Hashpower, True Names, and Dysonian Gods
|Venkatesh Rao||Aug 4, 2017|
I hope those of you living in or visiting the continental United States this month will do your best to get to a place where you can view the total solar eclipse on the 21st. I saw one in 1996, and believe me, it's not a sight to be missed. I plan to view this one too. It is a rare opportunity to look with the naked eye (only at totality, not even slightly before or you'll blind yourself) at the source of all life on earth: the Sun. Well, almost look at it. Paradoxically, a total eclipse is when the Moon totally obscures the disc of the Sun. Turns out though, this doesn't matter. There is very little of interest to look at on the surface of the Sun, compared to planets. With the right equipment, you get to see some sunspots, that's it. And they don't even last long enough (unlike say Jupiter's red spot or Earth's continents) to name. Jupiter, Mars, the Moon, are far more interesting sights in a small telescope. The point of a total eclipse is that you can see the corona, the intensely energized and rarefied local neighborhood of the Sun. And that is very interesting. Don't let ignorant people tell you that 90% partial is 90% as good as totality. A total eclipse is qualitatively different from even 99%. The corona is only visible at totality.
What do eclipses have to do with the themes of this this newsletter? They shed light -- or rather, useful darkness -- on possibly the single most dangerously misunderstood metaphysics topic in technology: centralization vs. decentralization.
Figurative illustration of a Dysonian god: a Dyson sphere used solely for raw hashpower
1/ Even the most ardent and technically sophisticated evangelists of software as a decentralizing force deeply misunderstand the subject.
2/ As a result technologists are repeatedly blindsided by centralization dynamics appearing in whatever they are currently ecstatically celebrating as the end of centralization.
3/ For most people, the starting thought is the not-even-wrong one that networks and hierarchies are in dichotomous opposition and "networks will replace hierarchies."
4/ This is often countered, by people who are slightly less clueless, with the gloating idea that "hierarchies are networks too." Again, not even wrong.
5/ Sure, a hierarchy can be represented as a graph just like a network, (broadly tree-like, with possibly some cycles, cross-edges), but that's a superficial map-level similarity.
6/ What distinguishes a hierarchy is that there is a power gradient: a direction pointing to an "up." There is a qualitative difference in what it is like to be on top (pleasant) vs bottom (unpleasant).
7/ You don't need a formal hierarchy for this. Any graph can manifest local/global power gradients, usually pointing towards highly connected nodes ("hubs").
8/ A graph topology is not necessary. Centralization/decentralization can manifest purely as power gradients in potential fields, without meaningful graph representations.
9/ Accounts of centralization/decentralization, understood without the scaffolding of graph topologies, generally approach their topic in terms of center versus periphery.
10/ The periphery may or may not have an obvious geometric relationship to the center. Though usually there is one: the Sun and central galactic black hole are geometric centers too.
11/ The city of Rome was not obviously a geometric center of the Roman empire on a 2d map, but on a communication network graph, as "all roads lead to Rome" suggests, it was.
12/ But there are non-geometric, non-topological, non-graph-theoretic reasons why the Sun, the central black hole, and Rome, are centers. They are power centers.
13/ A power gradient implies a power center. You climb a hill, you get to a "top." Whether you follow a network of roads to Rome or traverse free space sunwards is irrelevant.
14/ Some aspects of power are purely topological. Merely having a lot of inbound roads makes a city powerful. But this is only necessary, not sufficient, for a power center.
15/ But generally, power implies an intrinsic kind of power at a node that is not a function of its position or connectivity with its topological neighborhood.
16/ Graphs with intrinsic power distributions modeled generally label nodes as "sources" or "sinks" based on whether they produce or consume power.
17/ Batteries are sources of electric power in circuit diagrams for example. Grounding points are sinks. Capital cities are sources. Slums are sinks.
18/ The power of the Sun is based on thermonuclear fusion and sheer mass. A blackhole derives all its power from gravitational potential. Rome had military-political-financial power.
