who can actually afford to run if the mining reward sort of drops to half of what it is right now? because the fees right now are pretty much insignificant. So if the total reward in terms of US dollars halves, then it's going to be much more difficult for some of those mining operations to continue. And we've seen this already.

A large number of mining operations have shut down because it was no longer cost effective to mine. And then you're left with people who potentially have different incentives for mining. So their reward is not necessarily in terms of it costs them X amount of dollars to pay for the mining and they get sort of X plus one from doing it. You potentially have a situation in which some people can afford to mine a loss because they're using it as a means of actually moving currency around other ways. But I think the whole question of what's going to happen with the next reward halving is very unclear.

Yeah, in my view, it's not a, it doesn't sound like a good scenario it's coming up, right? Because either we sort of keep going as we are today, right? So, and let's say we increase the block size. So, of course, that means that there's a lot of space for blocks and because miners are competing, the fees will not go up essentially, right? And that's great from a user experience point of view. it's great from sort of also that story when it comes to adoption. You say Bitcoin is fast, Bitcoin is cheap.

Well, right, those are some of the main selling points that we've been talking about for years to everyone and people. That makes sense to people. And you can keep saying that then, right? But the problem is, right, then what you're saying, right? The hash rate drops in half. People turn off their manning hardware. And, well, someone you can rent that hardware to attack it or it just becomes, so you have a security problem, right? And then I guess the flip side is you now leave it at one megabyte.

You hope the transaction volume goes up and that then all of a sudden you start having people paying five million bitcoins fees and some lightning network and somehow. And it seems very unlikely to me that somehow this is going to work. And all of a sudden we will then have this fee market that compensate. Yeah. I don't see there's any hope realistically of a fee market that compensate. for the loss in mining rewards in 2016. I mean, there's just not enough time.

I mean, we're talking 12 months. And unless there is an absolutely phenomenal increase in transaction rates and a big change in the way people are prepared to fund those transactions, then I just can't see reaching sort of the fee level where you'd even get half of the loss in mining reward. I mean, we're talking about losing 12.5 bitcoins per block as a reward change.

and that would be a staggering amount of fees. And of course, if you actually make the block size larger, then that fee market will never exist. I mean, it's just, there's no incentive to pay any fee if you don't need to. So one of the things you wrote on your blog is one thing that is clear is there will have to be changes in the way which mining is funded. What sort of changes do you have in mind? Well, I don't really know.

I mean, I think that, I mean, if you look at, if you look at mining right now, there's a tendency to centralise towards wherever the energy costs is the lowest. Those energy costs could be either, you know, you put them somewhere where you've got cheap geothermal power, or it could be somewhere where you can get cheap subsidised power

and I guess that's happening in some of the Chinese operations right now. But it's, there's a tendency to do that. I think that the funding in the network is going to be an interesting problem overall, because, you know, when I wrote those comments, there was still a bit more time before you actually got to that reward halving as you get closer and closer to it, the potential to actually change that and actually move towards a fee-based model gets smaller and smaller. So I think that, you know, unless there's a big speculative bubble and you see the price of Bitcoin's double, there's going to be some significant amount of hash rate that has to be powered down. I mean, actually the one thing that's been quite surprising is that there hasn't been a significant reduction in hash rates over the last year while the prices were dropping. I mean, people have been turning mining off, but actually other people have clearly been turning more mining on.

And so there has been a steady hash rate increase, some of which is going to be technology driven. But it's actually surprising that there haven't been larger chunks of the network shut down. My guess is that people who've got the most efficient mining have kept it running so far. I don't find that so surprising at all. I mean, if people have the opportunity or think they have the opportunity to make money, they're obviously going to invest in. And I think most people don't understand some of these very complex models that we're discussing here and how they're going to play.

Well, you hope that the people are putting millions into hardware, do have some clue of what's going on. Well, may not be the case. May not be the case. I mean, I think in some cases you've certainly got some miners now who've been around for a while and they're making a long-term bet on the future of Bitcoin and the long-term bet on the fact that that investment is going to be worth something over the long haul, even if it's not necessarily making the money in the short term.

