The Myth of the Ivory Tower

I know I promised no more posts about D-Wave and its “commercial” “quantum” computer for a while. But will you look at the bait that D-Wave founder Geordie Rose has been dangling in front of me on his blog?

People tend to approach problems and form opinions through the lens of their expertise. This happens all the time when disciplines are close … but it also happens in wierder [sic] situations, where the area of expertise is entirely disjoint from the situation being analyzed — like when theoretical computer scientists have opinions about real computers for example.

In Geordie’s comments section, the message is clearer still. One commenter writes that “the Professors didn’t get there first and they are angry; all truth must first come from them.” Another imagines “the Aaronsons of the world” fervently hoping that “their fragile self-created self-contained ecosystem can be re-built just the way they like it.”

For commenters like these, it would seem that the issue has nothing to do with decoherence rates or scalability, or with what the evidence is that D-Wave is actually harnessing quantum effects to obtain a computational speedup. So in this post, I want to step back and try to understand what the real issue is.

I propose that more than a few technology enthusiasts — not just the D-Wave supporters quoted above — are in the thrall of The Myth of the Ivory Tower. According to this Myth, the basic function of academic scientists is to sit around in their armchairs, pompously declaring to be impossible what plucky inventors like Thomas Edison or the Wright Brothers then roll up their sleeves and do. Now, I might be an academic myself, but I’m also a proud American (currently residing in the 51st state), and I won’t deny that this most American of myths has a certain resonance even for me. In the end, though, I believe that the Myth tells us more about our Zeitgeist, or our collective psyche, or something like that, than it does about the actual history of technology.

The “evidence” for the Myth (when such is offered) usually consists of famous last words from distinguished scientific authorities. You know the sort of thing I’m talking about:

Heavier-than-air flying machines are impossible.
Radio has no future.
X-rays will prove to be a hoax.
-William Thomson (Lord Kelvin)

I think there is a world market for maybe five computers.
-Thomas Watson

There is no reason anyone would want a computer in their home.
-Ken Olsen

(Watson and Olsen were of course CEO’s, but for the purposes of the Myth they stand in here as “academics.”)

However, as soon as we think about these predictions and what they’re supposed to demonstrate, we notice some glaring problems. The first one is confirmation bias. No one compiles lists of pessimistic technological forecasts made by experts that turned out to be right — where would you even start?

The second problem is that many of the juiciest predictions come from a single individual: Lord Kelvin. Furthermore, they come from the twilight of his career, when he was considered to have lost his vortices even by most of his colleagues. Seeking to better understand this great physicist of the 19th century who was so wrong about the technologies of the 20th, I just read an excellent biography called Degrees Kelvin. One thing I learned is that, if the selective historians chose to focus on the first half of Kelvin’s career rather than the second, they could find equally exquisite anecdotes illustrating the reliability of academic opinions.

In the laying of the first transatlantic telegraph cable in the 1850’s, there were two colorful personalities: Kelvin and Wildman Whitehouse. Whitehouse, the “practical” man, detested any math or physics he couldn’t understand, and insisted that a transatlantic cable would just be a longer version of existing cables. Kelvin, the “theorist,” said that while a transatlantic cable was certainly possible, it would need thicker insulation, a different kind of receiver, etc. than previous cables to work reliably, and that more testing and research was needed. As it happened, after laying a cable that was every bit as unreliable as Kelvin said it would be, Whitehouse (1) had to use Kelvin’s receiver to get any signal through at all, (2) faked the transcripts to make it look like he used his own receiver, (3) fatally damaged the cable by sending 2,000 volts through it in a desperate attempt to get it to work properly, and then (4) insisted the cable was still fine after it had permanently gone silent. Eventually the cable companies learned their lesson.

Despite this and other successes (e.g., the Second Law of Thermodynamics), Kelvin’s doofus predictions in later life do illustrate two important points. The first is that, if you’re going to make skeptical pronouncements, you’d better distinguish clearly between the provably impossible, the presumably impossible, and the merely difficult and not yet achieved. The second is that, if you’re going to claim something’s impossible, you’d better have an argument, and you’d better understand what assumptions it rests on.

