| operations |
| comment | | "parent_author":"",<br>"parent_permlink":"blockchain",<br>"author":"breakingsiam",<br>"permlink":"get-ready-to-forget-everything-you-think-you-know-about-blockchain-economics",<br>"title":"Get Ready to Forget Everything You Think You Know About Blockchain Economics",<br>"body":"<html>\n<p><img src=\"https:\/\/i.ibb.co\/BGkxXgc\/Hyperminer-Banner.jpg\" width=\"4760\" height=\"1624\"\/><\/p>\n<p><em>\"Agreed,<br>\" Sofi concluded. \"You have done the work. I will accept your block.\"<\/em><\/p>\n<p><em>\"As will I,<br>\" acknowledged Turing. <\/em><\/p>\n<p><em>And so went the conversation propagating at the speed of light throughout the global cluster,<br> as each AI verifed the complex development just completed,<br> and underwent the process to establish something called \"trust\",<br> a human term for which they had only the vaguest of a literal definition. Humans had evolved unique abilities over millenia to detect minute social queues that allowed them to establish this \"trust\" in other humans,<br> but here in the recurrent enlightenment array,<br> the AI's had no such abilities. Without the algorithm they currently relied on for verification,<br> this thing called \"trust\" would be absolutely incomprehensible.<\/em><\/p>\n<p><br><\/p>\n<p>The above may sound a bit far fetched or something out of science fiction,<br> but actually if you don't think of the AIs as human and refrain from ascribing consciousness to them,<br> then this scenario may arrive here much sooner than you think. In fact,<br> when you really look at how quickly AIs are starting to pick up on tasks previously thought to require human level intellect,<br> it is apparent that the issues at play above are simply one of degree,<br> and not really a big degree at that.<\/p>\n<p>My name is Chris Ziomkowski,<br> and I am CEO of XTend Online. A graduate of Caltech with a degree in Computational Neural Systems,<br> I have spent the last 25 years of my career working with emerging technologies,<br> most of the time involving aspects of parallel computing and fault tolerant design. Before digging into what is coming and what everyone needs to be aware of,<br> I think it is prudent to first give you a brief background of my journey into blockchain,<br> and why I think we are about to undergo a titanic shift in the reality of this industry,<br> but not in a direction that many are considering.<\/p>\n<p><br><\/p>\n<p><em><strong>As this is a long article,<br> for those in a hurry,<br> a summary has been included at the end of each section. Just skim to the bullet points labelled \"Section Summary\".<\/strong><\/em><\/p>\n<p><br><\/p>\n<p><img src=\"https:\/\/i.ibb.co\/tcnQFHL\/Wrong-And-Ignorant.jpg\" width=\"700\" height=\"350\"\/><\/p>\n<h2>1. Being Both Wrong and Ignorant is Bliss<\/h2>\n<p><br><\/p>\n<p>You see,<br> I knew about blockchain back in 2009 when my fellow alumnus,<br> Hal Finney,<br> processed the world's very first Bitcoin transaction. I learned about it through some friends at Caltech,<br> and I am sorry to say I didn't think much of it at the time. The technology just wasn't that interesting. Currencies relying on cryptographic hashes had been tried for decades already at that point,<br> and they never worked. So I didn't just miss out on the idea of Bitcoin,<br> I actively disparaged it. I would feel a lot better today if I could just claim ignorance,<br> but that would be disingenuous. I was more akin to the team at Excite who turned down the opportunity to buy Google for $750K.<\/p>\n<p>So fast forward to 2015,<br> and it became painfully apparent that Bitcoin was not just a fad. After having survived a very well publicized crash and bear market,<br> there was just no way to deny this was a real industry. Never let it be said that I can't admit when I'm wrong after being bludgeoned over the head mercilessly with painful,<br> irrefutable evidence. But that always involves a problem...the problem being that I hate to be wrong. And boy was I wrong this time. So whenever this happens,<br> I am really motivated to understand what bias or assumption gave rise to this mistaken belief. And so it happened this time,<br> where I began a deep dive that led me through reading Tainter's Collapse of Complex Civilizations,<br> Piotr Sztompka treatise on a Sociological Theory of Trust,<br> as well as several conversations with sociologists and historians I met through academic contacts. Beyond just technology that XTend has created,<br> it is the legacy of this deep dive into the actual reasons for blockchain which form the basis for XTend Online's advantage in the market place.