I think they're making a bet that they want to be doing this for other things in the future potentially. So I recently read an interview with John Tonys. And one of the things he said there, I thought was quite interesting. And he said that the sort of flip side of the idea of increasing the block size is changing the reward scheme and basically questioning whether Bitcoin really should be limited to 21 million and whether we really should have that halving off the block reward every four years. Do you think, I mean, to me it seems clear that politically there's just no way. is going to be touched. I think Bitcoin will crash and die and collapse before people will come to an agreement that that can be changed. I think so, so I mean, I think that the majority of

people who've been mining and are holding substantial amounts right now are basing the fact that, you know, they know that there's a declining amount of Bitcoin's coming out in the future. So they're assuming that what they invested in early on is going to be worth more over time because there's not new ones going to be coming on. becoming available. I think the bigger question really is about the nature of consensus within the Bitcoin network, consensus both at a protocol level, but also consensus about what the core devs do and how Bitcoin as a system is actually allowed to evolve.

So, I mean, if you start changing the reward structure, and I'd actually propose changes just sort of half-heartedly, really. but if you start making those sorts of changes and accepting the fact that hard forks are okay then where do you draw the line? No, is it okay to define that there's a hard fault that says, well, we're not going to reduce that the mine reward or we'll change it

so it's on a declining scale so it steadily degrades off. What other changes do you make at the same time? And that to me, I think, is one of the bigger problems with this whole discussion about the block size. What's the next one? You know, where do you go next? Do you say, well, in nine months time, if all of the large miners say, well, we're going to pull all our hashing on the day that the reward halves because neither is going to make any money anymore, then do you have a knee-jerk reaction that says, oh, we've got to change the minor reward structure? How should those decisions be reached?

Well, I'm not sure that they can be reached. I think part of the problem is that you have a design. It may not be perfect, but making fundamental changes to it is actually very, very tricky. And I think that there's also a longer-term question. Right now, we can actually have this discussion about making those sorts of, you know, radical changes because the number of nodes is relatively small, and the complexity of those nodes is they're basically, you know, servers and large PCs. So they're all basically the same. If you'd actually had a ragingly successful Bitcoin in which you'd had millions of embedded nodes that were all running full nodes as opposed to running anything else,

You couldn't even have this discussion because vast amounts of the network would never accept the change. And I think that that's actually one of the things that's unclear to me. We're sort of able to have a conversation about changing it simply because it's not been massively decentralized. As soon as it becomes massively decentralized, then you can't make those sorts of changes anymore. And then the only option really is to evolve the value of the system to something else. And so rather than saying you change the nature of what Bitcoin is right now, you have a Bitcoin 2 or a Bitcoin 3 or whatever you want to call it, where you have a mechanism to actually move the value from the existing system

into that next system, which has the different set of rules. And I guess that's a large amount of what the Blockstream guys are doing and the side chain's ideas is actually going to lead towards some of that sort of thinking. It will allow people to think about how can they move value from the existing system to something else. Part of the problem, I guess, is that everybody's a bit nervous about having to wait for those sorts of things to happen.

But arguably, that is the safest engineering course would be to say, we'll keep it the design as it is. We know that there are some things we don't like about it, but we're working on approaches that will allow us to actually migrate things in different ways. But of course, it seems like if you do that, you're always just sort of kicking the can down the road. right, I mean...

Certainly true. I mean, but what at least you do get the ability to do is actually run meaningful experiments to work out what the next thing should look like. The worry with a hard fork is that once you've made that, assuming that the majority of people actually accept it and that that hard fork sticks, then you now end up in a different situation. And suddenly you've gone from an unknown, well, in a situation where you've got a relatively known problem to a situation a bit further down the track where you have,

a completely unknown network and unknown incentives. And I think the worry there is that, you know, you need to do that analysis. And that's one of the things that I haven't seen right now, which is surprising. There hasn't been that analysis of what actually happens when you migrated that network. What does that actually look like? What does it mean? Certainly, it's going to delay the introduction of a fee market by having larger blocks, but does it actually introduce any vulnerabilities into the system?

