Tag Archives: Ems

Em City By Combo Auction

By Robin Hanson · December 1, 2011 3:00 pm · Comments (13)

Yesterday I outlined how combinatorial auctions could help our cities better coordinate their land use and utility capacity, without granting great discretion to a central power. But I ended with:

It would be very hard to get agreement to change to this system from today’s system of property rights and regulatory restrictions. I despair of it happening in our comfortable and change-averse cities. So we might have to wait until a big disruption creates lots of other change. (more)

Two years ago I pointed to a big-enough future disruption:

Rich stable nations … feel little inclination to consider big disruptive changes. … This frustrates rich-nation would-be-rebels like me who see our business, legal, political, etc. institutions as far from optimal. … If you long to say “come the revolution,” you might wait three to fifteen decades for the “em rev“, the whole brain emulation revolution. …

Rapid [em] growth will require huge rapid changes in economic organization, and supporting changes to business, legal, and political institutions. … Locations vying to be one of those [first em] centers may be open to big institutional change. … So if you have a favorite radical change you’d like the world to consider, you might give some thought to how your change could support a local em rev. (more)

The first em cities may be especially open to change regarding how cities are run. How might combinatorial auctions help them?

Here are my best guesses about (mid-em-era) em cities. City centers would mainly house computers, mostly running brains, and supporting infrastructure, e.g., power, cooling, structural support, part swapping paths, security, leakage containment, etc.

City centers would mostly house ems in virtual bodies doing office work, meeting often with other city workers. In most meetings, brains would stay put and just send signals; physical movement would be much rarer. Em minds would be sped-up relative to human minds as far as possible, until doubling an em’s mental speed much more than doubled its computing costs.

Outside of city centers there would be more ems in physical bodies, mostly small, helping with physical activities such as mining, harvesting, manufacturing, transportation, dumping, etc. Air cooling in the periphery would give way to water cooling closer in, and perhaps molten salt cooling very close.

All this would put a huge premium on inner city computer speed, density, and bandwidth. Cities would be very 3D, and city center computers would likely have very small physical structures generating lots of heat, making cooling crucial. Also important would be power sources, and physical paths for the replacement of devices and parts.

Today big computing centers are centrally planned, mostly with uniform parts and regular structures. But this level or coordination is may be infeasible for large cities, where diverse organizations make coordination expensive and change hodge-podge. In such a context, combinatorial auction might help improve coordination.

In am em city combinatorial auction, bids for locations could specify:

  1. spatial volume, shape, and orientation
  2. part swapping portal locations and sizes
  3. line of sight to outside, or to specific parties
  4. surface temperature and chemical corrosively limits
  5. amount and form of power and cooling, with price limits
  6. specific chemicals piped in, fluid garbage piped out
  7. communication distance from other particular residents
  8. time delay and expense to move hardware out and in
  9. support force tensors (including weight) get, support can give
  10. max stress-strain to support during earthquake
  11. limits on incoming, outgoing vibration distributions
  12. chances of incoming, limits on outgoing, leakage
  13. chance of explosive destruction, correlation with distant backups
  14. legal rules covering disputes with neighbors
  15. time commitments on each of these, and penalties for violations

As with cities today, winning allocations would say who gets what spaces with what supporting utilities, limits, etc. Competitive utility suppliers could also bid their prices to use particular spaces to supply particular utility amounts to particular locations. and futures markets about future winning bids might help estimate opportunity costs of commitment. Auction revenue could pay for utility fixed costs and repay city investors, and futarchy might choose the basic auction rules.

Yes, there’s a lot we don’t know about the future, and I could get some things wrong here. Even so, it seems worth thinking about what the future might be like, and when big institutional changes might be feasible.

GD Star Rating
loading...
Tagged as: , , ,

Kurzweil Rejects Ems

By Robin Hanson · October 23, 2011 3:00 pm · Comments (2)

I posted recently on Allen & Greaves criticizing “the whole brain emulation argument that we can simulate a brain without understanding it.” Ray Kurzweil responds that while he is far more optimistic on AI progress, he doesn’t believe in emulation without understanding either:

Allen mischaracterizes my proposal to learn about the brain from scanning the brain to understand its fine structure. It is not my proposal to simulate an entire brain “bottom up” without understanding the information processing functions. We do need to understand in detail how individual types of neurons work, and then gather information about how functional modules are connected. The functional methods that are derived from this type of analysis can then guide the development of intelligent systems. Basically, we are looking for biologically inspired methods that can accelerate work in AI.