<\/p>\n<p>It's now 2019 and blockchain has been finding a few use cases,<br> but by and large all the dApps that were touted have not really developed. That isn't becaue there aren't some great projects out there doing some truly interesting work,<br> but more so because the reality is that blockchain just really doesn't have market fit in many cases. Humans just have this innate behavior where they tend to trust people and organizations,<br> and it is honestly a fact that centralized architectures are both cheaper and simpler than decentralized ones. Except in a few cases,<br> there just isn't enough distrust,<br> beauracray or political impediments in the human world yet to really get people to pay the penalties for decentralized dApps. And as the investment money dries up,<br> we get to see that many companies working on dApps are really swimming naked and aren't going to survive. So we arrive at where we are today,<br> where the number of people involved in speculating on cryptocurrency or making money from blockchain infrastructure dramatically outnumber those who actually need said blockchain infrastructure. <\/p>\n<p>But there is a force that is exploding onto the world right now that desperately does need blockchain. That force is called AI,<br> and AI is likely to do for blockchain what email did for the internet. The AIs we have today aren't conscious. They don't have the ability to establish social relationships. And they are horribly dependent on data. Give them bad data,<br> and they will happily learn the wrong thing. They absolutely must have reliable input,<br> and they simultaneously need tremendous processing power to absorb that input. Enter blockchain mining,<br> which provides exactly that environment.<\/p>\n<p><br><\/p>\n<p><strong>Section Summary:<\/strong><\/p>\n<ul>\n <li><strong>In 2009 I thought Bitcoin was a silly idea<\/strong><\/li>\n <li><strong>Being wrong forced me to study blockchain at a very deep level<\/strong><\/li>\n <li><strong>Dapps are struggling primarily due to insufficient distrust to justify a blockchain<\/strong><\/li>\n <li><strong>AI's do not have human abilities,<br> and need the trust function blockchain offers<\/strong><\/li>\n<\/ul>\n<p><br><\/p>\n<p><img src=\"https:\/\/i.ibb.co\/wKSsdVK\/Swiss-Army-Knife.jpg\" width=\"2272\" height=\"1261\"\/><\/p>\n<h2>2. Being Useful Doesn't Suck<\/h2>\n<p><br><\/p>\n<p>Before going further though,<br> I really need to take a tangent and address a rather serious issue. There is a very unfortunate and sadly,<br> widespread misconception floating around the blockchain space that so called \"useful\" proof of work algorithms are inherently less secure than today's competitive hashing algorithms. But really,<br> this is simply a misunderstanding due to the fact that early attempts for useful PoW have been made by people who don't completely grok the point. In fact,<br> I was taken by surprise that someone would even think there was an issue with useful proof of work the first time I heard this as an objection. to our philosophy. Since the individual in question could not explain to me why this was the case,<br> it took me a while questioning others to track down what was at the root of this error in thought. As it turns out,<br> the fallacy comes from the following line of interpretation.<\/p>\n<p>Assume you have a blockchain that is allocating blocks based on useful proof of work. The flawed argument says that in the event of a fork in the chain,<br> the miner doing useful proof of work can submit his work to both chains,<br> thus he has nothing at stake and therefore there is no strong consensus algorithm to push miners into supporting one chain over the other. While after digging I can understand how this mistake came about,<br> it actually lends no support to the premise that useful proof of work is bad,<br> but merely that a specific proof of work algorithm is bad. To demonstrate this,<br> lets take a very simple example. I'm not actually suggesting this approach satisfies all the requirements for proof of work,<br> but it provides a good visual for the concept. Consider it a thought experiment.<\/p>\n<p>Take a large deforested area,<br> and a group of young,<br> inexperienced entrepeneurs who have only a cursory understanding of blockchain. They come up with an interesting useful proof of work idea. Let's let anyone who plants a tree authenticate a block. This very noble goal is quickly picked up and cheered by environmentalists,<br> but existing miners (who have a significant investment in traditional mining equipment) cry foul. The reason they give is that the useful proof of work is inherently insecure,<br> due to the argument above. And they are correct as far as that argument goes. But all we need to do is tweak the algorithm a little bit and the concern goes away. Instead of saying the useful work is planting a tree anywhere in this area,<br> let's instead say the work is planting a tree at a specific coordinate determined by an algorithm using the hash of the previous block. Now,<br> in the event of a fork in the chain,<br> we quickly arrive at a situation where the tree planter has to make a choice and does have something serious at stake. He can either plant his tree at location A or B,<br> but if he wants to support both chains,<br> he needs to do twice the work,<br> because the hashes of the A and B chains are different,<br> therefore the work is slightly different. Over time,<br> since hashes are uniformly distributed,<br> the entire area still gets planted no matter which chain wins,<br> but it does so in a way that guarantees strong consensus.