Does it mean that potentially miners change their incentives? I mean, right now, for example, most mining pools are accepting reasonably large blocks. I mean, the vast majority of them will take at least a 750 kilowatt block out of the nominal one megabyte they can take, and some number take a full megabyte. But if you actually were to allow 20 megabyte blocks, potentially that puts some number of them at a huge disadvantage, and they may actually just decide to actually clamp the size of blocks they'll deal with to something much smaller because it actually makes more sense for them to run with a smaller size. So I think that again, this is where there's some analysis needs to be done. So how optimistic are you that all these things will somehow get resolved and there will be the good outcomes reached in all these different risks and scenarios and that we will be here talking in five years about the Bitcoin that's now being used by 500 million people? So I think that there are, there's actually some other fundamental challenges. certainly if you want to get to 500 million people using Bitcoin.

The big challenge there is that even if you make the block size 20 times what it is right now, you still can't do that many transactions on it. So you're going to have to see things move to other sort of payment channels. You're going to have to have side channels of one form or another that will allow things to happen off the main chain. And I think that's actually the thing that's potentially the most interesting. That's also where you get the potential for the most innovation. people can try ideas out and they can still use the main chain for actually moving the bulk of value around.

It loses some of the decentralized sort of ideas. It loses some of the thought that this is completely trustless in the sense that you don't have to trust anybody. But I think for the vast majority of people, if you wanted to actually use the Bitcoin adoption, that's going to have to happen anyway. And we've already seen it, the exchanges effectively do that anyway. So you have to put a set amount of trust in somebody to do things the right way. So I think the potential there is for those organizations to become more transparent. So if you have to trust them but you can verify that they've done the right thing, then that's actually a really interesting model.

So what you're saying here is that then Bitcoin becomes a settlement mechanism. and transaction So one example of this is a lightning network network of force which we've talked about on this show. Right. Do we get in a situation there? Because when I think about this, I have a difficult,

it's hard for me to see how that would happen so that we would have offshane transactions for the majority of transactions and Bitcoin being used as a settlement mechanism and at the same time being able to use Bitcoin to do small transactions if you wish, So if you want to be transparent and you want to be completely, well, somewhat anonymous and use Bitcoin for small transactions you could. But it seems unlikely that Bitcoin will be able to be used as a settlement mechanism and also for just day-to-day transactions if one wishes to do so.

Doesn't that go sort of against the corporate principles of what Bitcoin is? So I think that there's no reason why it has to be two different things. I think you certainly could have those micro-transactions being handled. via some of the mechanism and then being settled on the main chain. In fact, in some respects, it's easier if you do have that same unit being used for those transactions. But, I mean, certainly you have lost some of the decentralization at that point. So the question you have to ask is, from an end-user's perspective, do they actually care about the decentralization so much,

or do they care about the transparency of the system gives them? And does that allow them to do something in a way that they, wouldn't do with any previous sort of systems. And I think that the technology behind Bitcoin allows that. And I think there's a lot of people trying to work out how you can do those sorts of things. And you still do have the ability, there's nothing to stop you from actually making small payments using the main chain if you wish to do so. It's just the costs associated with it may be higher.

Yeah, but so that's, so what I mean is then I think that that would be undesirable. If you're stuck having to use some off-chain solution, because the cost for transactions on the Bitcoin blockchain on the main change is so high, I don't think that's a very good outcome for Bitcoin as a whole. So I think one of the bigger questions long term is, you know, there has to be a question of how do we align all the economic incentives. I think if Bitcoin had been able to stay on a very, very highly decentralized network where basically you were piggybacking on a vast number of people's compute power,

then it's very, very easy to sort of say, well, okay, everybody's carrying some of the cost for the network, and therefore everybody can take some benefit from it as well. It becomes more tricky when you end up with sort of the centralized mining infrastructures we have right now. I mean, there's a lot of centralized mining effectively happening right now, because those guys are controlling the security of the system, and they're also the ones who are extracting most of the value from it right now, because they're the ones taking the limited coins. They're the ones taking the transaction fees.

So I think there's some questions in terms of that mining infrastructure. But I think for Bitcoin as a whole, I don't actually see that there's a huge problem with having off-chain payment mechanisms. As long as those off-chain payment mechanisms can actually support the sorts of capabilities that the main chain itself supports. The only thing they potentially lose is some of the anonymity. which I suspect the vast majority of users isn't actually a huge deal. Or it could actually have more anonymity, you know, who knows, right?

