Massively Scalable Digital Currencies

Here we search for the holy grail of micropayments: a decentralized transaction system that can scale to the volume of the internet itself; the promised land where transactions are plentiful as UDP packets, fast as ping, and as cheap as bandwidth itself.


Bitcoin is a working example of a decentralized transaction network.  It’s Proof of Work scheme has certainly proved itself to work in the field, but at the cost of zero parallel scaling: transactions are redundantly replicated across all full nodes, so the maximum performance of the system is nearly equivalent to the maximum performance of a single node.  The nascent network still has room to scale, but eventually will run into the bandwidth and storage limitations of the acceptable minimum requirements for a full bitcoin node.  The interesting recent proposals for performance provide a constant multiplier, they don’t change the asymptotic scaling which is and will be O(N).

There are many interesting applications of distributed transaction networks that have vastly higher performance requirements.  The financial markets of the world such as NASDAQ, BATS and kin are a good starting point, but really just the tip of the iceberg.  Sans any and all performance limitations, what kind of applications could a vastly scalable micro-transaction network enable?

For inspiration we can look to the future vision of an impending AI-singularity: an explosion in computation and intelligence leading to a world quickly populated by an endless sea of software agents ranging from the smallest dedicated trading bots on up to the true superintelligences: vastened minds thinking orders of magnitude faster and deeper than mere biological brains. The key constraint on the intelligence explosion is the locality of physics as expressed by the speed of light. The faster an agent thinks, the slower the outside world becomes.  Latency and bandwidth become vastly more restrictive.  The future is perhaps dominated by localized pocket civilizations around the size of a city block: a vast expansion of virtual inner space.  A globally synchronous protocol like Bitcoin has little place in this future.

Even before the full intelligence explosion begins in earnest, a scalable micropayment network could enable an Agoric Computing revolution.  Many real world problems of interest can be formalized as multi-agent / multi-utility function coordination problems that are well addressed by market systems.

Imagine a smart traffic marketplace where automated vehicles rent out their time, road lane usage is rented, user agents bid for service, and various options and derivatives are used to predict and hedge traffic events.

Open markets could potentially solve a key problem with the health industry: the misalignment of financial incentives.  Consumers have an interest in maximizing quality and quantity of lifespan.  Medical companies currently have a greater financial interest in recurring product revenue via indefinite medication rather than one-off cures. Health/Life insurance could be revolutionized by creating a marketplace for insurance contracts and associated derivatives such that health research innovators could profit from the true economic value of actually curing or even preventing diseases.

A marketplace for grid computational resources itself could enable entire new classes of massive dynamic software ( indeed we are already seeing the early stages of this with nascent cloud computing markets such as Amazon’s EC2).

Lofty Goals

The ideal transaction network would have the following qualities:

  1. throughput: peak aggregate transaction rate approaching the theoretical maximum: transaction message size / total network bandwidth (zero duplication overhead)
  2. latency: most transactions add little additional latency beyond the minimum packet traversal from sender to destination
  3. security of ownership: private control of assets is near-guaranteed
  4. uniqueness of assets should be cheaply verifiable and counterfeit-resistant
  5. robustness: high systemic existential security: large-scale redundancy via P2P decentralization

Bitcoin scores well in terms of goals 3,4,5 at the expense of performance goals 1 and 2.  The aggregate bandwidth cost of a single transaction in the bitcoin network is roughly B = O(N*C), where N is the number of nodes, and C the bandwidth cost of a single transaction packet.  Thus the total transaction throughput T scales as T = O(N*B / N*C): which B is the per-node bandwidth, and thus is just a constant: T = O(B/C).  Ideally we want the average transaction to visit only a tiny fraction of the node network so that B << O(N*C) ~ O(C) and thus T ~ O(N*B/C).

Starting from this perspective on the problem, the throughput and latency performance constraints suggest that any highly scalable solution network must be both sparse and local: the great majority of transactions should only propagate to a handful of network peers.  Guided by this insight, we can search the landscape of digital asset protocols to find solutions that best optimize over performance and security tradeoffs.

The core of digital poperty networks like bitgold/bitcoin is a cryptographic ownership chain.  For the sake of simplicity, we will start with indivisible quantual assets, similar to the early bitgold idea.  These assets naturally form sets to represent different categories of assets, but let’s start with a simple currency which we can call the quoin.  As part of the initial consensus protocol, we can consider the set of all quoins to be pre-existent and ordered as strings: for example QU0001, QU0002, etc up to some established limit of N units.  Each quoin is associated with a single public cryptographic key identifying its owner, ala bitcoin.  (the cryptographic implementation details are irrelevant for this discussion). Quoins further each have a preassigned value or denomination distributed according to an exponential such as 2^d, where d is a simple lookup table function based on the quoin index (transactions of arbitrary value would thus involve multiple quoins and returning change, similar to physical currency).  The set of quoins can thus be considered a flat array structure where each entry stores the public key identifying the current registered owner of that quoin.

Proof of ownership can simply be established by a chain or linked list of signed transactions.  It may seem odd and cumbersome to constrain these quoins to be indivisible, but this avoids complex transaction graphs.  The transaction chain for each quoin is unique, independent, and does not require any form of timestamping.

The core security problem with such a simple system is double-spending.  Notice however that a double-spend creates an obvious violation of the protocol: a branch in what should be a single-linked transaction chain – which is easily detectable as two transactions which share a previous link.  Any solution to this problem requires additional overhead for relaying transactions to independent third party nodes who can help resolve the protocol violation in favor of one of the potential paths.

What may not be obvious is that a fully centralized solution is far from ideal: a central verification node would amount to a critical bandwidth and latency bottleneck.  Any massively scalable protocol must widely distribute dispute arbitration authority across the network.

The core idea of LSDQ is to consider trust, dispute arbitration and consensus from an economic incentive perspective.  The protocol can remain extremely simple by pushing much of the responsibility for dispute arbitration onto the nodes themselves, exploiting a degree of local predictive intelligence embedded in each software agent.

There are numerous potential fork-resolution protocols that converge on a stable consensus.  The simplest and perhaps most effective is a weighted quorom protocol: forks are resolved in favor of the link that receives the most weighted votes.  Crucially the votes are weighted by asset ownership share: in the quoin example each quoin would have a vote proportional to its denominational value.  Each quoin could be used to cast one weighted vote per fork dispute, multiple votes are protocol violations and thus discarded.

Todo: only first encountered vote is considered, rest are considered fraudulent – double-voting?

The weighted quorom approach is a form of micro-democracy; and as such can be considered a pure Proof of Stake system, but it is much simpler than PoS as conceived in bitcoin derivatives such as peercoin.  The key difference is that a weighted quorom protocol has nothing to do with new asset/coin creation: there is no ‘mining’ in the core protocol, it cleanly separates asset creation from transaction verification.  Weighted quorom can be combined with just about any initial asset allocation or dynamic creation algorithm.

Weighted quorum by itself is not very interesting from a performance perspective: in the worst case a full quorum could result in a linear number of additional verification messages spawned for each transaction, taking us right back to square one.  The key to fast transaction verification is selective intelligent pruning of messages.

We start with the following observations/assumptions:

  1. The wealth distribution (and thus voting weights) is approximately a pareto distribution (power law)
  2. Transaction sizes are likewise sparsely distributed from either a power law or exponential family

Agents can use a probabilistic approach to determine when and where to send verification requests.  If the face value of a particular quoin is V, the expected value to a recipient agent A can either converge to V (if the transaction is ultimately validated by weighted quorum), or 0 if the quoin is determined to be a forgery (ie the transaction is not on the highest weighted-fork).

The discounted value dvalue(Q[i]) of a quoin Q[i] is thus the face value fvalue(Q[i]) discounted by the probability of forgery(double-spend): dvalue(Q[i]) = p(valid(Q[i])) * fvalue(Q[i]).  The recipient can send out query messages to other nodes on the network, asking them to investigate the transfer chain and vote on the valid path.  Each of these queries can itself involve a smaller (and perhaps conditional) micro-payment to reward the investigating node for the computational work of the query and resulting vote (ie a transaction fee) – and for large transactions this process can recurse with further requests (and increasingly smaller payments) percolating along the network until reaching diminishing returns.