It makes sense that since Kurzweil is so optimistic about rapid progress in so many technologies, such as life extension, he’d be optimistic about rapid progress in modeling the higher level organization of brains. Ems seem more likely to pessimists like myself — although we think emulation should be possible with far less than quantum chemistry detail, since the brain is a robust signal processing system, we estimate that the rate of progress to date suggests a long slow road to understanding brain organization.

GD Star Rating
loading...
Tagged as: , ,

The Future Of Cities

By Robin Hanson · October 18, 2011 11:10 pm · Comments (20)

What sets city size? That is, what determines how many people all cluster together in an urban area? On the one hand, city size increases with feasible building height and with the gains to people and businesses from interacting closely with many others. On the other hand, city size decreases with how much space folks want, and with costs to transport people and goods within a city and from outlying regions. City size increases with more and cheaper nearby non-city economic activity.

Policy also matters; poor governance and positive externalities of density reduce city size, while being the center of government for a surrounding area increases city size. The size of big cities should be limited by the reluctance of nations to let their city activity be absorbed into big nearby foreign cities. And sunk costs and coordination failures can long delay the adaptation of city sizes and locations to changing circumstances.

In the farming era, cities held only a small fraction of the population, and so their size and locations were determined mainly by nearby farming activity. However, when most folks live in cities, then nearby non-city activity matters less, and decreasing transport costs make bigger cities more economical.

So how well have today’s city size and locations adapted to industry era tradeoffs? That is, how well do cities today trade the gains from more interaction in bigger cities for the added costs of transport and reduced personal space? While we expect optimal industry era cities to be more concentrated, i.e., fewer and larger, we also expect inertia, coordination failures, density externalities, and city mismanagement to slow the transition from an ideal farming era distribution of cities to an ideal industry era distribution. So cities today are probably too many and small. But how far off are they?

One clue – alas one that that is hard to interpret – is that today (log) city size follows a normal distribution, and (log) size changes follow a random walk. Another more informative clue is that in many large nations, a big (but not too big) fraction of the urban population is in the largest city. For example: South Korea 53%, Japan 44%, Egypt 43%, Argentina 37%, Bangladesh 34%, Philippines 28%, Mexico 24% (sources here, here). This weakly suggests that such cities might be running up against a political limit – the reluctance of neighboring nations to let these cities absorb their city activity.

How should we expect cities to change in a future em era, where trillions of human emulations live in virtual reality or in tiny android bodies? Since ems are easier to transport, require less space, and interact less with rural areas, optimal em cities should be even more concentrated than industry cities. Especially if ems learn to better subsidize density, to internalize today’s density externality. And since ems require quite different infrastructure from humans, and need large and rapid changes that most cities will initially be unwilling to allow, existing industry era cities may less constrain the size and location of em cities.

Together these suggest that em cities might be quite a bit more concentrated than our industry cities. Most ems might live within a half dozen or fewer really huge cities. Which would imply that only a half dozen nations would have substantial political power, allowing for easier global coordination.

If optimal em city concentration is really high, most ems might even live in just one biggest city. An analogy in the history of brains seems apt. Some of the first brains were spread out all over animal bodies, but then brains evolved to concentrate in one small region, to minimize signal delays within the brain.

Of course one big em city could be vulnerable to bad governance, so perhaps the biggest city would change as biggest cities became badly managed. Especially if ems had better ways (e.g. prediction markets) to coordinate their city switching activities. Creates an interesting picture of a competitive world government – at any one time most world economic activity might be under a single central city government, and yet cities might compete to offer the best world governance.

GD Star Rating
loading...
Tagged as: , ,

A Theory Of Status

By Robin Hanson · October 14, 2011 8:00 am · Comments (10)

A few days ago I asked for a theory of status, to help predict how status will change in a rather different future. Today let me offer such a theory.