Right. I mean, with Lightning Network potentially, it has more anonymity. One of the things you mentioned when we were talking about topics before is accepting unconfirmed transaction of that practice. What's a view in that? So I think that in the current network, zero confirmation transactions are, are worrying. I think that there was a certain amount of resistance to some of the recent software

developments that have tried to make them highly unreliable. But to a certain extent, the genie is out of the bottle there. Once people realize that zero confirmation transactions are risky, there will be... But they're not risky right now, right? Well, they're not risky as long as the mining calls actually don't accept replacement fees. And so, you know, double spends are something that would only be a problem if miners were prepared to say, oh, there's a bigger fee associated with this. So the replaced by fee, for example, would potentially allow a double spend or even a triple spend of the same coins.

And so anybody who's receiving those is at risk if they actually try and do anything without waiting for a sufficiently large number of confirmations. But I think that that's actually a natural phenomenon that's going to happen. I mean, we've already seen with the spontaneous forks that happened not that long ago, you have mining pools that aren't actually running full validation and full verification of transactions. So relying on the mining pools to actually do anything, I think, is questionable, at which point you have to argue that the whole idea of a zero confirmation transaction is actually itself highly debatable.

Wait, can you explain what you mean by mining pools or not doing full validation of transactions? So there were some number of the mining pools who were basically doing SPV verification. They weren't actually doing full verification because they could gain a small advantage in terms of how quickly they could process things. And so we had, I'll forget the exact date was a couple of weeks ago, where we had effectively a fork where certain mining pools weren't actually taking into account one of the new forms of transactions that could be. could be used. And so they were effectively creating a second chain. But they weren't

running full nodes. They were running a light-awake version of verification. So what are you suggesting for payment processors like BitPay and Coinbase and Bitnet, etc? Because

it's a pretty miserable user experience if you can't rely on So to a certain extent On compayment payments. Yeah, I mean, I think the trick is that with those guys who are acting as payment processes, to a certain extent, they have to take a judgment on the risk that's associated with doing it. So they might actually take a view that the risk is small,

and there may be some heuristics they can come up with that actually measure what the real risk is. And this is another reason I think statistics are really interesting for the Bitcoin network. If those guys have good analysis of the actual real double spend attempts and they can see any patterns to those and they can see what the actual risk is, they can actually take a judgment as to whether they want to accept zero confirmation transactions or not. I think that the bigger risk is for anybody in general taking those zero comp transactions. I mean, commercial entities can obviously make their own judgments about whether they'll accept them or not. I mean, the other risk with zero confirmation transactions is really that

if you look at the original Satoshi white paper, there's an analysis there of what would actually be a safe number of transaction confirmations to get a certain confidence level about whether something can be reversed or not. And this is another area that, again, I think the statistics can help with where if your largest miner controls less than 10% of the mining hash rate and there's no collusion between miners, then sort of six, confirmations is great because you get a very, very low probability of six confirmations being overturned. If you have a situation where a minor is much larger, then the risk is that you need a dramatically large number of confirmations to avoid or to get the same level of confidence that those transactions can't be overturned. And so I think that that's actually another area where potentially when you're talking about low confirmation rates, not necessarily zero, but low confirmation numbers, you actually need to understand how centralized the network potentially has become. The more decentralized it is, the smaller numbers you actually need.

Today's show is brought you by our friends at ShapeShift. ShapeShift is the fast and easy way to trade all coins. They now support over 40 of the most popular altcoins, including flooring coin, I.O. coin, positron, infinite coin, and so many others. When you want to trade altcoins, forget about using exchange. That's so 2013, man. Just go to Shapeshift.com.

