Caplan Audits Age of Em

When I showed Bryan Caplan an early draft of my book, his main concern was that I didn’t focus enough on humans, as he doesn’t think robots can be conscious. In his first critical post, he focused mainly on language and emphasis issues. But he summarized “the reasoning simply isn’t very rigorous”, and he gave 3 substantive objections:

The idea that the global economy will start doubling on a monthly basis is .. a claim with a near-zero prior probability. ..

Why wouldn’t ems’ creators use the threat of `physical hunger, exhaustion, pain, sickness, grime, hard labor, or sudden unexpected death’ to motivate the ems? .. `torturing’ ems, .. why not?” ..

Why wouldn’t ems largely be copies of the most “robot-like” humans – humble workaholics with minimal personal life, content to selflessly and uncomplainingly serve their employers?

He asked me direct questions on my moral evaluation of ems, so I asked him to estimate my overall book accuracy relative to the standard of academic consensus theories, given my assumptions. Caplan said:

The entire analysis hinges on which people get emulated, and there is absolutely no simple standard academic theory of that. If, as I’ve argued, we would copy the most robot-like people and treat them as slaves, at least 90% of Robin’s details are wrong.

Since I didn’t think how docile are ems matters that much for most of my book, I challenged him to check five random pages. Today, he reports back:

Limiting myself to his chapters on Economics, Organization, and Sociology, [half of the book’s six sections] .. After performing this exercise, I’m more inclined to say Robin’s only 80% wrong. .. My main complaint is that his premises about em motivation are implausible and crucial.

Caplan picked 23 quotes from those pages. (I don’t know how picked; I count ~35 claims.) In one of these (#22) he disputes the proper use of the word “participate”, and in one (#12) he says he can’t judge.

In two more, he seems to just misread the quotes. In #21, I say taxes can’t discourage work by retired humans, and he says but ems work. In #8 I say if most ems are in the few biggest cities, they must also be in the few biggest nations (by population). He says there isn’t time for nations to merge.

If I set aside all these, that leaves 19 evaluations, out of which I count 7 (#1,4,9,13,17,19,20) where he says agree or okay, making me only 63% wrong in his eyes. Now lets go through the 12 disagreements, which fall into five clumps.

In #6, Caplan disagrees with my claim that “well-designed computers can be secure from theft, assault, and disease.” On page 62, I had explained:

Ems may use technologies such as provably secure operating system kernels (Klein et al. 2014), and capability-based secure computing systems, which limit the powers of subsystems (Miller et al. 2003).

In #5, I had cited sources showing that in the past most innovation has come from many small innovations, instead of a few big ones. So I said we should expect that for ems too. Caplan says that should reverse because ems are more homogenous than humans. I have no idea what he is thinking here.

In #3,7, he disagrees with my applying very standard urban econ to ems:

It’s not clear what even counts as urban concentration in the relevant sense. .. Telecommuting hasn’t done much .. why think ems will lead to “much larger” em cities? .. Doesn’t being a virtual being vitiate most of the social reasons to live near others? ..

But em virtual reality makes “telecommuting” a nearly perfect substitute for in-person meetings, at least at close distances. And one page before, I had explained that “fast ems .. can suffer noticeable communication delays with city scale separations.” In addition, many ems (perhaps 20%) do physical tasks, and all are housed in hardware needing physical support.

In #2,23, Caplan disagrees with my estimating that the human fraction of income controlled slowly falls, because he says all ems must always remain absolute slaves; “humans hold 100% of wealth regardless .. ems own nothing.”

Finally, half of his disagreements (#10,11,14,15,16,18) stem from his seeing ems them as quite literally “robot-like”. If not for this, he’d score me as only 31% wrong. According to Caplan, ems are not disturbed by “life events”, only by disappointing their masters. They only group, identify, and organize as commanded, not as they prefer or choose. They have no personality “in a human sense.” They never disagree with each other, and never need to make excuses for anything.

Remember, Caplan and I agree that the key driving factor here is that a competitive em world seeks the most productive (per subjective minute) combinations of humans to scan, mental tweaks and training methods to apply, and work habits and organization to use. So our best data should be the most productive people in the world today, or that we’ve seen in history. Yet the most productive people I know are not remotely “robot-like”, at least in the sense he describes above. Can Caplan name any specific workers, or groups, he knows that fit the bill?

In writing the book I searched for literatures on work productivity, and used many dozens of articles on specific productivity correlates. But I never came across anything remotely claiming “robot-like” workers (or tortured slaves) to be the most productive in modern jobs. Remember that the scoring standard I set was not personal intuition but the consensus of the academic literature. I’ve cited many sources, but Caplan has yet to cite any.

From Caplan, I humbly request some supporting citations. But I think he and I will make only limited progress in this discussion until some other professional economists weigh in. What incantations will summon the better spirits of the Econ blogosphere?

GD Star Rating
Tagged as: ,
Trackback URL:
  • arch1

    Robin re: “half of his disagreements”-
    I guess you think that an em scanned from one of the most productive humans wouldn’t be made *more* productive by e.g. tweaking her self-focus close to zero and her satisfy-the-boss focus close to one. I guess Bryan might think the opposite. But there isn’t any literature on such tweaking of such ems, is there?