Alright, so let’s move on to Watson and Olsen’s predictions about the computer industry. The funny thing is, these predictions weren’t nearly as stupid as they sound! Why? Because there’s nothing inevitable about the concept of a personal computer. Instead of billions of home PC’s, we could just as easily imagine most of the world’s computing power concentrated in a few servers, accessible remotely to anyone who wanted it. In this alternate universe, your desktop PC would be little more than a glorified information portal — a “browser,” if you will — while most of the actual application software (email, calendars, maps, etc.) ran elsewhere. I admit that this is just a fanciful, hypothetical scenario, but what does that matter to a theorist like me?

Speaking of which, the Internet was of course the child of DARPA and NSF, raised to adolescence in university CS departments. (DARPA has since reoriented itself toward projects with shorter-term payoff, its previous funding model having failed so disastrously.) The Web was created by Tim Berners-Lee at CERN, and the first popular web browser by Marc Andreessen at the University of Illinois. (And yes, Al Gore had a nontrivial role in funding this work.) R, S, and A were all at MIT. If you’re going to argue for the irrelevance of academic research, the Internet is not the place to start.

But what about some of the other spectacular inventions of the last fifty years: the laser, the transistor, the fiber-optic cable, the communications satellite? Didn’t those come from the private sector? As it happens, they came from Bell Labs, which is interesting as the sort of mammoth exception that proves the rule. Because of AT&T’s government-sanctioned monopoly, for much of the 20th century Bell Labs was able to function like the world’s largest university, devoting billions of dollars to “irrelevant” research. So in the 1980’s, when Congress decided to deregulate the phone system, many people predicted that Bell Labs would die a slow, agonizing death — a prediction that’s been borne out over the last 25 years.

But surely other companies must have picked up the slack? No, not really. While Microsoft, IBM, NEC, Xerox, and a few others all provide welcome support for basic research, none of them do so on the old Ma Bell’s scale. From a CEO’s perspective, the problem with basic research is obvious: a rising tide lifts all boats, your competitors’ as well as yours. (The famous cautionary example here is Xerox PARC, which made the “mistake” of giving the world the windowing system, the mouse, and the laser printer.)

For those who adhere to the religion of capitalism, have the Arrow-Debreu Theorem tattoed across their chests, etc., it might be difficult to understand how a system based on peer review rather than the free market could lead so consistently to technological breakthroughs. I mean, all those ivory-tower academics growing fat off government grants: what incentive could they possibly have to get the right answers? Without picky customers or venture capitalists breathing down their necks, what’s the penalty for being wrong?

I’m lucky enough to be friends with Robin Hanson, a brilliant economist and futurist who starts where Ayn Rand would’ve suffered a loss of nerve and keeps going from there. Robin has long argued that the scientific peer review process is broken, and ought to be supplanted by a futures market that would reward scientists for making correct predictions. As he writes:

The pace of scientific progress may be hindered by the tendency of our academic institutions to reward being popular, rather than being right … Academia is still largely a medieval guild, with a few powerful elites, many slave-like apprentices, and members who hold a monopoly on the research patronage of princes and the teaching of their sons …

Imagine that academics are expected to “put up or shut up” and accompany claims with at least token bets, and that statistics are collected on how well people do. Imagine that funding agencies subsidize pools on questions of interest to them, and that research labs pay for much of their research with winnings from previous pools. And imagine that anyone could play, either to take a stand on an important issue, or to insure against technological risk.

Personally, I hope that Robin’s science futures market gets tried on a significant scale, and I can’t wait to see the results. (Naturally, even the marketplace of ideas has to compete in the marketplace of ideas!) I agree with Robin that academic science is often tradition-bound to the point of absurdity, and that its institutions ought to be as open to scrutiny and replacement as its theories. But I don’t go as far as he apparently does in the direction of the Myth of the Ivory Tower. For me, the interesting thing about science is not that it’s broken, but rather that it’s about the least broken enterprise in the whole sorry history of our species.

68 Responses to “The Myth of the Ivory Tower”

  1. Robin Hanson Says:

    I’ll agree, academia does seem to be nearly the least broken innovation institution humans have ever seen. In the twenty years since I wrote the above quote, I’ve seen academia’s severe failings often, sometimes up close and personal. But yes, that just suggests how very sorry is the history of our species. On average, an average person is far better off taking the word of a respected academic, rather than believing the opposite. So I’m happy to take Scott’s word on D-Wave. But I do suspect we could do even better if we allowed betting markets on such things.