and get it done on less than one minute with no counter signup required. The currency conversion tool makes trading all coins just about as easy as using Google Translate. And we've also got some use for you. Shapeshift, just put out an iPhone app. Here is my iPhone and here's the app for the iPhone. It's an iPhone app for the iPhone. And you'd think that if you would make an iPhone app that then you'd use to exchange altcoins, well, you're going to have to go to one wallet to send it to another to get the address

and the ShapeShift to like execute it. that would be a pain and a lot of copying and pacing and stuff. Well, actually, it wasn't. It was super easy. I put a little bit of Bitcoin in to get some gems in my James wallet, and it literally took about 30 seconds. So they've done an amazing job at that. So if you have an iPhone, do you sell a favor and download the app. If you are someone like Sebastian who uses Android, then do you sell a favor, throw away the Android, get an iPhone, and then download the app. we would like to thank our friends at ShapeShift for supporting the show and go to Shapeshift.io, give it a try.

So you work at Piranova, so you've been involved with mining at that company and also worked in ship manufacturing. Can you talk to us about what you see as the future of mining? So I think mining is actually very interesting. mining intrinsically is something that's going to burn a lot of energy and so it has some some implications because of that that i don't know of anybody who's yet come up with a good proposal of how you avoid that as a problem i think there are some intrinsic challenges though for mining which is that um you you really want to find a way where in some respects where you can actually avoid it sort of running away and just consuming vast amounts of

energy. That would be a sort of a desirable thing. I don't think anybody has a good solution to that yet, but I think that trying to find solutions to that would be extraordinarily useful. And if we can find a way of doing that, that would be a huge deal. I think the other challenge is if you want a network, which is largely based on having no centralized point of trust within it, there has to be some work done in terms of decentralising the mining itself. And not actually the hashing. The hashing is the least interesting part of the decentralization. There has to be a big push towards decentralizing the transaction selection and transaction processing.

So this is really down to the way the mining pools work, for example. We really need to get to a stage where there are no large mining pools. And that may not be something that's possible within the current Bitcoin network. It may be possible. I mean, certainly it's possible with a hard fork. But I think hard forks themselves are problematic. But I think that's another area that needs to be looked at. into because if you could have a thousand pools, each of which control 0.1% of the network

and could be demonstrably not actually colluding in some respect. You have a much better decentralized network. The worry right now is that it's very easy for one mine is again a very, very large share of the total hash rate. And it'd be very difficult for anybody to necessarily know that was happening. One mining-related story that has always gotten a lot of attention, and it sounds really exciting. It has been 21 Inc. They've raised a ton of money.

And from what they've told, it sounds like they will put miners in toasters and toilet seats and shoes, perhaps. Who knows? But no, basically embedded in devices, that has never made any sense to me. I don't understand it. Right. And you actually worked at Qualcomm, which is an investor in 21 Inc. What is going on with that? So I have no idea what Qualcomm were doing.

I actually know a bunch of people at Qualcomm who were really interested in Bitcoin for various reasons. But none of them, as far as I'm aware, were involved in those discussions either. The idea of actually embedding the mining into lots of devices is great in one respect, which is that if you could actually couple that with decentralising the mining pools as well, then it would give you a tremendous amount of decentralisation. But when you actually start to run the numbers,

I don't think it adds up from an economic perspective and I don't think it adds up from an energy perspective. The energy in people's homes is usually pretty expensive. It's not like the hashing would be free. It's not like a byproduct of something else that's happening already. So I think that, you know, from a global energy perspective, that doesn't make any sense. And I can't see how the model where 21 provide the chips to go into all these embedded

things, I can't see how that would work economically in anybody's interests anyway, apart from 21 potentially, because in order for them to actually gain any transaction fees or gain any mining fees from that, they're going to act as a huge pool. so 21 suddenly become the biggest pool in the network, which doesn't seem like it's actually decentralizing anything. They've decentralized the hashing and centralized the control of that hashing. So that doesn't make sense either. I mean, it basically sounds like their idea is to run a massive botnet.

Right. Effectively, that's what it seems like. I mean, it also makes very little sense in terms of power and the devices that are actually naturally connected. One of the problems is that you need devices that are actually going to be connected well, and you need that to provide some utility to the people who are going to be powering that hashing.