Here are our main clues about status:

  1. Status is a socially shared way to evaluate and rank people
  2. Status seems mostly relative; you can’t raise everyone’s status
  3. Human status has two main parts: dominance and prestige
  4. Most animals have dominance, which is who would win a pairwise conflict
  5. Prestige status seems to not exist in animals with simple social relations.
  6. Unlike other shared rankings, like sexiness or dominance, we seem unaware of what exactly prestige ranks.

So what is prestige? That is, what sort of ranking would be useful for human-like primates to track about each other, but also be something illicit, so that foragers would be reluctant to admit its true function? One obvious candidate stands out to me: one’s value as an ally in coalition politics. That is, how much better off is a typical coalition with this person as an ally, relative to not having them as an ally. This is clearly an important concept, well worth tracking. It only makes sense in groups with complex coalition politics, and foragers have norms against overtly engaging in such politics.

Since an ability to win pairwise contests is useful to coalitions, we expect dominance to add to prestige. But humans and similar primates can also add value to a coalition by having skills that make them useful associates, and by being on good terms with other good-ally-material folks. And both skills and associations also seem to make important contributions to human prestige. Note that this theory predicts that other primates with complex coalition politics, like chimps, will also have a prestige status distinct from dominance status.

If prestige is about one’s value in coalition politics, what does that predict about em prestige? Of the list I gave, items 2,5,7,12,16, should be substantially related to prestige.

GD Star Rating
loading...
Tagged as: , , ,

Allen & Greaves On Ems

By Robin Hanson · October 13, 2011 4:50 pm · Comments (29)

Paul Allen and Mark Greaves say the “singularity” is over a century away:

This prior need to understand the basic science of cognition is where the “singularity is near” arguments fail to persuade us. …. A fine-grained understanding of the neural structure of the brain … has not shown itself to be the kind of area in which we can make exponentially accelerating progress. … By the end of the century, we believe, we will still be wondering if the singularity is near.

But what about the whole brain emulation argument that we can simulate a brain without understanding it? They say:

For example, if we wanted to build software to simulate a bird’s ability to fly in various conditions, simply having a complete diagram of bird anatomy isn’t sufficient. To fully simulate the flight of an actual bird, we also need to know how everything functions together. In neuroscience, there is a parallel situation. Hundreds of attempts have been made (using many different organisms) to chain together simulations of different neurons along with their chemical environment. The uniform result of these attempts is that in order to create an adequate simulation of the real ongoing neural activity of an organism, you also need a vast amount of knowledge about the functional role that these neurons play, how their connection patterns evolve, how they are structured into groups to turn raw stimuli into information, and how neural information processing ultimately affects an organism’s behavior. Without this information, it has proven impossible to construct effective computer-based simulation models.

This seems confused. No doubt a detailed enough emulation of bird body motions would in fact fly. It is true that a century ago our ability to create detailed bird body simulations was far less than our ability to infer abstract principles of flight. So we abstracted, and built planes, not bird emulations. But this hardly implies that brains must be understood abstractly before they can be emulated.

Yes you need to understand a system well in order to know what details you can safely leave out and still achieve the same overall functions. But if you can afford to leave in all the details, you don’t have to understand what is safe to leave out. We apply this principle every time we play a song or movie. Since we know that a song or movie recording contains enough detail to reproduce a full sound or visual experience, we don’t have to understand a song or movie in order to be able to replay it for someone, and achieve most of the relevant artistic experience.

Projecting trends like Moore’s law suggests that our ability to simulate low level brain processes should increase by fantastic factors within a century. These factors seem plenty sufficient to model entire brains at low levels of detail. So if we have not understood brains well enough by then to know what details we can safely leave out, we should be able to reproduce their behavior via brute-force simulation of lots of raw detail.

Added 10p: As I explained in January:

We should expect brain emulation to be feasible because brains function to process signals, and the decoupling of signal dimensions from other system dimensions is central to achieving the function of a signal processor.

GD Star Rating
loading...
Tagged as: ,

What Status Ems?