19/ Power has this unusual property that the potential to produce/retain it increases with concentration. Power production of any sort enjoys serious efficiencies of concentration.
20/ In fact, ALL economies of scale are economies of power in disguise. You standardize processes in order to apply power at scale. Variety causes diseconomies of scale.
21/ Variety -- variations from a "powerful" standard -- is pure information. The more information variety there is, the harder it is to apply central power to it.
22/ This is neither good, nor bad. It just it is. If you make weird electric equipment that cannot run on grid voltage/frequency, it is harder to power it from the grid.
23/ This brings us to my definition of decentralization: the separation of information and power. To understand the motivation for this, we must first look at the nature of pure power.
24/ Electric power is the clearest example. You can store energy in decentralized ways (batteries) but generation requires concentrated production of raw watts.
25/ Even solar power illustrates this: it isn't decentralized generation: it is decentralized collection of radiant energy from an actual centralized power source, the Sun.
26/ A Dyson sphere is a theoretical solar farm that would wrap around the sun spherically. For a solar-system-scale civilization, it would be a centralized electric power plant.
27/ In computing, the migration of compute power from mainframes to minicomputers and then to PCs/mobile was not pure decentralization: the picture is cloudier than that.
28/ From early datacenters to modern clouds, the sheer scale of centralized computing power has gotten awesome. The Internet exists partly to access it. All roads lead to Cloud.
29/ But nothing illustrates the tendency of power production towards centralization, and its relationship to information, which tends to decentralize, better than blockchains.
30/ The utopian decentralization dream turned into a massive centralization of hashing power. Small-minded decentralizers don't get the beauty of this.
31/ I find the evolution of blockchain technology to centralized hashing operations measured in a power-like unit of gigahashes/second profoundly interesting/revealing.
32/ Newer blockchains like Ethereum use clever methods to mitigate this centralization, but I suspect other kinds of centralization dynamics will kick in.
33/ If information wants to be free (decentralize) power wants to clump (centralize). One way or another, power -- undifferentiated potential of some sort -- wants to concentrate.
34/ Arguably the essence of decentralization is differentiation (to spread and separate is to diverge informationally), and of centralization is erasure of differences (converge informationally)
35/ To understand why, think about what a hash is. A hash (read up on it; it is basic technical literacy for 2017) is something like a name.
36/ An ordinary hash is to an arbitrary bunch of information as a word is to its dictionary definition. A word is the "name" (or hash) of the associated informational concept.
37/ A word is also a "key" to its meaning. The informational "handle" that marks its variation from the standard empty word (all zero bits) and allows you to uniquely manipulate it.
38/ A cryptographic hash such as is used in blockchain technology is the demigodly relative of a regular dictionary-grade hash. It is associated with a pair of keys, public and private.
39/ To oversimplify, public keys allow you to look up definitions in dictionaries. Private keys allow you to change the definitions, like a dictionary editor. A key pair is elemental thingness, an informational stem-cell.
40/ Producing a public/private key pair is harder than producing a non-cryptographic hash key. But producing a hash with pre-defined special properties is harder still.
41/ Here is the relationship between information and power. To produce the special cryptographic key pairs required to make blockchains work you need raw power.
42/ "Hash power" turns raw energy watts into hashes with special properties (a certain number of leading zeroes in the hashes identifying newly minted "bitcoins" for example).
43/ These special hashes can be considered the "true names" of the bundles of information they get attached to (the blocks) in a global namespace (a blockchain).
44/ To true-name a thing is to breathe life energy into it. To elevate it from indistinguishability and interchangeability and give it a minimal-viable numbered stem-cell identity.
45/ Naming is the foundation of computing. You can't process information without naming things to distinguish them first. Key-pairs are unwidgets, relative to the widgets of MBA textbooks.
46/ There appears to be no cheap short-cut to the production of true names. It burns up a certain minimum amount of energy. It is the hardest (physics sense) part of computing.