So it makes no sense to me to say you're going to be collecting a few hundreds of toisies a day from your toaster, because what are you going to do with them? There's nothing useful to do with that sort of smaller quantity. And also your toaster's not networked right now. So from an end user's perspective, there's a huge problem, which is if you suddenly have to network your toaster, well, now you've gone from your toaster, which is pretty unintelligent, you've now got to have a CPU in there, you've got to have a mining chip in there, you've got to design this thing so it can, you know, connect to something else. You've got to put Wi-Fi in there. And then I don't know if anybody's actually spent a lot of time with Wi-Fi, but I used to work on Wi-Fi design.

and there's an awful lot of Wi-Fi access points out there that will fall over if you put too many clients on them. So your toaster could be one device too far and suddenly your network stops working. That's not a great end-user experience. You put them in your refrigerator, or the refrigerator is usually somewhere inaccessible. Probably doesn't have great Wi-Fi coverage either.

So these sorts of things don't really make a lot of sense. And as a thought experiment last year, I was actually thinking, does it make sense to put these things into something like a router, as I used to design routers. On paper, that's the natural place to put sort of a hashing engine if you're going to do one. But then there comes a cost problem, which is those devices are designed with a certain thermal footprint. They're not designed to go above a certain temperature. The enclosures are designed to run without fans.

As soon as you start putting any sort of high-energy-consuming device in there, those thermal devices, those thermal characteristics go out the window, you need more expensive enclosures and it suddenly starts to get more expensive to run. So not only does it cost you more energy, but it costs you more up front as well. So it just seems like it doesn't work as a concept. So is there, there's no way you can think of that their plans make any sense? I keep trying. I haven't found one yet, but good luck to them if they can find a way of doing it. Well, I mean, one interesting area to perhaps consider would be central heating. If you could have a central heating component in a house that then distributes a heat through ventilation system and you get some mining reward from that.

Would that be a good idea? Sure. If you have something which already takes electricity and needs to generate heat, then that would make some sense. I think another one I heard somebody suggested was like greenhouse heating and things like that for growing things. those make sense because you're already turning electricity into heat in the first place. They're not necessarily the most, I mean, A6 aren't the most efficient way of doing that, but at least there is some value there because the intent was originally the burn energy to actually generate heat in the first place.

But I think the question then becomes, is it the most efficient way? And would people sign up for that for large-scale heating? And of course, even the issue, of course, is let's say people do sign up on a large scale for something like that, Well, I mean, the thing is with Bitcoin, of course, that would mean the difficulty rises and then the revenues would drop again, right? So there seems to be like, no way. This makes any sense. Exactly. Unless, of course, you were to do a hard fork and save the 21 million cap disappears and you could now start having, you know, keep collecting bitcoins forever.

So I want to ask you regarding the ASIC chips. So I believe we're getting towards the threshold in terms of what those chips are going to be able. to produce in hash rates, like physical nanometers. I'm not especially knowledgeable about chip manufacturing. But you mentioned in a talk that we were coming to that threshold. And also something interesting, you mentioned that in terms of energy savings, as we were getting towards closer to that threshold, the energy savings began to drop. So does that mean that as chips continue to evolve, there will be more expensive.

expensive to run. And secondly, what comes after A6? What comes after we've reached that threshold and can't go any further? I think lots of people would like to know what comes after A6. Intel and everybody else. Yeah, the, if you actually look at Bitcoin, I think I mentioned earlier, Bitcoin sort of raced through five generations of technology in five years, which was absolutely astonishing. And you got to the stage where you're 28 nanometre or 20 nanometer of devices, which are pretty much state of the art for what most people are using. There are a few now in 16 and I guess 14 as well. But those are, at the moment, those are leading edge devices.

I think there has been some discussion of some research now that would take us down under 10. But, I mean, the amount of time to go through those generational gaps is getting bigger. It's getting harder and harder for the scientists to actually come up with ways of actually shrinking things. more than they have already. So every so often, you know, people say, oh, we're not going to get any smaller. And somebody always finds a way of doing it, but it's taking longer each time. So Bitcoin has reached the point where the A6s are pretty close to the best they can do in terms of the geometry. I mean, yes, the geometries will continue to shrink. And, you know, as those technologies become available, you'll be able to use those for hashing as well. And that will certainly improve the power

efficiency for that generation. But you're then dictating the pace at which you can do things based on what the semiconductor fabs can do. And can they actually produce anything smaller or more power efficient? So the other problem is that as these things get smaller, it used to be that you'd sort of get this sort of square log gain where as everything halved in size, the power efficiency would go as a square of that.