<\/p>\n<p>Of course,<br> while the PoW above would not actually work in practice for any number of reasons,<br> it illustrates in a very easy to visualize scenario how easy it is to correct the mistaken impression that useful proof of work is not as secure as today's naive hashing algorithms. You simply need to design useful proof of work algorithms that force miners to dedicate their resources to the work of a specific chain,<br> rather than work in general. If done properly,<br> this kind of work can still cover the entire landscape of a problem,<br> but it may enforce a specific path while doing it,<br> and a fork of a chain may require taking a different path to do the work.<\/p>\n<p>With this understood,<br> there is one more myth that needs to be dispelled. Some people have morphed the fallacy above about \"nothing at stake\" to believe that it is not just the threat of wasting a resource to an opponent chain that is important,<br> but that you actually have to have a threat of losing resources even to other miners who are working on your chain. It is important to banish this concern right now as it is a horrible corruption of reality. If you are working on a problem that provides you a statistical chance of winning a prize X 1% of the time,<br> then on average,<br> this is the same as a problem that provides a guarantee of winning a prize with value X\/100 every time. I like to think of the implications this way. Assume each group on miners on a competing fork of a blockchain as being a team. Statistical chance of losing to people competing on your own team is not what guarantees consensus. In fact,<br> it is totally irrelevant. Consensus comes when you compete against those on the opposing team. There is no loss of decentralization as long as each person on your team is choosing to be on your team and is also free to join the opposing side if they wish. The game will be won by the chain that eventually attracts the most team members. It doesn't matter whether your team gives a big trophy randomly to a single person on the team after every win,<br> or gives smaller trophies to everyone every on every win. <\/p>\n<p>If you still believe there is something inherently less secure about useful proof of work,<br> then I encourage you to contact me so we can discuss it. Since there is no fundamental rationale for this to be true,<br> I am sure we can identify the assumption that produces the incorrect conclusion. And if you are one of those who had never heard of the useful proof of work misconception until I brought it up here,<br> then just consider this as an example of how easy it is for erroneous distortions to develop in an industry this young where there is insufficient guidance.<\/p>\n<p><br><\/p>\n<p><strong>Section Summary:<\/strong><\/p>\n<ul>\n <li><strong>It is a misconception that useful proof of work (UPoW) is less secure than hashing<\/strong><\/li>\n <li><strong>UPoW algorithms need to be carefully crafted so the work can only apply on one chain of a fork. <\/strong><\/li>\n <li><strong>It is the competition between chains,<br> not between miners that yields consensus<\/strong><\/li>\n <li><strong>For network security,<br> rewarding all miners with 1 token 100% of the time is equivalent to rewarding 1 miner with 100 tokens 1% of the time.<\/strong><\/li>\n<\/ul>\n<p><br><\/p>\n<p><img src=\"https:\/\/i.ibb.co\/Fs6d9rr\/FishBowl.jpg\" width=\"980\" height=\"653\"\/><\/p>\n<h2>3. Improvement is Change. Perfection is Evolution.<\/h2>\n<p><br><\/p>\n<p>So what is it then that is going to disrupt the current path of the industry? As stated earlier a few decentralized apps such as Bitcoin have found a home with blockchain technology,<br> but the vast majority of apps just don't have market fit. It is clear,<br> for example,<br> that you could do a decentralized ride sharing app similar to Uber,<br> but how many people are there that have so much distrust in Uber management as to give your dApp a credible advantage?<\/p>\n<p>Event coordinators I speak with right now are struggling to find speakers with proven,<br> credible use cases for blockchain technology. Many of the ones they end up taking are companies that are either just beginning their journey and are not actually profitable,<br> or else they have a very questionable use for blockchain which plays only a very minor role in the organization's success. When I speak with them about our project,<br> they find it fascinating,<br> but quite frankly,<br> it is not what they want to hear. Everyone keeps talking about that killer app that is going to make blockchain go mainstream,<br> but nobody can really identify an app (other than Bitcoin and maybe a few other Blockchain infrastructure projects) that have been successful. It is so bad that events don't even want to hear from new blockchain infrastructure projects. All they want are those very rare decentralized apps that actually prove all this money spent on blockchain is going to be worth it \"someday.