Superhumans Live Among Us

Computers are impressive machines, and they get more impressive every year, as hardware gets cheaper and software gets better. But while they are substantially better than humans on many important tasks, still overall humans earn far more income from using their smarts than do computers. And at past rates of progress it looks like it will take centuries before computers earn more income overall.

The usual explanation for why humans are so much more capable is their flexibility, which probably results mainly from their breadth. A computer doing a task usually has available to it a far smaller range of methods, knowledge, and data. When what it has are good enough, a computer can be far more accurate and cheaper than a human. But when when a computer lacks important relevant method, knowledge, and data, then you just can’t do without that human flexibility and breadth. You might hire a human to work with a computer, but still you need that human on the team.

In our world today, most people are specialists; they spend years learning the methods, knowledge, and data relevant to an existing recognized specialty area. And when your problem falls well within such an existing area, that is exactly the sort of person you want to work on it.

But often we face problems that don’t fall well within existing specialty areas. If we can give a short list of specialty areas that cover our problem, then we can collect a team with members in all those areas. Because talking between people is much less efficient that communication within one person, this team will take a lot longer to solve our problem. But still, eventually such teams are usually up to the task.

However, sometimes we face problems where we don’t know which kinds of expertise are relevant. In such cases what we really need is a person who is expert in far more areas than are most people. Let me call such people “polymaths”, though that word is often used for people who have wide interests but not wide expertise. A polymath with expertise in enough areas has a far better chance of solving broad hard-to-classify problems. A polymath is to an ordinary human as that human is to a computer. At least in terms of relative flexibility and breadth, and thus generality.

Quite often a specialist will see that some of their tools apply to a problem, and not realize that there are tools from other areas that also apply. And if specialists from other areas tell them that other tools do apply, they will usually not have sufficient expertise to directly evaluate that claim. And so the usual human arrogance will often lead them to disagree. Specialists from each area will say that they can help, and discount the possibility of help from other kinds of specialists.

Now a clear long track record showing that teams that include several kinds of specialists tend to solve a certain kind of problem better may convince many specialists that other specialists are relevant. But we often lack such clear long track records. In such cases, we often get stuck in a pattern of having a particular kind of expert deal with a particular kind of problem, even when other kinds of experts could help.

The same thing applies when humans know more than computers. Usually there’s nothing the human could say to prove to the computer that it is missing important relevant tools and knowledge. The computer just doesn’t understand these other tools well enough. So the computer has to just be told to defer to the human when the human thinks it knows better.

Bottom line: superhuman really live among us, whose better abilities compared to us really are analogous to the way we are so much better than computers: they have more flexibility, due to more breadth of expertise. But without clear track records, they usually don’t have ways to convince us to listen to them. Once we’ve found one kind of expert relevant to a problem, those experts tend to tell us that other kinds aren’t needed, and we tend to believe them.

Superhumans walk among us, but don’t get the respect they deserve. We reserve our highest honors for those who are best at specific recognized specialty areas, and mainly only recognize polymaths when they are good enough at one such area.

Added 22Apr: Actually, someone with multiple expertise areas isn’t what I meant if they haven’t worked to integrate them. Compared to computers, the human mind can not only do many things, it has integrated those tools together well. When areas overall, one needs a common representation to accommodate them both. Is one a special case of the other? Do they focus on different parameters in a common parameter space? I mean to refer to a polymath who has successfully integrated their many areas of expertise.

GD Star Rating
Tagged as:
Trackback URL:
  • Steve Burrows

    As a polymath that fits your description, I’ve found the only way to be gainfully employed is as a small business owner. Decent at most functions, master (not grand) of one or two.

    • Radford Neal

      Yes. And more specifically, farmers typically are capable in many areas – growing crops, caring for animals (including basic veterinary skills), maintaining machinery, building fences and other structures, business management, etc. This is an economic necessity, since all the profit would go away if you hired other people to do this.

      (At least, this was how it was a few decades ago…)

      • Dave Lindbergh

        There’s a difference between merely being a generalist (like your farmers) and a polymath.

        A polymath has enough knowledge in many narrow domains to see things from from the insider’s perspective. That is, from the viewpoint of an expert in those fields, even if they’re not experts themselves, they’re close enough to think like one.

        A successful farmer is probably an expert at farming, but doesn’t have a physician’s view of medicine, or a mechanical engineer’s view of machine design, or an architect’s view of buildings.

        And each of those are of course broad domains, not narrow specialties. A farmer doesn’t know *anything* about, say, antibody production by immunization with keyhole limpet hemocyanin, or something equivalently narrow in mechanics or construction.

        A farmer – a generalist – doesn’t even know those sub-fields exist. A polymath does.

      • John Shakespeare’s Fool

        Might there be an immunologist or two who are also farmers?