  2. none Says:

    wow, they even misquote you here:

    http://en.wikipedia.org/wiki/D-Wave_Systems

  3. Jonathan Vos Post Says:

    “(Naturally, even the marketplace of ideas has to compete in the marketplace of ideas!)” — a strong conjecture that the mechanism of evolution in the ideocosm is identical (modulo a constant) for ideas and meta-ideas. My years in an academic CS department (1973-1977) led me to agree with the conjecture, as I beta-tested John Holland’s breakthrough book on the genetic algorithm. Holland, of course, was an ivory tower academic — who launched a sub-industry worth over a billion dollars!

    see also: [Sir Arthur C. Clarke’s Laws]

  4. Nagesh Adluru Says:

    Scott when you prove a theorem do you check the proofs of theorems/lemmas you rely on, all the way down to assumptions (once for every unique path down)?

    I ask this because there might be some people who think this job might be prone to errors especially if they don’t understand it intuitively.

    P.S.: I firmly believe that intuition is not the best verifying technique we got.

  5. anonymous Says:

    How come they can’t use their fancy quantum computers to prevent them from getting pwned?

  6. Dave Bacon Says:

    Damnit after those quotes I was hoping that you, Scott, would write something like “A quantum computer will never be built!” (maybe even in all CAPS) so that years from now we could use you as an example of the bad predicitions being blasted out of the ivory towers.

  7. Scott Says:

    Scott when you prove a theorem do you check the proofs of theorems/lemmas you rely on, all the way down to assumptions (once for every unique path down)?

    No.

  8. Nagesh Adluru Says:

    But then I guess you expect peer-review process to have weeded out any errors. But what if the peer-review is not perfect and is not independent except for subjectivity of how deep down each path is verified. And as the time goes by the things you know things can get really buried down.

    You argue very beautifully about how scientific predictions (especially objections) better be provable not presumable. I am curious to know your view on how reliable even this proof business is.

  9. Scott Says:

    Nagesh, until the day when all proofs are done by computer in ZF set theory, mistakes will happen and which proofs one can rely on will always be a judgment call. The good news is that, if a theorem is important enough to form the basis for lots of subsequent work, then generally it’s also important enough for lots of people to have verified it. In my own work, when I’ve used a previous theorem at all, it’s usually been something like Ambainis’s adversary theorem or Markov’s theorem on polynomial degree, which has been checked by dozens of people (or in the case of Markov’s theorem, thousands of people).

  10. Nagesh Adluru Says:

    Does the strength of limitation theorems in QC come from “complexity” based theorems like adversarial arguments and degree-bound arguments or from “quantum” based theorems like those proven by mathematical physicists?

    If it is latter then how reliably one can predict the limitations since the assumptions of quantum mechanics might still be too fragile (?!)

  11. nextquant Says:

    Thus there is some hope (which I do not share) that within the next 20 years we will be able to factor the number 6, or even 15, by using Shor’s algorithm!

    Quantum computing: a view from the enemy camp by M. I. Dyakonov (October 2001)

  12. Jonathan Jones Says:

    nextquant (quoting Dyakonov in October 2001) writes

    Thus there is some hope (which I do not share) that within the next 20 years we will be able to factor the number 6, or even 15, by using Shor’s algorithm!

    Actually the factorization of 15 by Shor was published that December…

  13. Scott Says:

    Nagesh: I’m not sure I understand your question, but all the limitation theorems for QC’s that I know about are based on complexity theory. Even so, those theorems certainly assume the validity of quantum mechanics! For example, if quantum mechanics were nonlinear, then the lower bound theorems wouldn’t apply.

    nextquant: 15=3×5 is old news by now! I’m still waiting for the factors of 21…

  14. Slawek Says:

    > until the day when all proofs are done by computer in ZF set theory

    That would be a sad day… But I am looking forward to the day when it is common to use machines to verify proofs done by humans. It is technically possible today, it is just that the inhabitants of the Ivory Tower (typically) can not stand the tedium of writing real proofs.

  15. Scott Says:

    Slawek: Yeah, an implicit assumption in my comment was that, if we ever switch to machine-verifiable proofs, then machines will have to play a large role in finding the proofs — otherwise it’s just too tedious.

  16. Kaus Hackula Says:

    Excellent post. It needed to be said.

  17. Vladimir Levin Says:

    The article about creating a sort of futures market where scientific problems and their solutions are the commodities is kind of interesting. I may simply not have the needed background, but I find myself wondering if there is any difference between that idea and the idea of just privatizing all scientific endeavours.