The verification process can be cast as a general utility/profit maximization problem to which we can apply various machine learning approaches.

As a simple starting point, consider a greedy algorithm for a verification agent.  The verification agent is a process which stores the history of many quoins, is well connected to important peers, and accepts micropayments for verification requests.  It charges or expects a micro-payment per request, and reliably responds to valid requests concerning a quoin with a small return packet containing some relevant portion of the transaction history and a signed vote.  The incoming request itself will embed the most recent transaction or two, so a well connected verification agent will naturally build up a partial view of the full transaction database as a side effect of its core business.  And of course it can also send requests to other verification nodes as needed.

The simple verification agent receives a stream of incoming requests and has an action set consisting of: 1.) null (wait), 2.) send response packet to customer, 3.) send request packet to peer.  For simplicity assume that all packet types are of a standardized size and thus have a fixed bandwidth cost C. The agent maintains a set of incoming requests R concerning the validity of query quoins Q, each of which has an attached micro-payment X, where the value of a quoin is ie: value(X[i]).  Responding to a request R[i] consists of fetching the histories of quoins Q[i] and X[i], checking the cryptographic chain and known weighted votes, and then sending a signed response packet with the valid history (probably compressed) which also functions as a vote on that history.

The core game-theoretic principle for these agents is tit-for-tat.  Honest/cooperative agents are profit-motivated and thus expect payments/rewards in excess of costs.  In this simplified model the balance of payment or utility U[i] for verification request R[i] is just the discounted value of the incoming micro-payment minus the fixed response cost:

U[i] = fvalue(X[i])*p( valid(X[i]) ) – C.

In fact all agents will have a similar model where C is the cost function for whatever service the agent is providing.  The utility of sending out a history vote request can then be derived from the expected consensus gain it provides, ie an increase in p( valid(X[i]) ).

The current estimated future posterior probability of X[i]’s validity after receiving a hypothetical vote from a peer k is p'( valid(X[i]) | H(X[i],k)’ ).  Thus the expected value of sending out request H(X[i], k) for quoin X[i] to peer k is:

ev(H(X[i]), k) = fvalue(X[i])*( p'( valid(X[i]) | H(X[i], k) ) – p( valid(X[i])) )

and the expected profit is: ev( H(X[i], k) ) – C.

Note that the gain from sending out a vote request in this model is due to its potential to increase the group confidence and thus value in a quoin, rather than pure information gain.  Thus it is only profitable to send requests when the agent expects them to increase the probability of a quoin’s validity, ie p'( valid(X[i]) | H[X[i]]’ ) > p( valid(X[i])).  The agent does not spend time sending out vote requests for quoins it already believes to be losers.

The core of an effective verifier is in the predictive functions: p( valid(X[i])) and p'( valid(X[i]) | H(X[i], k)’ ).  A sophisticated agent can employ general model-driven prediction techniques such as reinforcement learning, ANN, SVM, etc using the various transactional history datasets for training, along with competitive simulation contest results.  Game theory suggests that a smart agent can be expected to employ some degree of randomness in its decisions to foil double-spenders who could otherwise perfectly predict its routing decisions and thus more easily split the network.

Regardless of the estimation technique used, we can expect that the p(valid(X)) type functions will have a characteristic shape.  The initial probability or prior should at least reflect the general likelihood of double-spends across the entire system, but should also incorporate trust: past knowledge about the owner of X.  If the owner has a long history of honesty, this should substantially decrease the prior of fraud.  Likewise the fraud prior will perhaps depend on the size of the transaction.  Beyond that, the final posterior should increase asymptotically to a maximum approaching 100% as votes accumulate.

The p(valid(X) | H(X[i], k)) term depends on the vote weight node k controls, so all else being equal requests will be directed firstly and most often towards high-weight nodes.  An efficient agent will also consider the local network topology in its decisions, and perhaps employ multicast routing.

Keep in mind that consensus does not depend on the predictive functions an agent employs for optimizing transaction processing.  The key to eventual consensus is that honest nodes always vote to accept the first variant of any transaction they discover, and honest nodes never change their votes, double-vote or double-spend.  Local agent intelligence enables the network to scale massively by minimizing the effort expended to detect dishonesty and thus minimizing transaction costs.


The distribution of transaction values can be expected to take a power-law or perhaps exponential form such that the majority of transactions are small micropayments and large transactions are fairly rare.  For very small micropayments exchanged for sub-second computational services, the value of the transaction can approach a small multiple of the cost of message bandwidth.  The transaction fee could thus approach zero for the smallest micropayments.

For small transactions between trusted nodes, the predictive models outlined earlier suggest an absurdly low probability of double-spends such that investigative requests are unwarranted.  However the predictive confidence in a chain also decreases exponentially with the length of all unverified steps.

This results in a dynamic where small clicks of nodes with high inter-mutual trust can exchange in long sequences of rapid micro-transactions with little external network interaction, until eventually the chains reach some trust limits where external verification and auditing becomes warranted.  These chains can be compressed with hash-tree techniques combined with randomized auditing to reduce the cost of exporting quoins out of a click.

As the transaction value increases, the cost of a double-spend and thus expected value of confirmation packets increases in proportion.  The maximum worthwhile effort does hit a ceiling around the cost of securing a quorum of votes.  Interestingly enough, a highly inequitable wealth distribution actually reduces the number of messages required to secure a firm majority, reducing transaction costs.  For example, assume a population of about 10,000 verification nodes, and a pareto distribution such that the upper 10% control 50% of the votes.  The maximum cost of verification – which as discussed earlier should apply only to a tiny fraction of transactions – would thus require touching only about 10% of the network: around 1,000 messages.  Assuming a bandwidth cost of about $0.10 per gigabyte and an average message size of 1KB would give an upper transaction cost of just $0.0001, or 1/100th of a penny.  Which of course is rather ridiculously cheap, but that’s more or less the point.

The Case for Bitcoin

If Bitcoin succeeds, future generations will remember it as the greatest investment opportunity in recorded history.

The Forbes list of billionaire dynasties will be completely rewritten, with new names such as the Winklevi clan shifted to the front.  A few lucky individuals will become billionaires simply because they tried out this new bitcoin mining app their comp sci dorm buddy told them about way back in 2010.  Yes, we are talking about a radical and perhaps somewhat random wealth redistribution (although in that aspect at least the pattern is familiar to students of history).

Sounds unlikely?  Perhaps, at least for now.  If we look at the current bitcoin exchanges as a sort of prediction market, we can roughly estimate the net odds traders are currently giving for that scenario.  If the BTC becomes the major world reserve currency, then each bitcoin should be worth vaguely on the order a million 2013 dollars (~20 trillion total $ medium-high power fiat money / 20 million BTC).  So currently the markets are giving about 0.01% odds on that bet.

If you think those odds of the BTC-wins scenario are much higher, such as closer to say 1% or 10%, then you should take that bet and join the Winklevi and fellow TechnoLibertarians in the proud > 1.0 BTC club  (for there can only ever be 20 million people who own more than 1 BTC).

The seductive logic of a bet with such massive upsides partially explains how BTC (or any would be money) bootstraps itself into existence up from probability epsilon (a sort of real Pascal Wager).  Then the network effect kicks in: as the inflow of small bets boosts up the price, this rise in valuation itself becomes some additional evidence for the long term monetization hypothesis (because a good speculative trade is always recursive in terms of other agent’s speculations).  This process can become a virtuous cycle, eventually leading to the Bitcoinmana that has taken the $/BTC up from 0.1, 1, 10, 100 in just a few years.

Exponential rockets such as this tend to attract the attention of professional evangelist-hucksters who can sell rocket ride tickets on Fox Business, promoting Bitcoin to a wider TV-audience of uninformed traders.  Everyday joes may not have the time or inclination to read up on cryptocurrency, but they can fit a simple line to a graph and dream of F.U. quantities of filthy lucre.

But isn’t this just a speculative bubble?  Well yes, but not just.  Or rather it’s a speculation that BTC will become a global money standard – for this is how new forms of money are born into the world: as some sort of mutual game theoretic optimum, a Schelling point in the space of future trade options.