47/ Bitcoin is an early, relatively inefficient, and non-renewable way to burn energy to produce true names, which is why "bitcoins" are limited by 21 million very expensive true names.
48/ Bitcoin is a countdown from 21 million to 0 (which is why the number of leading zeros is relevant in the "true name" hashes). Like a star burning down to a white dwarf.
49/ You can think of a string of zero bits as the only "standard" in information technology. Naming is the act of creating elemental variation -- information -- to drive all else.
50/ Information cannot be created or destroyed, but meaning can be turned into noise. Naming is the opposite act: turning noise into elemental meaningful information.
51/ To get esoteric for a moment, it turns out, setting all the bits of a random bit string to zero also takes energy, just like creating a true name takes energy.
52/ This is called Landaeur's theorem and is the reason the 2nd law of thermodynamics applies to computers. Information cannot be destroyed, but meaning can be -- for an entropic price.
53/ So there is an interesting symmetry here: both elemental information operations, true-naming a thing, and erasing the name of a thing, produce entropy.
54/ Stepping back, a datacenter with massive hashing power is a centralized power generation source that does one thing and one thing only: produce true names.
55/ Aside, here is how you make a basic God: make a Dyson sphere around the sun, pipe all energy into a hashpower cloud that broadcasts true names (aka words of god)
56/ Now you can see why it is useful to define decentralization as the separation of information and power. The boundary between the two is the production of true names.
57/ A hash -- a unique, true name of a thing -- is a degenerate, minimally distinguishable information fragment. A stub for future thingness. A word with a blank entry in the dictionary.
58/ This is why naming is an act of such cultural significance, as is the unique numbering of banknotes or limited edition prints. We recognize this instinctively.
59/ Makers of limited edition block prints follow a practice of destroying the block with a gash after the run (these defaced engravings are valuable; gashes are anti-hashes).
60/ This is why art can serve as currency. Individually valued fine art pieces can be considered blockchains of length 1. Prints from defaced block engravings are longer artisan blockchains.
61/ Bitcoin is a series of 21 million limited-edition numbered prints block-printed (heh!) algorithmically by the print-maker Satoshi. The transactions are just the ink.
62/ Ethereum is a little more subtle, it is based on a rate-limit on the production of new ether, combined with a promised cheaper (energetically) future hashing technology.
63/ Back to names. To know the name of a thing is to know the absolute minimum possible about it to distinguish it in a universe that is not just an undifferentiated ball of energy.
64/ So to decentralize everything except the production of true names is to decentralize maximally. Blockchains only centralize the minimum necessary: power (watts), and true naming (hashes/second)
65/ This incidentally suggests a way to resolve the question of whether blockchains are singular or plural, fluid or discrete. Is it blockchain, the blockchain, or a blockchain among blockchain_s?_
66/ If you mean the fungible hashing power that can be reallocated across specific blockchains, you could drop even the, the way you just say water for the substance qua itself.
67/ For example, following the recent fork in Bitcoin, some miners are reallocating hashing power to the BCH fork. You could say blockchain is being reflowed across forks (I don't recommend this usage).
68/ If you mean the centralized, deconflicted/no-collision namespace or hashspace, the definite article, the blockchain is appropriate. It's a bunch of the disconnected islands.
69/ As an example application, if you accidentally send ether to a bitcoin address, you could say it fell through the cracks of the blockchain. This is the blockchain like the internet.
70/ Global true names require a certain minimum level of coordination. Or do they? It may be possible to break even this constraint. This is the premise of Zcash.
71/ There is a lovely idea known as Zooko's triangle -- that you can only have 2 of 3 properties: human-meaningfulness, security, and decentralization in a naming system.
73/ Not surprisingly, one of the consequences of being able to do that is that Zcash, unlike Bitcoin or Ethereum, is a blockchain that offers true anonymity.
74/ To create true anonymity, you need the ability to create and destroy true names. Shiva-grade hashpower. Zcash had to be initialized with some interesting destruction.
