Introduction. Welcome to Part IV of my (that is, 7/.V) “In the Hands of a Few”(ITHOAF) article series. Here I show how we get “fooled by randomness” – with a slight abuse of Nassim N. Taleb’s terminology. If you search the Internet for “4,448,419” you will find the U.S. patent “Electronic Gaming Device Utilizing a Random Number Generator for Selecting the Reel Stop Positions” filed on February 24, 1982, by Norwegian Inge S. Telnaes. The number 4,448,419 is not important. But randomness is: the proposed pseudo random number generator makes a slot machine mathematically (“almost surely”) smart by Law of Large Numbers (LLN). The U.S. patent #8,449,378 of IGT (2013) describes an even smarter slot machine utilizing bitcoins. Soon, machines will become super-intelligent. But the intelligence of distributed hardware hinges on centralized software – and that fools us.
[1.] Before explaining why, some background: About 30-40% of all the Bitcoin mining power is reportedly in the hands of Chinese. While there certainly are more Chinese than just “a few,” most of the Chinese mining power is concentrated in just a few mining pools. Secondly, a large portion of bitcoin’s traded volume is taking place on Chinese exchanges (e.g. BTCC) – although it may partly be a mirage due to intentional wash trades. Thirdly, almost all of the powerful mining chips, application-specific integrated circuits (ASICs), are made in Taiwan (part of China). So, there really appears to be much going on in China. It may be because of their cheap electricity from hydropower sources or system abuses. But it is also cultural: Chinese have a nag for gambling – and control. In fact, China’s central bank is said to study the prospect of issuing its own digital currency.
[2.] On April 8th, 2016, I accidentally stumbled on a piece of news about black holes’ jet temperature: 18 trillion degrees Fahrenheit or 18 million million (18*10^12) F. Similarly to the two aforementioned U.S. patents, the numbers are not so important. What they conceal is: the jet temperature finding was achieved by linking together telescopes to get a finer resolution of a very distant object, a process known as interferometry in astronomy. Bitcoin operates analogously: its public ledger (blockchain) is updated via a network of distributed machines, all with (supposedly) aligned interests. It has been stated that, back in 2013, the computational power of the Bitcoin mining network, 7*10^19 Floating-point Operations Per Second, was already roughly 300 times more than the combined power of the Top 500 supercomputers. Now the network hash rate has rocketed to 16,159,415 peta-FLOPS.
[3.] Similarly, LLN is one of the most powerful statistical laws ever invented. It states that with a large number of repeated trials in an appropriate environment, the observed (by us) outcome (estimate) will approach its theoretical counterpart (truth) ever more accurately. In gambling, this implies that as long as the odds are even slightly in favor of the house, in the long run the house will win – almost surely. Essentially, LLN exerts control over the players like a government would like to exert power over its citizens; the result becomes totally predictable in the long run – a utopian like governance. But like every proper utopia, LLN cannot reign over everything: if its components have a large enough variety and/or are correlated in various enough strong forms, it fails to work well. To function, LLN requires (nearly) independent and identically distributed (IID) components.
[4.] Granted, the number 4,448,419 is large, but it is still 310 times less than the number of Chinese, as of April 17th, 2016 (U.N. estimate). What does that have to do with Bitcoin? Nothing directly. But indirectly: the U.S. patents are governed by the United States Patent and Trademark Office (USPTO). Chinese are governed by People’s Republic of China (PRC). Both entities exert power over individuals in their respective domain of attraction (DOA), like finite variance IID disturbances belong to DOE of a Gaussian distribution – Bitcoin belongs to DOA of its miners and core developers; whilst its technology is open and “free,” its management lays “In the Hands of a Few.” It exerts power over the other coders. It is also the network’s Achille’s Heel. LLN fails to function properly; the system components are non-IID – correlated and/or highly varying – that require careful governance.
[5.] In January, 2016, one of the core Bitcoin developers, Mike Hearn, quit the project and announced that the “Bitcoin experiment has failed.” It immediately sent the price of bitcoin to a free-fall – although only for a moment. Interestingly, Hearn’s criticism echoes a concern put forward in my earlier “Imperialistic Democracy of Bitcoin” (ITHOAF Part II) article although he focuses more on the more evident problem of centralization of the Bitcoin code repository “In the Hands of a Few” rather than what I call imperialism. The chaotic large-scale dynamics that are likely to emerge from code centralization are worth a more theoretical treatment. Here, I use LLN – and Central Limit Theorem (CLT) later on – to probe in the governance of (ideally) IID components. After all, a process with IID components that would guarantee – almost surely – long-term capitalistic growth has not yet been invented.