    • there isn’t any literature on such tweaking of such ems, is there

      Data, if it exists, on whether the most productive humans are least self-focused would seem relevant. Seems to me it depends on the occupation.

      [I’d think there are order-giver ems and order-taker ems. Their relations seem more interesting to me than those between ems and humans.]

      • arch1

        My guess is that ems would be different in this respect.

    • The whole point of “ems” — as opposed to constructed AI — is that their internals are not understood to human civilization, and so can’t be “tweaked” at all. You only get copies of existing humans, you don’t get to modify them.

      Once you can design human-level computer intelligences, you move into the realm of an AI world, which is totally different from Robin’s em world.

      • arch1

        Thanks Don. I made an apparently rash assumption about what Robin meant by “mental tweaks and training methods to apply”, in the process forgetting that a merely scanned brain is not understood, much less configurable.

        I *am* left scratching my head about the “mental tweaks” mention (which is probably a good thing – it will encourage me to read the book before spouting off again:-)

  • Lord

    Monthly doubling would be astounding. More likely, what is in abundance falls steeply in price while what is not will get extremely expensive and always remain the bottleneck though just which one could continually shift. While physical resources may fall in price, they will likely not fall nearly as fast as computing which will limit its contribution. It might cure unemployment though, always having the next physical tasks ready.

    If really based on scanning, we might see mental aging and faster the faster they operate. Attempts to prevent this may destroy its operation. While they could be reset, this could lose everything not turned into an artifact.

    They won’t act like robots because robots need to be told what to do and who would be able to keep up with them.

    • Jake

      The volume of space within our lightcone grows cubically in time, hence the amount of matter-energy accessible by time t should vary at most cubically with t, assuming bounded fluctuations of density of the universe.

      Therefore, even under WILDLY idealized optimistic assumption, exponential gdp growth can only last for a short time.

      E.g. the Milky way has ~10^68 atoms and a radius of 50,000 light-years. If GDP were just getting new atoms, and it doubled every month, we’d have the whole milky way in just 225.9 months < 19 years. If GDP is also better organizing the atoms, let's say we manage an improvement similar to that of enriched uranium over meat in terms of energy density; 1 kg of U-238 contains 11 million times more energy than 1 kg of meat. That buys us another < 2 years of month doubling GDP, for a total of < 21y.

      If instead you go off of the improvement in FLOPS from early human brains to current computers, or even all computers in the world combined (~2B computers, < 1 TFLOPs each, assume 1 human can do one FLOP a minute), you get maybe an extra 4y of such doubling, for a total of 25y.

      If we very generously double that figure (which in exponential growth is like SQUARING gdp), we come to a 50y time frame for monthly-doubling gdp to eat the entire galaxy AND optimize the atoms to a degree that is several orders of magnitude larger than anything that has ever been seen in 100k years of human history.

      The problem is, it necessary takes 50 THOUSAND, not 50 years to even have a chance of reaching half the atoms in the galaxy (remember, the radius is 50k LY and we are at the edge).

      • The book only focuses on a roughly 2-year period anyway (thousands of subjective years from the ems perspective). Growth could immediately level off at the end for all we know.

        I’m skeptical that ems are necessary for this kind of extreme growth. Simple von Neumann probe with minimal supervision could convert the entire solar system to useful machinery in a few dozen doublings. Since that’s just a matter of automating existing tech, it could happen much sooner than brain emulation.

  • marshall bolton

    I think this topic is much too important to be confined to economists! Brian and Robin are just wrangling about freedom v’s efficiency (again). But there are broader perspectives that each belong to other fields. For example I would wonder whether bodyless brains would not go mad (and bad)….entering into loops of craziness (and software rot). Selecting only from the most productive people would also seem to predispose to – well let’s call it “nerdiness” – a property which is only sometimes useful and sometimes crazy but always unbalanced. So anthropologists, opticians and the man-in-the-street, unite.

    • I’m happy to hear reviews from experts in most any field.

    • Selecting only from the most productive people would also seem to predispose to – well let’s call it “nerdiness” – a property which is only sometimes useful and sometimes crazy but always unbalanced.

      In any event, there would seem a drastic restriction in the range of personality types.

  • Dan R

    Bryan’s concern about who gets uploaded and how they’re treated seems like a red herring. In the end, in the absence of strong centralized control, efficiency wins. Sometimes we apply external morals and call the result evil, and other times we call it progress. One of the lessons of economics is that when regulations, perceptions, and values prefer (significantly) less efficient results, it takes massive coordination to prevent competition from causing change. This is particularly true when the growth rate potential is as fast as your model indicates.

  • Pingback: Overcoming Bias : The Future of Slavery()

  • There will be android AI slaves, gen 3 or 5 or more before a useful Em. I don’t think Ems will be the slaves, except maybe to gov’t that allows them to become “immortal” this way…

  • Pingback: Overcoming Bias : Oarsman Pay Parable()