  18. Nagesh Adluru Says:

    I guess I should have asked ‘Is it possible to asses how reliable the assumptions of quantum mechanics are?’ Because that degree of reliability indicates the actual strength of the limitation theorems.

  19. Jonathan Vos Post Says:

    Slawek Says: #14,

    > until the day when all proofs are done by computer in ZF set theory

    “That would be a sad day… But I am looking forward to the day when it is common to use machines to verify proofs done by humans.”

    See the fascinating discussion by the 1998 Fields medalist Richard Borcherds at
    http://borcherds.wordpress.com/experimental-mathematics/#comment-42

    including:

    “… Rather than find proofs by themselves, computers might be quite good at finding formal proofs with human assistance, with a human guiding the direction of the proof (which computers are bad at), and the computer filling in tiresome routine details (which humans are bad at)….”

  20. Scott Says:

    Nagesh: If quantum mechanics was wrong, that would be the biggest physics discovery since the 1920’s (and depending on how it was wrong, maybe the biggest physics discovery of all time). And that goes for humans as well as asses. :-)

  21. Coin Says:

    The thing that’s really standing out to me here as bizarre is that as I’m understanding this situation, D-Wave seems to be dragging out this “you’re just a theoretical ivory tower theoretician!” carnard in response to demands that they produce actual data or evidence that their product works.

    Maybe I’m not being entirely fair to D-wave here but it seems to me that when someone starts to view empiricism as some kind of far-out detached-from-reality academic thing, there is a problem.

  22. Vladimir Levin Says:

    Oops, sorry for my earlier post. I see now that the difference is that the proposal for a market as described really is a kind of gambling, since the players in this market would not be involved in any kind ownership stake.

  23. Nagesh Adluru Says:

    Thanks Scott. So building a QC as claimed by D-Wave should have to be the biggest discovery of all time because for them not only QM is wrong but they seem to have figured how it is wrong so that they can solve NP-Complete problems:)

    This is close to impossibility without people like you on the team:)

  24. Bram Cohen Says:

    A few things –

    Lord Kelvin may be the model of the pompous academic, but was he actually prissy?

    The field to which I contribute, high-level networking protocols, is clearly a poor fit to the academic model. Apparently this is due to it being mostly messy engineering, while the academic structures for it are designed along the lines of theoretical CS. This has reached cartoonish proportions, to the point where the most important recent work in it was done by an unemployed unheralded decidedly non-academic hacker working essentially alone.

    Something rather like your friend Robin’s idea in scientific futures exists in the form of modern venture capital, which does a fairly good job of funding certain kinds of research which weren’t previously. It has its own clear biases though, much more like traditional research labs than academia. I rather suspect that his idea won’t work though, because I can’t see how such markets could get liquidity.

  25. Scott Says:

    Nagesh: In fairness to D-Wave, they’re not claiming that quantum mechanics is wrong, or even that NP⊆BQP. As I understand them, they’re “merely” claiming that they’ve built a scalable quantum computer, and that they’ll soon be able to solve NP-hard optimization problems “well enough in practice” in a way that wouldn’t be feasible with classical computers.

  26. Nagesh Adluru Says:

    Oh I see. So it’s just that they want to commercialize their product not their scientific breakthroughs at the cost of undermining scientific rigor which is being ranted against by experts. Thanks Scott for patiently participating in such discussions.

  27. anonymous Says:

    “Imagine that funding agencies subsidize pools on questions of interest to them, and that research labs pay for much of their research with winnings from previous pools.”

    So I guess a thesis on gender roles in Chaucer will have to be pro bono…

  28. Anup Says:

    In the US there are myths of far lesser towers that gain traction with the general public. What hope do scientists have of convincing the populace that quantum computing is unlikely to help with NP (an issue where even understanding what the argument is about is very challenging) when they seem to be unable to convince people that the universe is more than a few thousand years old etc etc.

    In the end it seems like most people don’t really have much of an interest in having their beliefs line up with truth… they’d much rather believe things that give them the most satisfactory “state of the world”.

    Most people aren’t academics and consider themselves very different from academics… so they’d like to believe that someone just like them one upped academia… that gives them the most satisfactory world view.

    I think it’s the same reason people have a bias to root for underdogs in sports.