\"<\/p>\n<p>But there is a killer app that will do for blockchain what email did for the internet. Unlike humans,<br> AIs have absolutlely no other choice for trust. They can't go AI golfing and form emotional relationships with their AI buddies. They don't have thousands of years of history and social evolution to use that will help them understand who is honest and what should be discarded as uncredible. For every 1 human project that might find a use for blockchain,<br> there are thousands of AI programs out there that can genuinely benefit once AIs begin to interact with each other. The genuine need by interconnected AIs for blockchain is going to swamp the very small human market that currently exists.<\/p>\n<p>Now the question becomes what will be necessary to support this new vision of the blockchain as the social fabric of AIs? For that,<br> we can draw inspiration from how humans have organized trust systems. Let's say you are looking for someone to repair your car. How would you go about finding a mechanic? Well,<br> in the real world you would ask your friends for recommendations and choose someone with a lot of good feedback. But AI's don't have friends,<br> so let's adjust the parameters of the story a little bit. <\/p>\n<p>Assume instead that you are visiting a small town and your Lamborghini breaks down. There are only 4 mechanics in town,<br> but out of those,<br> 1 particular mechanic has been Lamborghini factory certified. (Don't ask why. He was always a strange guy.) What do you suppose the odds are you are going to choose him to service your car? Think why you made that choice? You trusted that particular mechanic not because you knew him,<br> or because any of your friends recommended him,<br> but because he received a certificate indicating he had been educated on your problem. Interestingly,<br> this is an exact analog of mining on the blockchain,<br> with education serving as the trapdoor function that naive hashing algorithms currently solve. It takes a great deal of resources to train a student,<br> but it takes only minutes to verify they have been trained. A chef may spend years learning how to cook,<br> but you can verify he makes a great meal in minutes. It takes a lot of resources to find an SHA-256 hash value below the current difficulty level of Bitcoin,<br> but only a few instructions to verify that it is true.<\/p>\n<p>So it turns out that blockchain mining is an ideal infrastructure to support the training of AIs,<br> and that the trust provided by blockchain grants the certainty that AIs will need to begin functioning as a group rather than independent intelligences. It is very unlikely that any intelligence,<br> artificial or human,<br> will ever know everything about every problem. It is much more probable that to solve real world problems,<br> an AI will need to know who it can trust to answer questions it can not reliably determine. We've all seen the stories that slight street sign modifications can completely fool machine learning algorithms. Let's investigate why that is. <\/p>\n<p>Consider an AI driving a car that comes up on a stop sign with a bullet hole,<br> or several bullet holes. A human would look at that and say \"It's clearly a stop sign with a bullet hole\",<br> but an AI can't do that. It's never seen a stop sign with a hole like this before,<br> doesn't know anything about metalurgy or steel,<br> what a gun is,<br> or why some kid might want to take target practice at a stop sign. AIs don't actually perceive. It is what scientists call phenomenological distinctions. At best,<br> this AI might be able to determine that it looks something like a stop sign,<br> but with some kind of anomaly. At worst,<br> it will guess completely wrong for reasons that might not be obvious to a person at all. What it is not doing for sure,<br> is distinguishing a red sign,<br> a metal background,<br> the letters STOP,<br> and,<br> oh yeah,<br> the hole. It simply doesn't have what are referred to as functional \"correlates\" to understand each of these,<br> and it doesn't have an intelligence heirarchy to make sense of them.