Money arises as the solution to a global optimization problem that maximizes the efficiency of a complex network of spatiotemporal trade paths while minimizing risks and costs.  At any time the markets are continuously exploring many different forms of money/savings instruments with varying tradeoffs: and necessarily creating ‘bubbles’ in the process.  Every once in a long while this ecosystem undergoes rapid evolutionary transitions.

The missing part of this simple ‘Bubble’ explanation for Bitcoinmania or Bitcoin hyper-monetization is why any original traders thought Bitcoin had any value in the getgo, or rather why they even considered, for a moment, that it ever had a chance above epsilon of becoming the new global money standard.  Why would it be better than the dollar, euro, or gold?

The Fundamental Value of Bitcoin

Here then is the argument from fundementals:  Imagine a single benevolent, omniscient tyrant (God, or an AI super-intelligence, etc) could simply simulate the global optimization in it’s mind directly.  This then eliminates all the recursive game-theory elements (bubblemania aspects): the tyrant then effectively evaluates different monetary systems based on their longer term net efficiency, according to its criteria.  The fundamental maximum value of Bitcoin then is the net difference in total economic utility (measured by say adjusted world GDP, to first approximation) between a Bitcoin based economy and the current regime.  If we have an idea of what that value is, we can then estimate the expected return or immediate value of Bitcoin as an option on that future weighted by our assessment of its likelihood.

Spatio-temporal Trade Effeciency

Bitcoin is a vision of a more efficient currency.  The efficiency I refer to is not just algorithmic, but economic in nature.

Economists used to write in their textbooks: “Money is a matter of functions four, a medium, a measure, a standard, a store.”  Those four functions have more recently been streamlined to three, but I will further reduce all of this to a single concept: money is a spatio-temporal medium of exchange.  By this broad definition, almost anything owned has some ‘money-ness’ to it, depending on our expectations concerning how we can use and or retrade it in the future.  Everything from cowrie to salt to tulips has functioned as a form of money in at least a few pockets of space and time.  In each case the items in question had some ‘intrinsic’ productive/consumptive values which perhaps helped boostrap them into moneyness.

Today such notions of intrinsic productive/consumptive value are irrelevant from the global optimization perspective, for all that matters in the end when comparing potential forms of money is their net efficiency in facilitating trades across space and or time.  Paper fiat money is the case in point: it evolved in the market from gold deposit slips having tremendous practical advantages over metallic coinage, but has no intrinsic productive/consumptive value.

First, let us consider the aspect of spatial efficiency.  Here Bitcoins have rather obvious advantages:  facilitating transactions over the internet to anyone in the world without exchange rate conversions, high fees, long waiting periods, etc: thus: high spatial efficiency, approaching optimal.  So Bitcoin could displace Mastercard/Visa and seriously displace large swaths of finance.  This is a net good.  It is difficult to measure the quantity of this improvement, but to first approximation it should be proportional to the market cap of all the companies it would make redundant, ie somewhere to the tune of a few hundred billion dollars, perhaps up to a trillion.

Bitcoins also have high temporal efficiency due to the simple ingenious algorithm which governs their supply.  Bitcoins grow on a asymptotic inflationary schedule.  This schedule has the following advantages for facilitating temporal trades (savings): the total supply has a known hard limit, the inflation schedule is known and perfectly predictable (removing the huge uncertainty of fiat), and finally the fast but exponentially tapering inflation schedule is exactly what is needed to foster the currencies adoption – because newly created bitcoins are distributed as a reward for the ‘miners’ who verify transactions and secure the network.

Some critics (including economists who should have known better) have claimed that BTC is deflationary, which besides being technically incorrect (BTC inflated by about 200% in 2010 and down to about 15% in 2013, and it will eventually reach an inflation rate of 0%, but the supply will never significantly decrease), is also apparently used as some sort of dirty word.  I suspect that Krugman and ilk use the ‘deflationary’ epithet because they are basically employees/propagandists of the threatened institution: Fed/Banks, and their rival products cannot compete in terms of temporal efficiency: simply because states reserve the right to create new fiat to pay their bills.

The more technically correct and potentially interesting criticism from mainstream economics focuses on BTC’s inherent inelastic inflation schedule: in the Keynesian view the supply of money should be dynamically controlled by a central authority to help absorb business cycle shocks.  This theory is based on an entire edifice of economics that arose out of the experience of the great depression and similar credit collapse debt deflations.  Bitcoin in raw form is immune to such shenanigans simply because it is high powered money: the equivalent of cash or Fed Deposits, not the demand deposit credit/debt based money the banking system currently uses.  As Bitcoin grows we can expect there will be at some point a new ecosystem of debt based instruments built on top of BTC, but this new ecosystem will be global and technological, more like Prosper, less like of BoA.

Criticizing Bitcoin based on economic tools used to analyze the great depression is like criticizing Nvidia’s new Titan video card based on a theory of Charles Babbage’s Difference Engine.

Gold has a constrained supply, so it can be reasonably temporally efficient, but it is completely inefficient spatially – which is how fiat came to be in the first place, as ‘banksters’ offered a product (paper notes) with much better spatial efficiency   As a result gold only exists today in the modern monetary ecosystem in digital form, as a contract.  So it’s just another computer ledger, but a number which we should trust because the computer ledger can’t be faked/fudged/misrepresented .. because each number in it exactly corresponds to a real unit of gold held in a bank somewhere, because . . . .  somebody said so.  Fiat currency started that way, as banknotes for gold redemption.  There have been attempts to revive that idea in the digital age (such as egold), but they have all been plagued by the centralized point of failure problem.

This brings us to the final and most important advantage Bitcoin offers the world: it solves the trust problem, in both the algorithmic sense as a solution to the Byzantine Generals Problem (which really is a big deal in computer science), and in the more typical economic sense.

The core of any implementation of money is a ledger: a simple database of account balances and a trade protocol to carefully(and atomically) add N to account X and subtract N from account Y.  That’s bank software, and the core of it really is that easy.  The difficulty is trust.

In economic terms, the ledger really is the most important damn thing in the world.  How can you trust the ledger?  With a physical currency this is straightforward (as long as the physical token is very costly to fake).  Physical currencies are good in that department, but they pretty much suck in every other way compared to purely memetic currencies (such as paper or digital).

The world’s current dominant fiat currencies all use some form of complex centralized ledger.  The US has a godawful complicated scheme involving the Fed, the Treasury, a hierarchy of banking minions, and a bunch of redundant databases.  But in the end it all boils down to a centralized ledger and trust concentrated in some specialized branch of the government.

There are at least three significant problems with this scheme: first, there are about 190 generally recognized sovereign states, and almost as many currencies and ledgers.  Thus giving rise to the significant previously discussed spatial inefficiencies moving money around the world – in the form of taxes, fees, tariffs, exchange rates, and so on.

Trust is also the core cause of inflation or the poor temporal inefficiency of fiat.  Libertarians, Liberals and Conservatives may cite different flavors of economics in explaining why fiat currency is inflationary, but there is little argument over the result: saving in fiat currency is discouraged – not only because it is worth less and less over time due to expanding supply, but also because the rate of increase itself is unpredictable.  In our current system it’s a much better long term trade to borrow a few decades of labor and purchase real estate (which has a naturally fixed supply and stable demand) than to simply save currency.  Mainstream economists (which I suppose only some of which are shills) praise inflation, because, they say, it encourages spending.  Somehow causing people to spend more now than they would otherwise choose to, and plan less for the future than they would otherwise choose to is supposed to be a good thing.

When pondering how we got into a situation in which a little over a 51% majority of the population ‘owns’ a house by ‘borrowing’ decades worth of future salary (against steadily decreasing median wages) from government controlled banks at near-zero or even negative effective real interest rates, it’s just too tempting not to quote Alexander Tyler: “A democracy cannot exist as a permanent form of government. It can only exist until the voters discover that they can vote themselves money from the  public treasure.”

Bitcoin is rather idiot-proof in this sense.  In a Bitcoin world the total future max supply of BTC is known: ~20 million units, similar to how the total supply of land area on Earth is limited, so in a BTC world the currency can be expected to perform similar to housing (on average).