That's not happening so much now. So that's actually pretty tricky. And the other thing is, if you look at a Bitcoin ASIC, it's doing a fairly simplistic task, There's nothing too complex with it. It's pretty well defined. There are some optimizations. I think most of them are already being taken.

You could certainly do some custom layout that's going to improve the power efficiency, and I'm sure the guys at 21 are looking at this and others. So you can get some improvements there, but there's a limit to how far you can take that. Because what a Bitcoin ASIC does is already highly specialized. So I think that you will see Bitcoin at this point. the hashing will basically follow the same trends as cell phone processes and other CPUs. You're not going to see dramatic gains. And so this is one of the things I predicted last year was that there would be a slowdown in the hash rate

where the increases would now be pretty much dominated by just the pace of general technology and the replacement of previous generations of hardware. So Bitcoin 86 and these come sort of subject to, the same problems as the regular ship manufacturing industry. Yeah, they've hit the limits of the physics right now. Okay. So Dave, we're kind of at the end of our show.

We've gone pretty long. Before we wrap up, so you work for Pinova. Is there something like a few sentences you can say about the company and maybe if people want to learn more about it, where they can go? Sure. So, I mean, obviously, we start out in the Bitcoin space.

We firmly believe that the technology that's used in cryptocurrencies and crypto ledgers is extraordinarily useful. We have our own take on that, and so we've been working on some stuff that I guess we'll be announcing more in the next couple of months. I think, unlike a few other people, we decided we didn't want to talk about it until we finished building it, or at least building a good working prototype. But we're looking at building things on a different scale. So we're looking at how you can get the same sorts of guarantees and the same sorts of characteristics that you get with something like the Bitcoin network in terms of transparency and in terms of the ability for somebody to verify that it's correct,

but apply those for other applications. So we're looking at building sort of vast blockchain-like things but without being exactly the same as a blockchain. chain. So we have stuff up on our website now. In fact, we would just change the websites like there's a little bit more on there than there was before. And I guess we'll be announcing more in the next couple of months. Cool. Thanks so much. And of course, we'll have links to this and also to your blog and to some of the articles we read in preparation all in the show notes. So Dave, thanks so much for coming on. Okay, thank you very much.

So also some announcements. First of all, I signed job. So I'm going to be sort of a full-time employee. I think is the time when this podcast comes out. So I would be doing business development. So I'll be the head of business development at ERIS Industries. And some of you may know that because we've had them on the podcast before Press and Burn a while ago. And well, I won't talk too much about it now. I'm really excited about it. I think it's a great company and a great technology. But I'm sure I will talk about it more at some later stage. Regarding epicenter,

you know, we'll keep doing this in the same way, so there shouldn't be any changes there. And another thing is that, so we have decided, since you guys have not been very, not been very good listeners in terms of leaving iTunes reviews,

we've decided to resort to the low act of bribing. Guerrilla tactics. So we have some of these T-shirts. So what we've decided is that once a week, if someone leaves a reviews, we will give you, we will send you a one of a T-shirt. And it can be a negative review or it can be positive view. It doesn't matter. You can write whatever you want.

But, yeah. So if you want to do that, just send us an email at show at Epicenter Bitcoin with the sort of link or reference review and we'll send you a T-shirt. I believe we'll draw one a week. So it just like droves if you're doing it. Yeah. Of course, since we have no way of identifying you on iTunes, it's important that you send us an email to let us know that you've written a review so that we can identify you.

So that's show at epsenterbiccone.com. And by the way, congratulations on your new job, Brian. Yeah, thanks. I'm excited. Yeah, so thanks so much for joining us. So we put out new episodes of Epson and Bitcoin every Monday. You can subscribe to the show in iTunes, SoundCloud, or of course, in whatever podcast app you use. and you can watch the videos on YouTube.com slash FC

and yeah, that's it. So thanks so much and we'll be back next week.