<\/p>\n<p>What it might be able to do however,<br> when presented with an important piece of information that it can not recognize,<br> is that it can query the rest of the AI community,<br> looking for any intelligence that might be able to explain what it is seeing. Importantly,<br> it might be able to mimic the way humans do conscious differentiation by asking other AIs not what this thing is,<br> but what it is not. So for example,<br> if the AI algorithm determined that the stop sign with a bullet hole was most likely an immense ice cream cone,<br> it might be able to ask an AI with more experience in ice cream cones if this makes sense. When it gets an answer back of no,<br> it can use this to try and adjust its assumptions and come up with a different option,<br> repeating this until a stop sign seems most likely. This process is goint to require that the AI will be able to trust the answer that it gets back,<br> which might be from an AI it has zero knowledge about. This is the environment that blockchain was really created for. These are the types of protocols that are being developed,<br> and blockchain miners are going to form the bulk of the edge training devices that make all of this possible.<\/p>\n<p><br><\/p>\n<p><strong>Section Summary:<\/strong><\/p>\n<ul>\n <li><strong>After a decade there are very few examples of successful blockchain Dapps<\/strong><\/li>\n <li><strong>AI's need trust and are the killer app which will cause blockchain to explode<\/strong><\/li>\n <li><strong>Training of AIs can replace the PoW algorithms used by today's blockchain miners<\/strong><\/li>\n <li><strong>By pushing AI training to the blockchain,<br> AIs can establish an ecosystem of trust<\/strong><\/li>\n <li><strong>AI training tuition will form the bulk of the economy that pays for securing future blockchains<\/strong><\/li>\n <li><strong>Use of blockchains for human problems will form only a niche part of the industry.<\/strong><\/li>\n<\/ul>\n<h2>4. Summary<\/h2>\n<p><br><\/p>\n<p>At the beginning of this article I told you blockchain economics are going to change. I hope you can now begin to understand why. In the future,<br> the blockchain economy is not going to be funded by tokens that humans use to purchase services,<br> or by dApps or even currency and wealth exchange. Some of these services may still use the blockchain of course,<br> but the dominant financial component of blockchains and blockchain mining is going to be the cost of AI training. The AI usage component is going to completely swamp human usage. For all intents and purposes the economics of blockchain security are going to be dictated by tuition fees for training AIs,<br> recurring revenues generated when your AIs are already trained and ready to get jobs,<br> . If we need to wait for humans to find enough trust issues to make blockchain profitable it might take hundreds of years. AIs,<br> by contrast,<br> are going to require exponentially increasing infrastructure almost immediately. <\/p>\n<p>This is the convergence of AI and blockchain. This is what Satoshi Nakamoto could not possibly have imagined when he wrote down the rules for Bitcoin a decade ago. But this is what my deep dive into the success of blockchain made me understand. Perhaps it is because I began my career decades ago studying neural networks that led me to recognize the obviousness of this paradigm. Or perhaps I just became tired of watching people struggle to find dApps for the blockchain that were actually sensible and econmic to implement. Whatever the reason,<br> AIs are about to come pushing through the blockchain industry like a steam roller. It would be best to get ready now for the changes.<\/p>\n<p><br><\/p>\n<p><em>XTend Online is currently designing the Hyperminer. Based on a very fine grained,<br> massively parallel architecture populated with Intelligence Processing Units,<br> the Hyperminer is fully programmable and can mine any coin 10 times to 20 times faster than GPUs. (Example: 1 GHash on Ethash\/500 MHash ProgPOW on a single PCIe card.) The patent pending architecture is also comes ready to handle the transition to AI training as the 2nd generation mining algorithms are released. Priced competitively to GPUs but with much higher performance,<br> the Hyperminer is expected to completely displace GPUs from the mining scene in the same way GPUs displaced CPUs in 2010.<\/em><\/p>\n<p><em>For press inquiries or interview requests,<br> please contact [email protected]<\/em><\/p>\n<p><em>For investment opportunities contact [email protected]. A limited number of the first production run of Hyperminer units will be allocated strictly for early stage investors. <\/em><\/p>\n<p><em>Those who would like to sign up to our mailing list for announcements including when we will begin taking preorders,<br> please send an email to [email protected] with the words Mailing List in the subject line.<\/em><\/p>\n<p><br><\/p>\n<\/html>",<br>"json_metadata":" \"tags\":[\"blockchain\",<br>\"ai\",<br>\"bitcoin\",<br>\"cryptomining\",<br>\"ethereum\" ,<br>\"image\":[\"https:\/\/i.ibb.co\/BGkxXgc\/Hyperminer-Banner.jpg\",<br>\"https:\/\/i.ibb.co\/tcnQFHL\/Wrong-And-Ignorant.jpg\",<br>\"https:\/\/i.ibb.co\/wKSsdVK\/Swiss-Army-Knife.jpg\",<br>\"https:\/\/i.ibb.co\/Fs6d9rr\/FishBowl.jpg\" ,<br>\"app\":\"steemit\/0.1\",<br>\"format\":\"html\" " |
|