By Robin Hanson · October 12, 2011 6:50 pm · Comments (16)

While many things count for status in our society, most of us have a rough idea of their relative weight, at least for common evaluations. But we understand the origins of these weights poorly. This ignorance seems especially clear when we consider how status might change in the future. For example, I’ve been pondering the scenario of a future world dominated by ems (whole brain emulations), and realize that it seems especially unclear what would count more for status among ems. Some possibilities:

  1. Pure physical size or power
  2. Impressiveness in conversation or verbal sparring
  3. How well its personality embodies the ideals of its age
  4. How mental, complex, or abstract is its job
  5. # other statusful ems this em commands or controls
  6. The accomplishments of this copy, since its last split.
  7. # other high status ems know it personally
  8. Political influence of this em in local disputes
  9. Personal em wealth
  10. Current daily wages
  11. Current daily profit, of wages minus cost to exist
  12. The status its human had in the pre-em world
  13. Total time this mind has experienced subjectively
  14. Time expected until em forced to retire/archive
  15. # active copies expected of this em at a future date
  16. # active “clan” copies, all of the same pre-em human
  17. # active copies expected of this clan at a future date
  18. The total accomplishments of the entire copy clan
  19. # other high status ems know anyone in clan
  20. Total wealth of its copy clan
  21. Total potential political influence of its copy clan

Of course if we had a good theory of status, we could use that to predict future status. For example, if status were a measure of future evolutionary success, then #15 would make sense. But if status were instead a measure of the value of an ally in local coalition politics, then #8 would make more sense.

Many of these measures (e.g., #14) could produce an abrupt change in status when a new copy is created. Do abrupt status changes make sense when others’ opinions about you haven’t changed?

GD Star Rating
loading...
Tagged as: ,

Shulman On Superorgs

By Robin Hanson · October 4, 2011 9:45 pm · Comments (15)

It has come to my attention that some think that by now I should have commented on Carl Shulman’s em paper Whole Brain Emulation and the Evolution of Superorganisms. I’ll comment now in this (long) post.

The undated paper is posted at the Singularity Institute, my ex-co-blogger Eliezer Yudkowsky’s organization dedicated to the proposition that the world will soon be ruled by a single powerful mind (with well integrated beliefs, values, and actions), so we need to quick figure out how to design values for a mind we’d like. The main argument is that someone will soon design an architecture to let an artificial mind quickly grow from seriously stupid to super wicked smart. (Yudkowsky and I debated that recently.) Shulman’s paper offers an auxiliary argument, that whole brain emulations would also quickly lead to one or a few such powerful integrated “superorganisms.”

It seems to me that Shulman actually offers two somewhat different arguments, 1) an abstract argument that future evolution generically leads to superorganisms, because their costs are generally less than their benefits, and 2) a more concrete argument, that emulations in particular have especially low costs and high benefits. Continue reading "Shulman On Superorgs" »

GD Star Rating
loading...
Tagged as: , ,

A Galaxy On Earth

By Robin Hanson · July 28, 2011 9:00 am · Comments (34)

Our galaxy has about three hundred billion stars, and Earth today has about seven billion people. Assuming only half as many useable planets as stars, we could combine these two numbers into an initial crude guess for the size of a galactic civilization, and define a “galaxy of people” to be a thousand billion billion (or 1021) people. Now consider some famous galactic civilizations in science fiction.

One of the most popular science fiction stories ever was Issac Azimov’s Foundation series. It tells of the fall and rise of a galaxy-wide civilization, whose capital, Trantor, was a planet-wide city a kilometer deep into the ground. Trantor’s population was said to be forty billion, in a galaxy with millions of populated planets and a total population of a million billion (or one millionth of a “galaxy” as defined above).

Star War‘s Coruscant is also a planet-wide city and capital of a galaxy wide civilization, with planetary population of a thousand billion, in a galaxy also of millions of planets and a total population of a million billion. Some say Coruscant’s buildings averaged two kilometers tall. In Star Trek‘s Federation of 150 planets a few centuries hence, which controlled a few percent of the galaxy, each planet had no more than about our Earth’s seven billion, though some say the Federation held ten thousand billion people.

These all seem like dramatic underestimates to me. If Earth were paved over with a city the density of Manhattan today (1.6 million in 59 square kilometers), Earth would have a population of 14 thousand billion. Since Manhattan now has an average building height of 25 meters, a two kilometer deep version could hold a million billion people, and a two thousand kilometer deep version (Earth’s radius is 6400km) could hold a billion billion people.