    And maybe why people voted for GW twice (he dudn speak good, just like me).

    If you’re like me, I like you.

  29. Scott Says:

    Anup, the question is this: should one throw up one’s hands in despair, or should one work to decrease the total doofosity in the world by some positive amount? No, I haven’t gotten the message that quantum computers are unlikely to help with NP out to a billion people, but I think I’ve gotten it out to a thousand people. So you see, I’m already the cube root of the way there!

  30. James Says:

    There’s an ancient and perfectly reliable way to settle disagreements like this: make a bet. I use this occasionally when colleagues (math) insist something is true that I insist is false. I find the offer of a $1000 bet clarifies the situation immediately, at least for mild disputes at seminars. Of course, this is a public dispute. Perhaps something like $50,000 would be necessary to be taken seriously.

    On the other hand, if you can’t make your disagreement precise enough to bet on it, then you’re just arguing about counting angels on the head of a pin. If you can but don’t want to make the bet, then maybe you aren’t as sure about this as you act.

  31. Scott Says:

    James: I would be willing to bet, if Geordie and I could agree on the terms — which is really the nontrivial part. Presumably Geordie wouldn’t agree to a bet about factoring a specific 1000-bit integer, since he hates the factoring problem. I, on the other hand, would not agree to a bet about “building a 1024-qubit quantum computer” or “solving our customers’ practical optimization problems,” since the relevant question for me is whether D-Wave can use quantum effects to solve large optimization problems asymptotically faster than those same problems could be solved with the best-known classical algorithms. So any bet would have to be phrased extremely carefully (suggestions?).

    Two further points that might be relevant:

    First, I don’t accept your premise that if a disagreement isn’t precise enough to bet on, then it must be about angels on the head of a pin. To give one example, the disagreement might be about what should count as a “practical quantum computer.”

    Second, I’ve never said that what D-Wave is claiming to have done is impossible — merely that we haven’t seen any evidence that they’ve actually done it. Recall Bertrand Russell’s reply when asked what he would say if he died and met his Maker: “But my dear Lord, you didn’t provide enough evidence!”

  32. Robin Hanson Says:

    Bram, anonymous answers your question – liquidity could come from subsidies offered by research patrons. And, as with any institution, if there are no patrons interested in gender roles in Chaucer, they would have to fund themselves or skip it.

  33. JK Says:

    I don’t even understand how Robin’s “futures market” would apply to most of the best research in TCS. Take all the early work on the PCP theorem, for example. What would the bets have been just a few years before that whole line of research was started (when PCP was not yet even defined)? Once the PCP research started, would people then be betting on whether NP is in PCP(log, O(1)) or not?? The idea seems kind of silly to me, though maybe it works better in other disciplines.

  34. Joe Fitzsimons Says:

    I must point out that I do see a problem with this betting scheme: it introduces an incentive (and probably quite a strong one) for researchers to be less than objective.

    I don’t think it would be a good thing for science as a whole to have researchers become even more invested some position.

    Imagine, for example, that some researcher does a lot of work on say whether or not NP=P, and makes substantial progress in showing (in a long and complicated proof) that this would require some even more unlikely collapse in the hierarchy of complexity classes. If he were to discover a flaw in his reasoning, after publication, he would have a very strong incentive to keep his mouth shut about it. Even if his work is completely correct, but his institution chooses to bet, based on this work, that P!=NP, then looking at some promising approach which may prove P=NP is something which he will be strongly discouraged from doing.

    Obviously scientists become invested in their prior work, be it their ego, their reputation or any one of a number of other human or practical concerns. To some extent this is unavoidable, but is this really something we want to encourage?

  35. Dan Stowell Says:

    If said researcher does find a mistake in his reasoning, he would have information that the rest of the market does not and could gain from aligning his bet(s) with reality-as-he-now-understands-it.

    Researchers who play the market based only on their research stand to lose if they’re wrong. Hence the appeal of prediction markets for ideas.

    Of course, if the particular market the researcher invests in has a deadline (which would be ludicrous in this case – I wouldn’t buy anything tied to P=NP being resolved by a particular date) he could pump-and-dump by publishing (possibly-flawed) arguments for a side that he has invested in.

  36. Joe Fitzsimons Says:

    So now we need the FTC involved? That actually puts me off the idea even more.

  37. Jonathan Vos Post Says:

    “… an incentive (and probably quite a strong one) for researchers to be less than objective….”