75/ Back to less esoteric (but not by much) things: whether or not the naming layer of computing can be decentralized away, I suspect power will remain centralized.
76/ I don't have a good reason for this belief. Just an intuition that perhaps power ought to be defined as the creation and destruction of names (or true informational variety).
77/ I'll leave it to any friendly neighborhood black hole physicists/information theorists to figure out how this might work, but let us get back to lower-math-pay-grade stuff.
78/ It is no accident that power in culture is associated with un-nameability. Voldermort was "he who must not be named" in proportion to how much you feared his power.
79/ Most religions have a tradition of their central deity being unnameable in some way. It is both the void of no names and the fountainhead of all names.
80/ In Hinduism, a famous verse declares the central generative source of all creation to be unknowable/unnameable. Devout jews spell God G_d and don't verbalize it. The Tao that can be named is not the true Tao.
81/ At the other extreme, Vishnu is more an incarnation namespace than name. There is a long hymn, a favorite of my father's, known as the thousand names of Vishnu. To be all names is to be no name.
82/ On the sci-fi end, a friend recently suggested that perhaps humans are actually cryptocurrency tokens for god-beings (this is apparently a subplot in Charlie Stross' Accelerando)
83/ I am not religious, but I find the culture of naming around conceptualizations of the divine among the religious fascinating and instructive. There are some big ideas in there.
84/ The less power you have, the more nameable you are. Rumplestiltskin had a name but it was secret and he lost power if anyone found out. But he was only a minor power-being and got doxxed.
85/ We humans vary in how we present online, there is a whole sociology of nyms out there -- anonymity, pseudonymity, and so on. It takes power to hide online.
86/ Real names are at the center of informational power dynamics. Transgender individuals take offense at being "dead-named." Pronouns (relative names), doxxing, and identity theft are modern battlegrounds.
87/ To have an unlisted phone is to have power. To be an anonymous donor is to have power. To be Satoshi, pseudonymous creator of blockchain (no article) is to have power.
88/ If Google or Facebook accidentally outs "private key" aspects of your identity, like being a gay person in the closet, that's genuinely a game of power dynamics playing out.
89/ Both the ability to create true names and destroy them represents power. If bitcoin is the alpha of naming hashpower, Zcash is the omega.
90/ Stewart Brand once said, "we are as gods, we might as well good at it." In a way, even without talk of DysonHashing god spheres, blockchains are the first true god-technology.
91/ With blockchains we've stepped up to at least one God-challenge: creation and destruction of true names. The Bible got the history wrong. Adam is only now true-naming the beasts.
92/ Let's close with the symbolism of total solar eclipses. Looking at the corona of the Sun is the closest you'll ever get to see raw true-naming hash-power in nature (unless there is a supernova in your lifetime).
93/ As the Hitchhiker's Guide to the Galaxy said, real infinity is kinda flat and boring. There's not much, and nothing nameable, to see on the surface of the Sun.
94/ The central black hole of the Milky Way too is kinda boring, tiny and unimpressive though it's impressive that we've been able to make a movie of stars orbiting it.
95/ But to paraphrase the Hitchhiker's Guide, seeing something really really close to awesome power conveys a better idea of it than staring directly at raw power.
96/ Staring at true raw power of any sort is hazardous, which is perhaps why religions and monarchies tried to fake it by introducing taboos around looking at gods and kings.
96/ Anyhow, after that estoric technotheological riff, let me close with a final reminder to go watch the eclipse if you can. May the hashpower be with you.
_Feel free to forward this newsletter on email and share it via the social media buttons below. You can check out the archives here. First-timers can subscribe to the newsletter here. You can set up a phone call with me via my Clarity.fm profile page. _
Check out the 20 Breaking Smart Season 1 essays for the deeper context behind this newsletter. If you're interested in bringing the Season 1 workshop to your organization, get in touch. You can follow me on Twitter @vgr
Copyright © 2017 Ribbonfarm Consulting, LLC, All rights reserved.