There is roughly another thousand times as much useable material nearby, in other planets, comets, and the sun itself, allowing a solar-system population perhaps a thousand times larger. This brings us to a thousand billion billion, or a “galaxy” of people, the same as my initial crude population estimate for an entire galaxy above, and vastly larger than most science fiction galaxy estimates.

Furthermore, android ems (whole brain emulations in simulated bodies) could take up a lot less space than humans. I once somewhat conservatively estimated that an em might stand at 1% of human height (and run one hundred times faster). Since such an em would take up only one millionth of a human’s volume, a two kilometer deep Earth city could hold a “galaxy” (or thousand billion billion) of ems. And a solar system civilization might fit a billion billion billion ems, or a million “galaxies.”

Of course we have a long long way to go, not only to generate such huge populations, but also to develop the energy, manufacturing, heat-dumping, etc techs to allow us to support them. And yes, eventually we would run out of energy and material near our Sun, and need to go elsewhere to grow.

But we have strong economic reasons to stay close to one another as long as there is enough energy and material nearby, and especially as long as we continue to innovate. So most of our descendants’ economy should stay close to our sun until congestion here gets severe. We may well have a solar system population of a billion billion billion before the time comes when most of our descendants are closer to other stars.

Most science fiction seems to vastly underestimate the population that a single planet or star can hold, and the strength of the economic pressures to keep an economy close together, rather than spread across vast distances. Someday we will learn to tell stories that treat planets and stars as the vast spaces of possibilities that they really are.

Added 11a: Even an unmodified sun radiates enough energy to cover the calorie consumption of over a hundred “galaxies” of humans, and far more ems.

The timescale to grow from today’s population to a “galaxy” of descendants would be 600 years at an industry-style 15 year doubling time, 40 millennia at a farming-style thousand year doubling time, and four years at at next-singularity-style monthly doubling time.

GD Star Rating
loading...
Tagged as: , , , ,

Upload Skepticism

By Robin Hanson · July 14, 2011 10:20 pm · Comments (23)

David Linden in Boing Boing:

Kurzweil predicts that by the late 2030s, we will be able to routinely scan an individual’s brain with such molecular precision and with such a complete understanding of the rules underlying neuronal function and plasticity that we will be able to “upload” our mental life into a vastly powerful and capacious future computer. … I am a neurobiologist and I have spent the past 28 years engaged in studies of the cellular and molecular basis of memory and cognition. I am an optimist and a technophile, but I believe that I speak for the vast majority of brain researchers when I express serious doubts about Kurweil’s timetable. …

Kurzweil then argues that our understanding of biology—and of neurobiology in particular—is also on an exponential trajectory, driven by enabling technologies. … At some point in the 2020s, a miracle will occur: If we keep accumulating data about the brain at an exponential rate (its connection maps, its activity patterns, etc.), then the long-standing mysteries of development, consciousness, perception, decision, and action will necessarily be revealed. … That’s where I get off the bus.

Our understanding of biological processes remains on a stubbornly linear trajectory. … There have been a number of genuine paradigm-shifting insights in genetics in recent years. … But these discoveries, and most of the other key conceptual breakthroughs in this field, have come slowly, the result of stubbornly linear small science, and not of the huge technology-driven data sets that Kurzweil describes. … This linear progress also holds true for the growth in our knowledge of brain function. … The ploddingly linear increase in our understanding of neural function means that an idea like mind-uploading to machines being usefully deployed by the 2020s or even the 2030s seems overly optimistic. (more; HT Tyler)

I’m happy to defer to Linden’s brain science expertise. But I wish he’d get clear on two key points:

  1. All this talk of linear vs. exponential progress, with linear progress unable to finish by 2040, suggests Linden has in mind a rough estimate of how far along are we now, and how fast we have been proceeding. It would be very helpful if Linden would tell us his best guess. For example, are we now 20% of the way along, and progressing 5% per decade, suggesting we need 160 more years of linear progress?
  2. Linden talks about “mysteries of development, consciousness, perception, decision, and action.” But all we need for brain “uploads” (= emulations) are good enough models of individual cell input/output/state relations (i.e., how a brain cell’s output signals and internal states change as a function of its input signals). We don’t need to understand how that huge mess of connected cells actually produces high level brain functions. If we just focus on this more limited goal, then how far along are we, and how fast are we moving?