    Which, if it has ever been so, was quite rare.

    Once money is involved, one has a business, whether one calls it a university, a research group, a virtual corporation, a conspiracy, a consortium, a city-state, a guild, a union, and army, or a government.

    Disinformation has been a sophisticated technology since we became human beings; perhaps it coevolved with us.

    I have been working for some years with Professor Phillip V. Fellman, at Southern New Hamphire University, on Mathematical Disinformation Theory.

    Obviously: no prosepctus nor Annual Statement of a major coporation can be “objective” and 100% accurate. To do so would be to provide all necessary parameters of its pricing structure for a competitor to underprice it in any valuable contract. Nor can it be obviously inaccurate at many parameters; that is eventually found out by audits or the marketplace (Enron, Adelphia, etc.). There is, in a given market, an optimum level of disinformation in such public documents. One is trying to get one’s competitors to waste their computational resources in disproving correct numbers, in their search for incorrect numbers.

    Approximately, the equations governing the nonlinear (and chaotic) interactions of companies in a market are the same as those governing the interactions of species in an ecosystem. In that the companies are systems of intelligent agents, Game Theory applies. The argument about how much to bet is close to an argument about the Nash equilibrium of the Vaporware Quantum Computer Game.

    In this case, D-Wave is claiming a First Mover Advantage. This is, in general, spurious, as Fellman and I have shown in conference and journal publications. In this case, Scott Aaronson has shown the game is not even being played very intelligently by D-Wave. D-Wave needs to bribe (I mean hire with equity stake) some prominent academics as Advisors to their Board. These academics are to fight on the plane of ideas, so as to buffer the company from unanswered questions, and thus to add a more sophisticated level of disinformation, which nonacademics in the marketplace see as a reduction in the level of of risk.

  38. cody Says:

    i agree that there are flaws with academia, but i dont know how comfortable i am with the suggested solution. maybe the problem i have is that although a lot of science is done for the sake of profit, a lot isnt, and introducing money as a motivating factor tends to introduce corruption as well. i think some of the best scientists in the world are less concerned about money and more concerned with what the answers to their questions are. also, it seems that funding would go almost entirely to research that has practical applications, which would be very damaging in my opinion. would any one have bet on Einstein’s general relativity had he not been a prominent scientist already, since its only practical applications followed more than four decades later?

  39. Dan Stowell Says:

    So now we need the FTC involved? That actually puts me off the idea even more.
    Joe

    Yeah, you’re right, the necessary mechanics of such a market may prove too onerous.

    I’m not sure how much regulation we’d actually need, though. If you buy shares/futures based on invalid proofs, then tough luck. If you buy shares in an unprovable or non-falsifiable statement, it’s your own fault. You may have to regulate experimental results, though, especially if access to the experimental apparatus is limited or reproducing the experiments takes a prohibitively long time.

    Regulation == peer review?

  40. Anup Says:

    Scott, I’m not suggesting that you should stop trying to make a difference and I applaud your efforts.

    Maybe what you should do to get even more traction, is get a fake beard, move to the woods and announce to the world that after years of tinkering in your cabin, you have one upped DWave and found a problem that their computer will never solve (then announce some hard instance).

  41. Scott Says:

    Anup: Why a fake beard? Should I choose to adopt your excellent plan, I’m more than capable of growing a real one.

  42. Jud Says:

    Scott said: “Anup: Why a fake beard?…I’m more than capable of growing a real one.”

    After seeing the photo, I am of the opinion that it more than adequately answers your question.

  43. zydborg Says:

    Joe said:

    So now we need the FTC involved? That actually puts me off the idea even more.

    You got something against the Fundamental Theorem of Calculus?

  44. dave tweed Says:

    Here’s something I don’t understand about the “betting process”: most scientific questions aren’t, when considered completely, are either false or “more nuanced” rather than a straightforward true. Is special relativity a correct theory of physics?: sort-of, providing you don’t got beyond the weak gravity, etc. Is the standard model correct?: sort-of, in that it predicts virtually everything within its purview but no-one understands it. Is twistor theory correct?: well after 50 years no-one’s shown it’s false, beyond that it’s up in the air. And this is on relatively scientific questions before you hit things like my area of computer vision – which I’d argue belong in academia because the fundamental understanding we have to build won’t yield commercial products on a business timescale – where even experimental methodology is a heavy compromise (because the amount of ground truth data you’d need is mindboggling). So why wouldn’t there be quibbling about exact terms in all but the most obviously false bets?