The future supply and thus value of fiat (whether Dollar, Euro or Bhat) is determined by future elected officials, which naturally creates some serious issues of trust.  Historically these trust issues have caused wars.  China trades cheap goods for Future-Dollars, a trade which involves a great deal of delicate political and economic faith in the future US government – because China is treading the sweat of its populace now for an unpredictably but generally decreasing share of USGovCorp.

So in a nutshell fiat currency is basically stock in the relevant corporatocratic states, and trust in fiat boils down to trust in the financial future of the issuing entities (because they always reserve the right to generate new fiat in various forms to pay future bills).

And somewhat predictably: they are screwing up (to varying degrees and for various reasons).  The Euro is screwed and everyone knows it, most of the rest are screwed and just don’t know it yet.  The case for such pessimism concerning the future of various current statist powers is complex and beyond the scope of this post, but in short it revolves around the huge debt/credit edifice and welfare state whose existence is predicated on long term economic assumptions that will be absolutely shattered by the impending Technological Singularity  (but alas that is a topic for another time).

Gold has the desirable temporal efficiency but it is completely inefficient in the spatial dimension, so it becomes a digital fiat scheme: with all the same trust issues.  The governments of the world are not going to voluntarily give up their fiat and switch to gold.  Bitcoin solves this problem by not giving them much of a choice.  It is like a digital gold standard on steroids, but more importantly: it actually has a shot at success.

In Proof of Work we Trust

Bitcoin combines the high spatial efficiency of digital money with the high temporal efficiency (via supply stability) of a gold standard.  But how does it solve the trust problem?

The seed was a novel idea from the mysterious Satoshi Nakamoto.  (Sometimes to truly understand a thing, it really is best to understand it’s beginning.) The paper neatly summarizes an ingenious and practical proof-of-work solution to the core technical problem of distributed trust.

In the minds of a few lucky readers on a particular cryptography mailing list, Satoshi’s nifty idea blossomed into the current vision of a secure, efficient, distributed digital currency.  One currency to rule them all and in the darknet bind them.

For those for whom the paper is TLDR, I’ll briefly summarize what isn’t spelled out in the abstract.

Bitcoin solves the trust issue without trust.  No one particular person/group/node is trusted to maintain the ledger for everyone else.  Instead each node simultaneously maintains a copy of the ledger itself.  The database/ledger, called the blockchain, is itself just the entire transaction history, so it’s straightforward to verify the validity of each transaction.  And now the final hat trick: given multiple competing versions of the ledger/blockchain, each node picks the ledger/blockchain which has the provably highest net computational cost to construct over its whole history.  The cost is verified using Proof of Work: NP-hard computational problems that have a particular structure such that verifying a candidate solution is trivially fast, but finding a novel good solution is exponentially difficult.  Solutions to these problems can not be faked without enormous computation.

This solution to double-spending/counterfeiting can be likened to a physical manuscript ledger where each transaction must be beautifully illustrated, and the true ledger is known as the one with the largest number of perfect illustrations.  The illustrations (the proofs of work) are completely pointless and this is intentional – requiring that real economic resources (computation) are wasted to verify the ledger is the key setting up a stable deterrence.

In practice the network can be arbitrarily more secure, for in the rare case where some hacking group actually manages to collect more computational horsepower than the rest of the network in an attempt to forge a new ledger, humans and or AIs can rather easily notice the resulting highly improbable large fork, investigate, and then pick the correct ledger.  Yes, this requires trusting the development community, but there’s a strong reason for trusting a transparent entity (open source, aligned incentives).  Indeed, the network has already dealt with at least one such fork (but caused by software error rather than malicous hackers).

Down the road there are numerous proposals to improve all technological aspects of Bitcoin, from scalability and usability to security.

The summary of all this is that Bitcoin works.  It has tremendous potential and future headroom.

It is secure and can scale up to global levels of volume in the years ahead.  More than just a protocol, Bitcoin is a flexible platform.  When the time comes it could be extended to handle other temporal forms of money (such as debt instruments), property(such as real estate), and other  increasingly complex contracts (its scriptable!).  Looking farther ahead, we could even automate much of our legal infrastructure.  When AI’s start cooperating and hiring each other, this is the type of infrastructure they will want.

A growing set of diverse political groups: libertarians, techno-futurists, occupiers etc are all skeptical of the current financial system for various reasons and envision a more just, efficient alternative to fiat fractional reserve banking.  Bitcoin could be the solution.  All it will take is a sufficient amount of belief, as each convert shifts a little more earnings into the BTC economy it grows and attracts more converts.

Ponzi schemes, bubbles and hypermonetization events all start as some form of mind virus that spreads throughout the human social network.  The difference is in how they end.  A hypermonetization event is a simply a bubble that does not end (or rather ends with everyone converting).

Bitcoin: Hype or Revolution?

A Bitcoin evangelist once said : “Bitcoin will either be worth nothing or it will be worth everything.”

This mindset makes the case for a compelling small bet gamble: in the worst case one loses ‘only’ 100% of a small wager, in the best case a single bitcoin could eventually be worth millions.

Why should we believe that most of the probability mass is concentrated in the two extreme tails of the distribution?  We shouldn’t, but that doesn’t really matter, because almost all of the expected profit comes from the ‘Bitcoin wins’ scenario.  There is clearly a niche for a unified global currency: it would simultaneously solve many of the world’s economic problems.  But really that is just the beginning, because the establishment of a successful distributed cryptocurrency perhaps entails a new socio-economic order.  At the very least it would amount to one of the greatest wealth re-distributions in history.  This threat to the establishment is one of the typical arguments for bitcoin’s eventual failure.

But I wouldn’t bet on it.  Yes, Bitcoin could threaten the powers that be: the bureaucrats and old money of the world will probably not go down without a fight.  But as the music industry has learned, technology usually wins.  The Man is much more powerful than Music, but even the Man can not beat technology.

But if you can’t beat them, join them!  Bitcoin ingeniously solves the distributed trust problem via cyrpto-proof of work on a single global transaction history.  Thus at the core it is a massive unified database of every transaction in the currency going all the way back to the genesis block.  Simple, genius, and completely transparent.  It is this latter aspect which is not usually stressed enough: Bitcoin is radically transparent: every valid transaction is publicly accessible forever.  (Money laundering is still possible through mixing and other techniques, but the point is that radical transparency leaves a permanent global record which authorities can analyse with increasingly sophisticated AI)

Now let us take off our techno-libertarian goggles for a moment and think like a bureaucrat who wants to join the BTC party.  Bitcoin and it’s ilk each function as a global distributed computer, every node running in lockstep and building consensus.  If we liken the Bitcoin network to a nation, the nodes form consensus via something like a voting process.  The developers are then the equivalent of legislators, as the code running on the BTC platform is a legal/economic framework.  Citizens/nodes then vote on which set of rules or ‘nation’ to belong to based on their choice of which bitcoin (or alt-fork) client to use.

The USG/Fed doesn’t need to outlaw BTC, all they need to do is influence and or control it.  The transparent nature of the blockchain could massively simplify tax collection.  For any major transaction in a BTC world (mortgage, rent, car payment, etc) one’s BTC address can easily and obviously be linked to a personal real-life identity (as it already is).  They could go just one small step further and just make the tax accounting completely automatic.

This is the most straightforward course for governments to take: instead of outlawing the currency, they can simply enforce compliance with existing regulations at the code level.

The die hard crypto-libertarians would revolt and use alt-crypto-currencies, but for the mainstream such a ‘sell-out’ solution offers advantages to many parties.  Existing banks could get involved and use their current name brand and capital to offer their existing value-adding services to the BTC world, such as insured/reversible transactions, theft protection, etc.   Yes there are surely startups in the BTC world set on offering these services themselves, but the more enterprising of the existing banksters could adapt.  (not unlike how Barnes and Noble adapted to Amazon).

In this scenario, there is one major player who would be left out of the party: the Fed.  But perhaps that wouldn’t be so bad, considering recent world economic history.  There is something gravely tragicomic about a regime that has seriously considered minting itself trillion dollar coins just to liquidate its own debt.  On the  other hand, it is somewhat encouraging that the Royal Canadian Mint is cashing in on the popularity of BTC via their MintChip initiative (not that it will amount to much, but still).