    • Sharper

      Yeah, my question is, what’s the best way to monetize that status? I’ve found pretty much the same thing you have, which is most “jobs” are designed for specialists. Even if you’re a specialist in multiple fields, which makes you better than usual in a single field, it’s tough to find an employer who is going to make use of more than a single field. In some ways, it may be that scarcity keeps employers boxing up jobs in ways that they can always find someone to fill them, rather than optimizing any particular person’s job.

      So single self-employed business is one option. Another with a bit more security is where you essentially work in multiple fields, but do it across more than one job. Another might be CEO, although as a “lottery” style job, that’s a tough one to work towards. Related is the “talent stack” idea, where you can create an uncommon work function by having complementary expertise.

      Curious if anyone else here can think of highly valuable jobs best filled by someone who is an expert in a few areas and semi-expert in many others.

  • SK

    Robin, you are yourself a superhuman in the above sense. Can you give some other people you know of as examples?

  • Cesar Uliana

    While I get the gist of the argument, I think that the existence of said polymaths is up to discussion. Historical examples are very controversial, which is easy to see in the talk page of the wikipedia article for polymath. The prototypical polymath is usually given to be Da Vinci, who was a great artist, yes, but had basal knowledge of any branch of science for his time. Rene Descartes might have been good at math and philosophy, but so were Pascal and Leibniz, which goes to show that it was not unusual for the time to dabble in both. Marie Curie is the only person to win Nobel Prizes in two different sciences, yet it involved basically the same set of skills. We might say today that maths, physics and chemistry are different things, but in the early 1800s mechanics was just a branch of maths, and all the experimental sciences were just a big lump of things where people like Faraday and Thomas Young could shine. Von Neumann kicked ass in all areas of math, was he a polymath? What I’m saying is that the existence of polymaths requires a epistemological theory to justify that indeed one is an expert in different areas, and not just someone with breadth of knowledge.

    • SK

      I think Robin’s notion of polymaths will probably not be found among people at the very top echelon. E.g., my dad has worked variously as a soldier, an electrical engineer, a cook, a teacher, a mechanical engineer, a manager in various companies, an event organizer, an environmental consultant and so on. In all these jobs, he came in without much domain knowledge, and did lots and lots very hard work initially learning on the job. He doesn’t have a college degree, so he isn’t credentialed as an “expert” in any domain. In none of these domains can he reasonably compete in the long run with experts. But he’s gotten hired because he knew the people giving him the jobs and they needed someone flexible and whom they could trust.

      • Robin Hanson

        Flexibility comes in part from knowing many things and having integrated all that.

    • Dave Lindbergh

      I don’t think a polymath needs to be an expert in multiple areas (tho in my experience they’re usually expert in at least one).

      Instead, they need to be somewhat competent in lots of areas – enough to have an insider’s view of each field. That lets them know where to dive deeper to solve a given problem.

      Elon Musk may be the most obvious example today. Deep domain knowledge in many different fields – perhaps not deep enough to be expert, but deep enough to know how to exploit expert knowledge in multiple fields.

  • Contrary to your claim, the gap between computers and normal humans seems much larger to me than between normal humans and polymaths. (As someone who’s pretty specialized and not a polymath, I guess this is what I’d say.) Do you have any way to quantify gap size, or any data?

    • Robin Hanson

      I didn’t claim the gap was the same size, but instead say it is a gap of the same kind.

  • Andrew Flicker

    I think limited “polymaths” can be created within an organization or team if members are motivated and flexible- my own team at work has had a lot of success by occasionally rotating major job roles from year-to-year, so many team members are experts in 2-4 roles, even though their “job” might only be 1. Really improves your workflow when your lead dev was a marketer for a year, and your email guy used to run ebay, and your ebay guy used to run the blog, etc., etc.

  • Robin makes the facially implausible claim that superhumans live among us, but offers nothing in support. The key sentence is “A polymath with expertise in enough areas has a far better chance of solving broad hard-to-classify problems.” Indeed, but what remains to be shown (or even argued) is that such polymaths exist; or what is the same thing, how much is “enough.”

    How many fields does an ordinary human know that a computer doesn’t. Uncountably many. How many areas is it possible for a human to gain high expertise? An easily countable number, to be sure. A patchwork knowledge of several fields (say physics, economics, and computer science) doesn’t give you anything like the advantage a typical human has over a computer. The chance that one of a still small number of fields will happen to fit a random problem would seem small – at least without further argument.

  • Dave Lindbergh

    Lots of negative comments here from specialized experts.

    But I think you’re onto something real, Robin.

    My wife was a postdoc under a Nobel laureate (Medicine); she observed that the way to win a Nobel prize is to obsessively specialize and then obsessively work. The peak of the curve is what’s rewarded. That may be the only path to a Noble prize, but broadness – the area under the curve – has other rewards.

    Great entrepreneurs, for example, are almost always polymaths. Look at Elon Musk, Bill Gates, Steve Jobs, Jeff Bezos for example – all polymaths, I’ll argue.