[6.] I have proposed Social Entropy Index (SEI) in my “Big Data Brother” (ITHOAF I) article. It serves as another theoretical illustration tool: In capitalism with IID components, SEI gets maximised. In communism, SEI gets minimised. The inherent system dynamics of today’s capitalism lead towards more concentration according to the Rule of 3&4 of Bruce Henderson described in “2 Fast 2 Furious Need 4 Speed Antidote” (ITHOAF III). Then, statistically, the component distribution will have fat tails and infinite variance. It obstructs LLN. If only SEI stays high enough (close to 1), then all of the components will be small and (almost) identical, creating an IID system with a finite variance distribution. But as SEI gets progressively smaller, the dynamics of power concentration will take hold and prevent a predictable stable outcome governed by LLN. The system becomes unstable.
[7.] These suboptimal large-scale dynamics are now taking place in cryptocurrencies, most visibly in Bitcoin and its blockchain. Control of ware – software or hardware – “In the Hands of a Few” is actually consistent with a statistical fat tailed power law distribution (Paretian, say), which implies inherent instability and lack of a normal equilibrium. In practice, it means that tranquil and turbulent periods are characteristic of the system. It is then nearly impossible to predict the next move; booms and crashes alternate much like they do in financial markets (widely regarded being highly non-linear by nature). A way to control such a critical system is to stabilize its volatility either through (1) a government or other centralized unit (think of Keynesian economics) or to (2) fix the source of instability that lays hidden in the dynamics of economies of scale, and conflicts with the law IID LLN.
[8.] But let me now take a deep breath of smoky Las Vegas air and get into the business of creating pseudo randomness. Telnaes developed his patented software in the late 1970s while working for a large slot machine manufacturer called Bally Distributing (acquired by Hilton Hotels Corporation). His “T-mapping” innovation became a game changer even by Las Vegas standards; bought from Bally in 1989 by International Gaming Technology (IGT, not to be confused with ITG) it quickly got widely licensed. To comprehend its deep repercussions, I have to explain what in it is random and what is – almost surely – deterministic. It should then be clearer how we get fooled by randomness in Las Vegas, and later, how we get fooled by randomness in Bitcoin. IGT’s patent #8,449,378, “Gaming System, Gaming Device and Method for Utilising Bitcoins,” is only the tip of a Titanic size iceberg.
[9.] A normal mechanical 3-reel slot machine with a reel of 22 stops has a total of 10,648 (22^3) payoff combinations. For it to be fair, at least one jackpot has to be rewarded once in 10,648 runs. But then, in a small stakes slot machine, the jackpot cannot become large, which creates an incentive problem. But what if a computer – or its random number generator – would pick the winner and the mechanical reel would just visualize it? This T-mapping enables slot machine manufacturers, in Telnaes’s (or Bally’s) patented words, “to make a machine that is perceived to present greater chances of payoff than it actually has within the legal limitations that games of chance must operate.” Slot machines can then also be physically small – no need for thousands of mechanical reels – which enables a larger number of digital slot machines to be placed in the mazes of the Strip. “More == Better”!
[10.] But what you see is not necessarily what you get – not even in Las Vegas. Not even close: “Slot Machines: Pursuing Responsible Gaming Practices for Virtual Reels and Near Misses” describes a psychological perception bias allowed by the state law of Nevada. Pre-programmed EPROM chips inside slot machines are able to create an impression of a “near miss”: the symbols above and below the payline are not in fact random – they just give you an illusion of control and suppress your rational thinking. A feature known as “clustering” – a bit similarly to clustering of volatility and trade durations in financial markets – puts a series of blanks (say, 5) adjacent to a high-paying symbol in order to create a near miss with a near determinism. In this case the so-called “award symbol ratio” would be 5. Again you may be wondering what does all this have to do with Bitcoin and blockchain?
[11.] (Patience, it is only the 11th paragraph and LLN has not fully kicked in. Converging.) IGT’s patent #8,449,378 is about the use of bitcoins in the context of slot machines. That, and some other similar patents aside, there have been only few granted patents related to Bitcoin in the U.S. before 2014. But now the number of pending patents is increasing. Rapidly. Bank of America (BAC) is one of the large companies actively filing patents: in January, 2016, a set of 20 new patents were reportedly filed by BAC to the USPTO. On November 19, 2015, Goldman, Sachs & Co. filed a patent application on “Cryptographic Currency For Securities Settlement.” Obviously, they take this road to maintain their present business positions and, if God allows, control the future through protected, often vague, ideas. And God allows it in the U.S. – just watch and learn from the tv-series Shark Tank.
[12.] Although the patent race has started in blockchain and cryptocurrencies, and we are likely to face a scenario where patents are “In the Hands of a Few” like they are now in the mobile phone industry, a more imminent concern needs to be discussed. To do so, I will have to head back to China and in its history. See, Chinese have a different but powerful way of calculating large things – and by that I do not mean their avant-garde bookkeeping principles. I mean the way to calculate that started with their personal calculator, known as abacus about 5000 years ago, making the handling of large numbers easier. Whether it is due to their inherent desire for computing power or their number of people, Chinese modern mining pools know their abacus and have taken a commanding role. Chinese seem to control much of what is related to the price of Bitcoin – and its value.