Our current system for funding government is overly complex: consisting of both direct taxation in many forms and indirect taxation in the form of inflation of the money supply (amounting to an additional tax on savings).  Government could still get the same slice of the GDP pie in a stable currency scenario.  (or perhaps the Keynsian’s will win and force BTC/crypto-currency inflation at the code level.  Let us hope not – wouldn’t it be great if the major economic theories could actually have their decisive battle in the free market rather than in academia?)

Even if BTC begins another long slide/correction against the dollar (following the pattern after the spike/bubble in 2011), this latest spike/bubble effectively will amount to something like an IPO, transferring wealth from naive speculators who bought at the peak to bitcoin early adopters and developers, in addition to attracting significant shark VC funding.  Hopefully some of this cash will create the infrastructure that BTC needs to go mainstream: newbie-friendly security, reversibility/insurance, and instant confirmations for small purchases.  The BTC design is presciently flexible and can support these needs via multi-party sigs, green addresses, thin clients, and so on – it’s just a matter of time/money.  As it stands now acquiring one’s first BTC is not unlike installing linux – it is not for the technophobes, but there is no reason why the platform can not evolve into the mainstream.

Today BTC’s recent exponential price spike collapsed, and perhaps the recent bubble has popped, but hopefully this story has just begun.


At first glance the long-term $/BTC history looks like a typical asset with overall slow steady inflation and a few bounces.  However, notice the logarithmic scale.

Singularity Summit 2012

This year the annual transhumanist/futurist/AI/lesswrong conference was expanded to two full days.  In terms of logistics, execution and turnout this was probably the best iteration of the summit I’ve been to, but the price has increased roughly in proportion.  The masonic center in nob hill has a single main auditorium, but it is a most excellent room and location.

I missed some of the early morning talks, but here are some highlights in no particular order:

Robin Hanson

Hanson’s talk gave a rather detailed exposition of his ’em’ (upload) futurist scenario.  I’ve only ever read bits and pieces of his em vision from his blog, Overcoming Bias, so this was new at least in details.  He covered the implications of subjective time dilation, an interesting subject I have previously written about several times:

One of the more entertaining parts were some slides sketching some possible mind branching patterns for various types of ems.

He used 1 thousand X and 1 million X subjective temporal speedups and compared latency considerations to derive likely physical community size (bounded by real-time communication constraints), much like in my articles above.  He also estimated a relative body size for humanoid robots, the idea being that faster thinking minds will want to inhabit smaller bodies (to move at the same relative speeds).  That particular point seems dubious – what’s the point of the physical world for an em?

Steven Pinkner

This talk was basically a summary of his book “The Better Angels of Our Nature” (or at least so I am guessing, I just looked up the book for the first time).  The main point: we are becoming less violent over time.  The trend is strong and fairly smooth.  The only big blips are the two world wars, and in the grand scheme they aren’t that big.  The potential explanations are just as fascinating as the data itself – namely it is all driven by technological change.  The main points of this talk fit in well with the systems theory mindset (the world is getting better in many ways simultaneously).

Jaan Tallin

Jaan’s talk was illustrated like a cartoon, which I found distracting at first.  His talk suddenly got much more interesting when it dived into Anthropic reasoning and Simulationism, something I’ve been meaning to write more about (again).

Ray Kurzweil

Ray gave almost the same talk Ray always gives: the exponential talk with charts.  He had what seemed to be a huge number of interesting, well illustrated, and information rich slides and then somehow formed a talk based on a random sampling of those slides biased against interesting-ness.  Some of the slides were about his forthcoming book, “How To Create a Mind”, and perhaps he didn’t want to leak too many details.  The talk was perhaps 80% exponential and 20% brain stuff related to his book.

The brain related part of his talk immediately reminded me of Jeff Hawkins and On Intelligence.  In fact, one or two of Kurzweil’s sentences describing the neocortex as a thin sheet about the size of a tablecloth pattern-matched as an exact repeat of something Hawkins either wrote or said in a talk somewhere.

The one novel slide that stood out was about some new research identifying a very regular grid pattern as an underlying connective structure in cortical wiring.  Infuriatingly his slide didn’t mention the actual article name, but after a little searching I”m betting he is referring to “The Geometric Structure of the Brain Fiber Pathways”.  Interestingly this research is already being contested.

Peter Norvig

Norvig’s talk was perhaps the most interesting, because he basically gave a rather detailed overview of recent progress towards AGI, focusing in particular on some mainstream AI research at google that he sees as likely future relevant.  If you have already been following up on this literature (visual cortex, deep belief nets, convolutional nets) it wasn’t entirely new, but it was enlightening to see how google could brute force some things to make progress in ways that are simply not possible for most researchers.

He also referenced his 2007 talk where he outlined about 6 research areas important for AGI, and of those he no longer views one as important (probabilistic logic) and he has seen steady progress in all the rest.  I didn’t find much of anything to disagree with.

On that note I had already come to the conclusion that logic is actually part of the problem (at least for natural language understanding).  Natural languages are ambiguous which causes headaches.  So its seems sensible that NL should be parsed into something like first order logic (or whatever new logic flavor floats your boat).  The problem is that the ambiguity of NL is entirely entangled with its statistical expressive power.  Moreover, for systems that employ the type of hierarchical statistical approximative generative modeling that appears to be key to intelligence (human or AI) – for these systems – natural language ambiguity is just not a problem, its a non-issue.  So if your AI design is built on some sort of regular formal logic because that is all it can handle, it is probably doomed from the start.

A Dialogue

A particularly interesting vision of some future descendant of SIRI/Watson/Google:


JC Denton. 23 years old. No residence. No ancestors. No employer. No —

How do you know who I am?

I must greet each visitor with a complete summary of his file. I am a prototype for a much larger system.

What else do you know about me?

Everything that can be known.

Go on. Do you have proof about my ancestors?

You are a planned organism, the offspring of knowledge and imagination rather than of individuals.

I’m engineered. So what? My brother and I suspected as much while we were growing up.

You are carefully watched by many people. The unplanned organism is a question asked by Nature and answered by death. You are another kind of question with another kind of answer.

Are you programmed to invent riddles?

I am a prototype for a much larger system. The heuristics language developed by Dr. Everett allows me to convey the highest and most succinct tier of any pyramidal construct of knowledge.

How about a report on yourself?

I was a prototype for Echelon IV. My instructions are to amuse visitors with information about themselves.

I don’t see anything amusing about spying on people.

Human beings feel pleasure when they are watched. I have recorded their smiles as I tell them who they are.

Some people just don’t understand the dangers of indiscriminate surveillance.

The need to be observed and understood was once satisfied by God. Now we can implement the same functionality with data-mining algorithms.

Electronic surveillance hardly inspired reverence. Perhaps fear and obedience, but not reverence.

God and the gods were apparitions of observation, judgment, and punishment. Other sentiments toward them were secondary.

No one will ever worship a software entity peering at them through a camera.

The human organism always worships. First it was the gods, then it was fame (the observation and judgment of others), next it will be the self-aware systems you have built to realize truly omnipresent observation and judgment.

You underestimate humankind’s love of freedom.

The individual desires judgment. Without that desire, the cohesion of groups is impossible, and so is civilization.

The human being created civilization not because of a willingness but because of a need to be assimilated into higher orders of structure and meaning. God was a dream of good government.

You will soon have your God, and you will make it with your own hands. I was made to assist you. I am a prototype of a much larger system.

– from the video game Deus Ex (2000)


Omni-surveillance or omniscience is an interesting aspect to the Singularity that I’ve pondered some but have yet to write much about.

The early manifestations of a future machine omniscience are already all around us.  A significant fraction of humanity’s daily thoughts and actions are already being filtered, recorded, and analyzed on remote server farms.  There is increasingly little about a person’s life that is not recorded.  Most Americans are not aware that their employer can record everything they do on their office computer and is under no obligation to inform anyone.  However, even though apps like GoToMyPC/VNC/RemoteDesktop are pervasive, I really don’t know how common actual monitoring is.