    • These various college dropouts are polymaths? I don’t think they are anything of the sort. An entrepreneur must learn the parts of disciplines relevant to his enterprise. They don’t, typically, first amass knowledge relevant to a broad array of problems, as Robin posits. Nor do they have the broad thirst for knowledge associated with the polymath. To the contrary: they have a laser-like focus on financial success.

  • Romeo Stevens

    Let’s drill down. What, qualitatively, if anything, is a polymath doing that a specialist is not?

    To answer this i’ll import David Chapman’s framing of problem solving methods vs problem representation permutations. If we look at problem solving methods as a class in any particular domain, they are often fairly limited, and the work of experts in that domain is to recognize clever ways of making those methods fit various problem and data sets. Each field has a certain quiver of problem representations that it then tries to translate other domains into, so that it can apply its methods. This is useful, but sometimes a bit of a crap shoot. So what does a polymath do? I’d argue that it actually isn’t access to a broader variety of tools, though basic knowledge of them is often helpful, but access to a broader variety of problem representations, and more experience with translating between them.

    This could be explicitly trained but is not, because it is not recognized as a distinct thing separate from domain expertise in the same way that it took a long time for people to notice that deliberate practice was a set of abstractions present across high level practice in many domains.

    I think there is something useful around practicing with schemas that tools slot into rather than just practicing with tools, but I haven’t figured out what yet. The analogy would be a woodshop class where you train a day on each of the tools vs one where you are tasked with building a thing and learn the tools as needed when you recognize what sort of affordances will be required for your vision.

    • Robin Hanson

      Representations ARE tools, and some of our most useful ones.

      • Romeo Stevens

        Tool taxonomies are useful

  • Robin Hanson

    I added to the post.

  • Olio Pantalonia

    Philosophers have always tried to raise their status by writing essays arguing that philosophy is actually a generalist discipline, but this has always been interpreted as mere self-interested signaling.

    • Robin Hanson

      Useful generality mostly comes from generalizing from particulars. Trying to be general by staying with generalities mostly fails.

      • Mathematics would seem to be the decisive counter-example to this claim.

      • Robin Hanson

        Math, and some areas of philosophy, are fields almost defined as about how far one can go staying with generalities. That approach can go far WITHIN those fields, but it isn’t a substitute for learning OTHER fields outside those.

  • Robert Koslover

    I’ve had the privilege of meeting/knowing some impressive polymaths. I would assert that it is is not uncommon that a person with very great accomplishments in a specific field (e.g., physics) may also be highly-skilled, though less accomplished, in one or more other fields (e.g., music). Such a person could have easily been employed in that other field instead, if not for their primary area of expertise overshadowing their other talents and diverting their attentions from developing them. For those who are somewhat less gifted than true polymaths, specialization and hard work offer an opportunity to achieve real greatness, albeit in a narrow subject field. For many people, this is a very effective path to success. The key is to understand your talents and pursue the right specialty!

  • I mean to refer to a polymath who has successfully integrated their many areas of expertise.

    What’s the advantage of using this term differently from ordinary usage? The paradigm example of polymaths have unrelated accomplishments in disparate fields. (For example, ) Moreover, this isn’t just a different understanding; it’s a major change in your argument.

    • Robin Hanson

      As usual, one looks for the closest term one can find, and then if needed declares that one is locally defining a different sense of that term.

      • To each his own default approach, I suppose. My preference would be to invent a new term. But we’re limited by the restriction of this discussion to generalities ;), inasmuch as you haven’t provided any examples. Seems to me that individuals who are proficient in even two advanced disciplines and also proficient at integrating them are pretty rare, and what most distinguishes them is their integrative ability and interests within restricted domains rather than their breadth of expertise. (“Cross-disciplinary integrative thinkers”? I suppose I can see why you would prefer “polymaths.”)

  • You certainly deserve more recognition, Robin! 🙂

  • Good and important post.

    “We reserve our highest honors for those who are best at specific recognized specialty areas, and mainly only recognize polymaths when they are good enough at one such area.”

    One can speculate about why that is. One possible reason may be that it is difficult for all but the most competent people to reliably assess polymath/generalist competence. It is easier to assess specialist competencies (especially in empirical and technical fields).

    I have written a bit on these issues; e.g., here:

  • Stephen S

    It would be interesting to try to estimate the maximum intellectual output at a defined task of a team of two or more humans, each with realistic constraints re: learning and communication.

  • Pingback: Reading List for the Week ending May 12, 2017 – Reading Diet()

  • For those who like the integration between chess and AI, and who are concerned about potential for AI control over future cultures via
    coercive though reform processes.

    The concept made concrete in a manner usable for “repeated games”:

    If you worry that AI could brainwash everyone, this is a concise theoretical illustration of the HOW.

  • Richard Neumann

    If you have very high iq you are basically better at everything (like a faster CPU). This is not allowed. Therefore you are only allowed to be good at one thing (like the rest of us). If you claim otherwise you will be shunned.

  • brendan_r

    Agreed! Who are your favorite superhumans?