Yes Kurzweil seems too optimistic, but rather than criticizing Kurzweil it seem far more useful for Linden to offer his own best expert estimates. Brain emulations would have such enormous social implications that even if they will take a century or so to arrive, it is still very important to let people know, so we can start to prepare. I fear my economist colleagues will continue to ignore this possibility until top brain scientists like Linden tell them it really is coming.

GD Star Rating
loading...
Tagged as: ,

Stross on Singularity

By Robin Hanson · June 22, 2011 9:50 pm · Comments (9)

I’ve long enjoyed the science fiction novels of Charlie Stross, so I’m honored that he linked to my Betterness Explosion from his Three arguments against the singularity:

I periodically get email from folks who, having read “Accelerando”, assume I am some kind of fire-breathing extropian zealot who believes in the imminence of the singularity, the uploading of the libertarians, and the rapture of the nerds. … It’s time to set the record straight. … Santa Claus doesn’t exist. …

(Economic libertarianism is based on … reductionist … 19th century classical economics — a drastic over-simplification of human behaviour. … If acted upon, would result in either failure or a hellishly unpleasant state of post-industrial feudalism.) …

I can’t prove that there isn’t going to be a hard take-off singularity in which a human-equivalent AI rapidly bootstraps itself to de-facto god-hood. Nor can I prove that mind uploading won’t work, or that we are or aren’t living in a simulation. … However, … the prospects aren’t good.

First: super-intelligent AI is unlikely because … human-equivalent AI is unlikely. … We’re likely to leave out … needing to sleep for roughly 30% of the time, being lazy or emotionally unstable, and having motivations of its own. … We clearly want machines that perform human-like tasks. … But whether we want them to be conscious and volitional is another question entirely.

Uploading … is not obviously impossible. … Imagine most of the inhabited universe has been converted to a computer network, … programs live side by side with downloaded human minds and accompanying simulated human bodies. … A human mind would lumber about in a massively inappropriate body simulation. … I strongly suspect that the hardest part of mind uploading … [is] the body and its interactions with its surroundings. …

Moving on to the Simulation Argument: … anyone capable of creating an ancestor simulation wouldn’t be focussing their attention on any ancestors as primitive as us. … This is my take on the singularity: we’re not going to see a hard take-off, or a slow take-off, or any kind of AI-mediated exponential outburst. What we’re going to see is increasingly solicitous machines defining our environment … We may eventually see mind uploading, but … our hard-wired biophilia will keep dragging us back to the real world, or to simulations indistinguishable from it. …

The simulation hypothesis … we can’t actually prove anything about it. …. Any way you cut these three ideas, they don’t provide much in the way of referent points for building a good life. … It’s unwise to live on the assumption that they’re coming down the pipeline within my lifetime.

Alas Stross’s post is a bit of a rant – strong on emotion, but weak on argument. Maybe Stross did or will explain more elsewhere, but while he makes clear that he doesn’t want to associate with singularity fans, Stross doesn’t make clear that he actually disagrees much. Most thoughtful singularity fans probably agree that where possible hand-coded AI would be designed to be solicitous and avoid human failings, that simple unmodified upload minds are probably not competitive creatures in the long run, and that only a tiny fraction of our distant descendants would be interested in simulating us. (We libertarian-leaning economists even agree that classical econ greatly simplifies.)

But the fact that hand-coded AIs would differ in many ways from humans says little on the key issues of when AI will appear, how fast they’d improve, how local would be that growth, and how fast the world economy would grow as a result. The fact that eventually unmodified human uploads would not be competitive says little on the key issues of whether uploads come before powerful hand-coded AI, how long nearly unmodified uploads would dominate, or just how far from humans would be the most competitive creatures. And the fact that few descendants would simulate ancestor humans says little on the key question of how that small fraction multiplied by the vast number of descendants compares to the actual number of ancestor humans. (And the fact that classical econ greatly simplifies says little on the pleasantness of libertarian policies.)

Stross seems smart and well-read enough to have interesting things to say on these key questions, if only he can overcome his personal revulsion against affiliating with singularity fans, to directly engage these questions.

GD Star Rating
loading...
Tagged as: , , ,