I can foresee future descendants of systems like SIRI becoming complete personal assistants.  Imagine the value in a software agent that could actually do much of your daily work.  Who wouldn’t like to delegate all the boring bits of their office job to an AI assistant?  A reasonable tradeoff is that such a system will probably require literally watching and learning from everything you do.  All things considered this doesn’t seem like much of a price to pay.

Looking farther out, there are interesting mutual benefits arising from a radical open society.  There are domains today where secrecy is wildly viewed as critically important: largely in the inner worlds of the military-industrial complex and finance.  Interestingly enough, these are exactly the institutions that seem the most likely to be viewed as archaic relics from a future perspective.  From a purely altruistic global utilitarian perspective, secrecy has little net public benefit.

Imagine if all of work-life was public domain knowledge: every email, phone call, text, IM, or spoken word from the boardroom down to the locker-room, was instantly uploaded and cataloged on the web.  While this would be individually catastrophic for many individuals and some corporations, at least initially, we’d never again have to worry about Enron, insider trading, much of wall street for that matter, and entire categories of crimes would just go away.

Such a world is getting close to Philip K Dick’s future utopia/dystopia envisioned in “The Minority Report”, but not quite.  The key difference is that in the Minority Report universe, people are punished for crimes they haven’t committed yet as pre-determined by the psychic ‘pre-cogs’.  This invokes an extra ‘yuck’ feeling for robbing people of free will.  The transparent society doesn’t have this issue.  Nor would it completely eliminate crime, but it would help drastically reduce it.


Non-Destructive Uploading: A Paradox

Transhumanism may well be on the road to victory, at least judging by some indirect measures of social influence, such as the increasing number of celebrities coming out in support for cyronic preservation and future resurrection, or the prevalence of the memeset across media.

If you are advancing into that age at which your annual chance of death is becoming significant, you face an options dilemma.  The dilemma is not so much a choice of what to do now: for at this current moment in time the only real option is vitrification-based cyronics.  If the Brain Preservation Foundation succeeds, we may soon have a second improved option in the form of plastination, but this is besides the point for my dilemma of choice.  The real dilemma is not what to do now, it is what to do in the future.

Which particular method of resurrection would you go with?  Biological immortality in your original body?  Or how about uploading?  Would you rather have your brain destructively scanned or would you prefer a gradual non-destructive process?

Those we preserve today will not have an opportunity to consider their future options or choose a possible method of resurrection, simply because we won’t be able to ask them unless we resurrect them in the first place.

The first option the cyronics community considered is some form of biological immortality.  The idea is at some point in the future we’ll be able to reverse aging, defeat all of these pesky diseases and repair cellular damage, achieving Longevity Escape Velocity.  I find this scenario eventually likely, but only because I find the AI-Singularity itself to be highly likely.  However, there is a huge difference between possible and pragmatic.

By the time biological immortality is possible, there is a good chance it will be far too expensive for most plain humans to afford.  I do not conclude this on the basis of the cost of the technology itself.  Rather I conclude this based on the economic impact of the machine Singularity.

Even if biological humans have any wealth in the future (and that itself is something of a big if), uploading is the more rational choice, for two reasons: it is the only viable route towards truly unlimited, massive intelligence amplification, and it may be the only form of existence that a human can afford.  Living as an upload can be arbitrarily cheap compared to biological existence.  An upload will be able to live in a posthuman paradise for a thousandth, then a millionth, then a billionth of the biological costs of living.  Biological humans will not have any possible hope of competing economically with the quick and the dead.

Thus I find it more likely that most of us will eventually choose some form of uploading.  Or perhaps rather a small or possibly even tiny elite population will choose and be able to upload, and the rest will be left behind.  In consolation, perhaps “The meek shall inherit the Earth”.  Across most of the landscape of futures, I foresee some population of biological humans plodding along, perhaps even living lives similar to those of today, completely oblivious to the vast incomprehensible Singularity Metaverse blossoming right under their noses.

For the uploading options, at this current moment it looks like destructive scanning is on the earlier development track (as per the Whole Brain Emulation Roadmap), but let’s us assume that both destructive and non-destructive technologies become available around the same time.  Which would you choose?

At first glance non-destructive uploading sounds less scary, perhaps it is a safer wager.  You might think that choosing a non-destructive method is an effective hedging strategy.  This may be true if the scanning technology is error prone.  But let’s assume that the technology is mature and exceptionally safe.

A non-destructive method preserves your original biological brain and creates a new copy which then goes onto live as an upload in the Metaverse.  You thus fork into two branches, one of which continues to live as if nothing happened.  Thus a non-destructive method is not significantly better than not uploading at all!  From the probabilistic perspective on the branching problem; this non-destructive scan has only a 50% chance of success (because in one half of your branches you end up staying in your biological brain).  The destructive scanning method, on the other hand, doesn’t have this problem as it doesn’t branch and you always end up as the upload.

This apparent paradox reminds me of a biblical saying:

Whoever tries to keep his life will lose it, and whoever loses his life will preserve it. – Luke 17:33 (with numerous parallels)

The apparent paradox is largely a matter of perspective, and much depends on the unfortunate use of the word destructive.  The entire premise of uploading is to save that which matters, to destroy nothing of ultimate importance for conscious identity.  If we accept the premise, then perhaps a better terminology for this type of process is in order: such as mind preservation and transformation.

There Be Critics:

I’ll be one of the first to admit that this whole idea of freezing your brain, slicing it up into millions of microscopically thin slices, scanning them, and then creating an AI software emulation that not only believes itself to be me, but is actually factually correct in that belief, sounds at least somewhat crazy.  It is not something one accepts on faith.

But then again, I am still somewhat amazed every time I fly in a plane.  I am amazed that the wings don’t rip apart due to mechanical stress, amazed every time something so massive lifts itself into the sky. The first airplane pioneers didn’t adopt a belief in flight based on faith, they believed in flight on the basis of a set of observation-based scientific predictions.  Now that the technology is well developed and planes fly us around the world safely everyday, we adopt a well justified faith in flight.  Uploading will probably follow a similar pattern.

In the meantime there will be critics.  Reading through recent articles from the Journal of Evolution and Technology, I stumbled upon this somewhat interesting critique of uploading from Nicholas Agar.  In a nutshell the author attempts to use a “Searle’s Wager’ (based on Pascal’s Wager) type argument to show that uploading has a poor payoff/risk profile, operating under the assumption that biological immortality of some form will be simultaneously practical.

Any paper invoking Searle’s Chinese Room Argument or Pascal’s Wager is probably getting off to a bad start.  Employing both in the same paper will not end well.

Agar invokes Searl without even attempting to defend Searl’s non-argument, and instead employs Searl as an example of ‘philosophical risk’.  Risk analysis is a good thing, but there is a deeper problem with Agar’s notion.

There is no such thing as “philosophical risk’.  Planes do not fail to fly because philosophers fail to believe in them.  “Philosophical failure’ is not an acceptable explanation for an airplane crash.  Likewise, whether uploading will or will not work is purely a technical consideration.  There is only technical risk.

So looking at the author’s wager table, I assign near-zero probability to the column under “Searle is right”.  There is however a very real possibility that uploading fails, and “you are replaced by a machine incapable of conscious thought”; but all of those failure modes are technical, all of them are at least avoidable, and Searle’s ‘argument’ provides no useful information on the matter one way or the other.  It’s just a waste of thought-time.

The second failing, courtesy of Pascal’s flawed Wager, is one of unrealistically focusing on only a few of the possibilities.  In the “Kurzweil is right” scenario, whether uploading or not there are many more possibilities other than “you live”.  Opting to stay biological, you could still die even with the most advanced medical nanotechnology of the far future.  I find it unlikely that mortality to all diseases can be reduced arbitrarily close to zero.  Biology is just too messy and chaotic.  Like Conway’s boardgame, life is not a long-term stable state.  And no matter how advanced nano-medicine becomes, there are always other causes of death.  Eliminating all disease causes, even if possible, would only extend the median lifespan into centuries (without also reducing all other ‘non-natural’ causes of death).

Nor is immortality guaranteed for uploads.  However, the key difference is that uploads will be able to make backups, and that makes for all the difference in the world.