[13.] At the Chinese lucky number 13, it is appropriate to talk about a tide change. The good “chip mining days” of Macau – a special administrative region of the PRC known as the Asian Las Vegas – are gone. Mega-casinos like Caesar Entertainment Co. and Wynn Resorts know that. Now, regular digital slot machines are not as attractive nor revenue creating as modern crypto slot machines, such as bitcoin. Bitcoin mining and its speculation have skyrocketed in PRC, and much of the volatility is addressed to Chinese. No wonder IGT and others seek a piece of the pie by filing patents on randomness. Bitcoin miners are themselves essentially playing a game of roulette with a payoff expected every 10 minutes. Increasing anonymity by an extra round of roulette at every so-called Masternode is done by Dash coin. The Western big players surely want to scrape off the Western bad omen of 13 in the Year of Fire Monkey, 2016.
[14.] (I cannot wait for LLN to take full effect, so I will have to accelerate its convergence.) The mega-jackpots of mega-casinos that typically capture the imagination of Vegas goers rely on the distribution of slot machines, all operatively IID, but connected to a central computer. That Masternode dictates if and when a player should be rewarded. Although it may conceptually appear to be far removed from the IID ideology of blockchain, the difference is not so dramatic: in both the decentralized blockchain and the centralized jackpot lottery, the production software version – so also the control of hardware – is “In the Hands of a Few.” Now comes the hardest theoretical thread of it all: a centralized fragile system does not obey the (near) determinism of CLT. If CLT would be allowed to function well, the sum of its nearly IID components would converge to a robust and theoretically unbreakable (Gaussian) product.
[15.] Like the ASIC chips planted inside the modern bitcoin mining rigs, Erasable Programmable Read-Only Memory (EPROM) chips – based on the U.S. patent #3,660,819 “Floating Gate Transistor and Method for Charging and Discharging Same” granted in 1972 to Dov Frohman-Bentchkowsky of Intel Corporation – are planted inside the digital T-mapped slot machines. A public list of EPROM manufacturing companies reveals where they are from, with a large portion of manufacturing done in Asia. But Asian hardware manufacturers do not exert control over people. Software does. For that reason it is more important to know who controls the software than who controls the hardware; with enough intelligent software even IID hardware components can be connected, pooled, and hacked. So, the control of software reigns supreme – as many hackers may agree with.
[16.] The award symbol ratio – or clustering of non-paying blanks – implies the blanks in slot machines are not IID. The clustering of Bitcoin mining power and the fall of blockchain’s software control “In the Hands of a Few” are like the slot machine’s blanks; they create an illusion. A clustered system is in a critical state and never in a stable equilibrium. Its characteristic large booms and busts alternate unpredictably. Small disturbances can propagate into massive effects, in sharp contrast to where LLN and CLT promise smooth convergence when their components belong to their DOI. British historian Edward Hallett Carr is known to have viewed history as “a considerable extent a matter of numbers,” which stresses the importance of social forces (large numbers) – and less of individuals; it is harder to disturb an IID system by an individual effort once it has stabilized.
[17.] Deceivingly, the form of authority – still very much centralized – is also changing shape; Big Data Brother gets harder to spot, sort, and purge. Like discussed in “Imperialistic Democracy of Bitcoin” (ITHOAF II), I expect the characteristics of instability to persist or get stronger. If the digital economy slides towards more concentration (SEI gets closer to 0), disruptive volatility will increase. To prevent that, blockchain should self-enforce reversion to IID ideology. Unfortunately, it is not clear how to accomplish it. In this sense, I agree with Hearn; in my view, only if there exist strong incentives to maximize SEI, can Bitcoin truly become robust. Now, it is the case that SEI is quite far from 1 (cf. China), so my expectation is low. Having many options of cryptocurrencies is good, but history has shown a few times that a suboptimal technological solution can hack its way to the top without adjustment.
Conclusion. To continue in my tradition (since ITHOAF II) of citing The Wired Magazine in relation to Bitcoin, their 2011 article “How One Man Hacked His Way Into the Slot-Machine Industry” tells a story of a Latvian counterfeit slot machine business that got out of hand and ultimately whacked by IGT. Searching the USPTO web site for “slot machine” now returns 7958 patents; patenting has always played a key role in the gaming industry. It is likely to do so in the blockchain industry. The Wired story runs in parallel with the likely future of blockchain – if not necessarily of Bitcoin. Today, the total number of cryptocurrencies is 3328 of which most have no or little market capitalization. The moral of the story is that the control of software, open-sourced or not, will have difficulty staying democratic. All important decisions will likely be – you guessed it – “In the Hands of a Few.” □
[Signed: 7/.V]