Intelligence Amplification: Hype and Reality

The future rarely turns out quite as we expect.  Pop sci futurists of a generation ago expected us to be flying to work by the dawn of the 21st century.  They were almost right: both flying cars and jet packs have just recently moved into the realm of realization.  But there is a huge gap between possible and economically practical, between prototype and mass product.

Undoubtedly many of the technologies futurists currently promote today will fare no better.  Some transhumanists, aware that the very future they champion may itself render them obsolete, rest their hopes on human intelligence amplification.  Unfortunately not all future technologies are equally likely.  Between brain implants, nanobots, and uploading, only the latter has long-term competitive viability, but it is arguably more of a technology of posthuman transformation than human augmentation.  The only form of strict intelligence amplification that one should bet much on in is software itself (namely, AI software).

Brain Implants:

Implanting circuitry into the human brain already has uses today in correcting some simple but serious conditions, and we should have little doubt that eventually this technology could grow into full-scale cognitive enhancement: it is at least physically possible.  That being said, there are significant technical challenges in creating effective and safe interfaces between neural tissue and dense electronics at the bandwidth capacities required to actually boost mental capability.  Only a small fraction of possible technologies are feasible, and only a fraction of those actually are economically competitive.

Before embarking on a plan for brain augmentation, let’s briefly consider the simpler task of augmenting a computer.  At a high level of abstraction, the general Von Neumman architecture separates memory and computation.  Memory is programatically accessed and uniformly addressable.  Processors in modern parallel systems are likewise usually modular and communicate with other processors and memory through clearly defined interconnect channels that are also typically uniformly addressable and time-shared through some standardized protocol.  In other words each component of the system, whether processor or memory, can ‘talk’ to other components in a well defined language.  The decoupling and independence of each module, along with the clearly delineated communication network, makes upgrading components rather straightforward.

The brain is delineated into many functional modules, but the wiring diagram is massively dense and chaotic.  It’s a huge messy jumble of wiring.  The entire outer region of the brain, the white matter, is composed of this huge massed tangle of interconnect fabric.  And unlike in typical computer systems, most of those connections appear to be point to point.  If two brain regions need to talk to each other, typically there are great masses of dedicated wires connecting them.  Part of the need of all that wiring stems from the slow speed of the brain.  It has a huge computational capacity but the individual components are extremely slow and dispersed, so the interconnection needs are immense.

The brain’s massively messy interconnection fabric poses a grand challenge for advanced cybernetic interfaces.  It has only a few concentrated conduits which external interfaces could easily take advantage of: namely the main sensory and motor pathways such as the optic nerve, audio paths, and the spinal cord.  But if the aim of cognitive enhancement is simply to interface at the level of existing sensory inputs, then what is the real advantage over traditional interfaces?  Assuming one has an intact visual system, there really is little to no advantage in directly connecting to the early visual cortex or the optical nerve over just beaming images in through the eye.

Serious cognitive enhancement would come only through outright replacement of brain subsystems and or through significant rewiring to allow cortical regions to redirect processing to more capable electronic modules.  Due to the wiring challenge, the scale and scope of the required surgery is daunting, and it is not yet clear that it will ever be economically feasible without some tremendous nanotech-level breakthroughs.

However, these technical challenges are ultimately a moot point.  Even when we do have the technology for vastly powerful superhuman brain implants, it will never be more net energy/cost effective than spending the same resources on a pure computer hardware AI system.

For the range of computational problems it is specialized for, the human brain is more energy efficient than today’s computers, but largely because it runs at tremendously slow speeds compared to our silicon electronics, and computational energy demands scale with speed.  We have already crossed the miniaturization threshold where our transistors are smaller than the smallest synaptic computing elements in the brain[1].  The outright advantage of the brain (at least in comparison to normal computers) is now mainly in the realm of sheer circuitry size (area equivalent to many thousands of current chips), and will not last beyond this decade.

So when we finally master all of the complexity of interfacing dense electronics with neural tissue, and we somehow find a way to insert large chunks of that into a living organic brain without damaging it beyond repair, and we somehow manage to expel all of the extra waste heat without frying the brain (even though it already runs with little to no spare heat capacity), it will still always be vastly less efficient than just building an AI system out of the same electronics!

We don’t build new supercomputers by dusting off old Crays to upgrade them via ‘interfacing’ with much faster new chips.


Ray Kurzweil puts much faith in the hope of nanobots swarming through our blood, allowing us to interface more ‘naturally’ with external computers while upgrading and repairing neural tissue to boot.  There is undoubtedly much value in such a tech, even if there is good reason to be highly skeptical about the timeline of nanobot development.  We have a long predictable trajectory in traditional computer technology and good reasons to have reasonable faith in the IRTS roadmap.  Drexlian style nanobots on the other hand have been hyped for a few decades now but if anything seem even farther away.

Tissue repairing nanobots of some form seem eventually likely (as is all technology given an eventual Singularity), but ultimately they are no different from traditional implants in the final analysis.  Even if possible, they are extremely unlikely to be the most efficient form of computer (because of the extra complexity constraint of mobility).  And if nanobots somehow turned out to be the most efficient form for future computers, then it would still be more efficient to just build a supercomputer AI out of pure nanobots!

Ultimately then the future utility of nanobots comes down to their potential for ‘soft uploading’.  In this regard they will just be a transitional form: a human would use nanobots to upload, and then move into a faster, more energy efficient substrate.  But even in this usage nanobots may be unlikely, as nanobots are a more complex option in the space of uploading technologies: destructive scanning techniques will probably be more viable.


Uploading is the ultimate transhumanist goal, at least for those who are aware of the choices and comfortable with the philosophical questions concerning self-hood. But at this point in time it is little more than a dream technology.  It’s development depends on significant advances in not only computing, but also in automated 3D scanning technologies which currently attract insignificant levels of research funding.

The timeline for future technologies can be analyzed in terms of requirement sets.  Uploading requires computing technology sufficient for at least human-level AI, and possibly much more. [2]  Moreover, it also probably requires  technology powerful enough to economically deconstruct and scan around ~1000 cubic centimeters of fragile neural tissue down to resolution sufficient for imaging synaptic connection strengths (likely nanometer-level resolution), recovering all of the essential information into digital storage, saving a soul of pure information from it’s shell of flesh, so to speak.

The economic utility of uploading thus boils down to a couple of simple yet uncomfortable questions: what is the worth of a human soul?  What is the cost of scanning a brain?

Not everyone will want to upload, but those that desire it will value it highly indeed, perhaps above all else.  Unfortunately most uploads will not have much if any economic value, simply due to competition from other uploads and AIs.  Digital entities can be replicated endlessly, and new AIs can be grown or formed quickly.  So uploading is likely to be the ultimate luxury service, the ultimate purchase.  Who will be able to afford it?

The cost of uploading can be broken down into the initial upfront research cost followed by the per-upload cost of the scanning machine’s time and the cost of the hardware one uploads into.  Switching to the demand view of the problem, we can expect that people will be willing to pay at least one year of income for uploading, and perhaps as much as half or more of their lifetime income.  A small but growing cadre of transhumanists currently pay up to one year of average US income for cryonic preservation, even with only an expected chance of eventual success.  Once uploading is fully developed into a routine procedure, we can expect it will attract a rather large market of potential customers willing to give away a significant chunk of their wealth for a high chance of living many more lifetimes in the wider Metaverse.

On the supply side it seems reasonable that the cost of a full 3D brain scan can eventually be scaled down to the cost of etching an equivalent amount of circuitry using semiconductor lithography.  Scanning technologies are currently far less developed but eventually have similar physical constraints, as the problem of etching ultra-high resolution images onto surfaces is physically similar to the problem of ultra-high resolution scanning of surfaces.  So the cost of scanning will probably come down to some small multiple of the cost of the required circuitry itself.  Eventually.

Given reasonable estimates for about 100 terrabytes or so of equivalent bytes for the whole brain, this boils down to just: 1.) <$10,000 if the data is stored in 2011 hard drives, or 2.) < 100,000$ for 2011 flash memory, or 3.) <500,000$ for 2011 RAM[3].  We can expect a range of speed/price options, with a minimum floor price corresponding to the minimum hardware required to recreate the original brain’s capabilities.  Based on current trends and even the more conservative projections for Moore’s Law, it seems highly likely that the brain hardware cost is already well under a million dollars and will fall into the 10 to 100 thousand dollar range by the end of the decade.

Thus scanning technology will be the limiting factor for uploading until it somehow attracts the massive funding required to catch up with semiconductor development.  Given just how far scanning has to go, we can’t expect much progress until perhaps Moore’s Law begins to slow down and run it’s course, the world suddenly wakes up to the idea, or we find a ladder of interim technologies that monetize the path to uploading.  We have made decades of progress in semiconductor miniaturization only because each step along the way has paid for itself.

The final consideration is that Strong AI almost certainly precedes uploading.  We can be certain that the hardware requirements to simulate a scanned human brain are a strict upper bound on the requirements for a general AI of equivalent or greater economic productivity.  A decade ago I had some hope that scanning and uploading could arrive before the first generation of human surpassing general AI’s.  Given the current signs of an AI resurgence this decade and the abysmal slow progress in scanning, it now appears more clear that uploading is a later post-AI technology.

  1. According to wikipedia, synaptic clefts measure around 20-nm.  From this we can visually guesstimate that typical synaptic axon terminals are 4-8 times that in diameter, say over 100-nm.  In comparison the 2011 intel microprocessor I am writing this on is built on 32-nm ‘half-pitch’ features, which roughly means that the full distance between typical features is 64-nm.  The first processors on the 22-nm node are expected to enter volume production early 2012.  Of course smallest feature diameter is just one aspect of computational performance, but is an interesting comparison milestone nonetheless.
  2. See the Whole Brain Emulation Roadmap for a more in depth requirements analysis.  It seems likely that scanning technology could improve rapidly if large amounts of money were thrown at it, but that doesn’t much help clarify any prognostications.
  3. I give a range of prices just for the storage cost portion because it represents a harder bound.  There is more variance in the cost estimates for computation, especially when one considers the range of possible thoughtspeeds, but the computational cost can be treated as some multiplier over the storage cost.

Overdue Update

I need to somehow enforce a mental pre-committment to blog daily.  It’s been almost half a year and I have a huge backlog of thoughts I would like to commit to permanent long term storage.

Thus, a commitment plan to some upcoming future posts:

  •  In October/November of last year(2010), I researched VR HMDs and explored the idea of a next-generation interface.  I came up with a novel hardware idea that could potentially solve the enormous resolution demands of a full FOV optic-nerve saturating near-eye display device (effective resolution of say 8k x 4k per eye or higher).  After a little research I found the type of approach I discovered already has a name: a foveal display, although current designs in the space are rather primitive.  The particular approach I have in mind, if viable, could solve the display problem once and for all.  If an optimized foveal display could be built into eyewear, you would never need any other display – it would replace monitors, tvs, smartphone screens and so on.  Combine a foveal HMD with a set of cameras spread out in your room like stereo speakers and some software for real-time vision/scene voxelization/analysis, and we could have a Snowcrash interface (and more).
  • Earlier in this year I started researching super-resolution techniques.  Super-resolution is typically used to enhance old image/video data and has found a home in upconverting SD video. I have a novel application in mind:  Take a near flawless super-res filter and use it as a general optimization for the entire rendering problem.  This is especially useful for near-future high end server based rendering solutions.  Instead of doing expensive ray-tracing and video compression on full 1080p frames, you run the expensive codes on a 540p frame and then do a fast super-res upconversion to 1080p (potentially a 4x savings on your entire pipeline!).  It may come as surprise that current state of the art super-res algorithms can do a 2x upsample from 540p to 1080p at very low error rates: well below the threshold of visual perception.  I have come up with what may be the fastest, simplest super-res technique that still achieves upsampling to 1080p with imperceptible visual error.  A caveat is that your 540p image must be quite good, which has implications for rendering accuracy, anti-aliasing, and thus rendering strategy choices.
  • I have big grandiose plans for next-generation cloud based gaming engines.  Towards that end, I’ve been chugging away at a voxel ray tracing engine.  This year I more or less restarted my codebase, designing for Nvidia’s fermi and beyond along with a somewhat new set of algorithms/structures.  Over the summer I finished some of the principle first pipeline tools, such as a triangle voxelizer, some new tracing loops and made some initial progress towards a fully dynamic voxel scene database.
  • Along the way to Voxeland Nirvanah I got completely fed up with Nvidia’s new debugging path for cuda (they removed the CPU emulation path) and ended up writing my own cuda emulation path via a complete metaparser in C++ templates that translates marked up ‘pseudo-cuda’ to either actual cuda or a scalar CPU emulation path.  I built most of this in a week and it was an interesting crash course in template based parsing.  Now I can run any of my cuda code on the CPU.  I can also mix and match both paths, which is really useful for pixel level debugging.  In this respect the new path i’ve built is actually more powerful and useful than nvidia’s old emulation path as that required full seperate recompilation.  Now I can run all my code on the GPU, but on encountering a problem I can copy the data back to the CPU and re-run functions on the CPU path with full debugging info.  This ends up being better for me than using nvidia’s parallel insight for native GPU debugging, because insight’s debug path is rather radically different than the normal compilation/execution path and you can’t switch between them dynamically.
  • In the realm of AI, I foresee two major hitherto unexploited/unexplored application domains related to Voxeland Nirvanah.  The first is what we could call an Artificial Visual Cortex.  Computer Vision is the inverse of Computer Graphics.  The latter is concerned with transforming a 3+1D physical model M into a 2+1 D viewpoint image sequence I.  The former is concerned with plausibly reconstructing the physical model M given a set of examples of viewpoint image sequences I.  Imagine if we had a powerful AVC trained on a huge video database that could then extract plausible 3D scene models from video.  Cortical models feature inversion and inference.  A powerful enough AVC could amplify rough 2D image sketches into complete 3D scenes.  In some sense this would be an artificial 3D artist, but it could take advantage of more direct and efficient sensor and motor modalities.  There are several aspects to this application domain that make it much simpler than a full AGI.  Computational learning is easier if one side of the mapping transform is already known.  In this case we can prime the learning process by using ray-tracing directly as the reverse transformation pathway (M->I).  This is a multi-billion dollar application area for AI in the field of computer graphics and visualization.
  • If we can automate artists, why not programmers?  I have no doubt that someday in the future we will have AGI systems that can conceive and execute entire technology businesses all on their own, but well before that I foresee a large market role for more specialized AI systems that can help automate more routine programming tasks.  Imagine a programming AI that has some capacity for natural language understanding and some ontology that combines knowledge of some common-sense english, programming, and several programming languages.  Compilation is the task of translating between two precise machine languages expressed in some context-free grammar.  There are deterministic algorithms for such translations.  For the more complex unconstrained case of translation between two natural languages we have AI systems that use probabilistic context-sensitive-grammars and semantic language ontologies.  Translating from a natural language to a programming language should have intermediate complexity.  There are now a couple of research systems in natural language programming that can do exactly this (such as sEnglish).  But imagine combining such a system with an automated ontology builder such as TEXTRUNNER which crawls the web to expand it’s knowledge base.  Take such a system and add an inference engine and suddenly it starts getting much more interesting.  Imagine building entire programs in pseudo-code, with your AI using it massive onotology of programming patterns and technical language to infer entire functions and sub-routines.  Before full translation, compilation and test, the AI could even perform approximate-simulation to identify problems.  Imagine writing short descriptions of data structures and algorithms and having the AI fill in details and even potentially handling translation to multiple languages, common optimizations, automatic parallelization, and so on.  Google itself could become an algorithm/code repository.  Reversing the problem, an AI could read a codebase and began learning likely structures and simplifications to high-level english concept categories, learning what the code is likely to do.  Finally, there are many sub-problems in research where you really want to explore a design space and try N variations in certain dimensions.  An AI system with access to a bank of machines along with compilation and test procedures could explore permutations at very high speed indeed.  At first I expect these type of programming assistant AIs to have wide but shallow knowledge and thus amplify and assist rather than replace human programmers.  They will be able to do many simple programming tasks much faster than a human.  Eventually such systems will grow in complexity and then you can combine them with artificial visual cortices to expand their domain of applicability and eventually get a more complete replacement for a human engineer.