    All the cases of bets I know of in physics are for trivial amounts made by people who are more interested in being “known as someone who makes bets” rather than in winning for financial purposes. I also note that in the general world, most bets are placed on highly things with highly formalised evaluations like football games, roulette wheels, etc.

  45. dave tweed Says:

    As is obvious I’m no physicist, so note in relation to #44 I’m aware the word “predict” in relation to the standard model should be something more like “can calculate”, or maybe even more nuanced.

  46. Douglas Knight Says:

    Is the standard model correct?

    No, because it says that neutrinos are massless.
    There are corrections, but I don’t think any of them is “standard.”

  47. James Says:

    Betting on theories!? No, you bet on experiments, and you use the theories to inform your bets. The theories themselves are really just ideologies. If there is no experiment for which theories X and Y predict different results, who cares which one you use? For example, to this day many people who think of themselves as being scientifically educated will say that the geocentricity of the solar system is false, when in fact it is merely meaningless, like heliocentricity.

    This is one thing that gets me about these evolution debates. Most of the arguing I hear is about what ideological spin to put on the same evidence. What all these scientists who are crying about the teaching of evolution should do is propose bets to creationists based on the outcomes of experiments, or since probably no creationist would accept such a bet, make it known that they would enjoy accepting bets from them. If there is such an experiment (I’m no expert) and the creationists accept the bet, then you win. If in every challenge, the opposing group agrees with the outcome expected by the challenger, then the two sides will be forced to admit that it really is a disagreement about ideology.

    Of course, science moves slowly, and sometimes you won’t be able to make things precise to this standard till you’ve put in a few more years of research. That’s fine, but if you complain about other people ignoring you in the meantime, you won’t get much sympathy from me.

  48. Scott Says:

    James, can you distill the argument in your comment into the form of a bet? Otherwise, aren’t you just blowing ideological hot air? ;-)

  49. Dani Fong Says:

    Oh dear. Unheralded, Bram? Still not getting respect from the plebes?

  50. Bram Cohen Says:

    Dani, the tense is ‘was … unheralded’. Some credit has been dispensed since then :-)

  51. James Says:

    Scott: Touche. But how about this? I bet that if that if you take a supposedly scientific theory and try to push it on any population without giving them specific ways in which it can make any conceivable difference whatsoever in their lives, then they won’t care about it and that all discussions will degenerate into politics and ideology. Same for any scientific population, same for any mathematical population. Let’s first try it with engineers and the axiom of choice.

  52. Scott Says:

    James, plenty of people buy books about black holes and Fermat’s Last Theorem, despite those topics making no conceivable difference in their lives. Admittedly, not nearly as many as buy Oprah’s book club selection, but enough that one can make a living writing about such things. On a smaller scale, among the regular readers of this blog (where the axiom of choice is not an unknown topic) there are probably at least a thousand engineers.

    If intellectual curiosity ever died out entirely — if everyone started to care exclusively about winning bets, and not about the nature of the world being bet on — then I guess I’d have to find a different line of work. I can only feel grateful that I haven’t had that problem yet.

  53. cody Says:

    what worries me about introducing monetary incentives is that it motivates people to be deceitful. it is unfortunate that an adamant, charismatic debater is more influential than a boring correct one. it is unfortunate that when telling the truth you must be honest of your uncertainties, but when lying you can say and omit whatever you please. in business it makes sense to lie, that is (the only) bright side to how little we pay our teachers. if there were grand profits to be had in research, how much of it would be legitimate and how much would be scams?

  54. Markk Says:

    I think a market for research funds is bad and would hurt science. Intellectual discovery is NOT a commodity. I cannot buy inspiration. To me, this is making the, all too common today, “intellectual property” fallacy: That there is some kind of ownership rights and tangible thing you can pay for to control ideas. A market is made to be manipulated in ways that are orders of magnitude more egregious than the manipulation of advisory councils.

    I actually think the way the western world has funded science for the last 50 years is unbelievably amazing in the history of mankind – have smart people try to figure out the best way to spend research money. Sure there are issues but I laugh at all the people who criticize the current US administration (mostly deservedly) – compared to anybody else in history they are spending unbelievable amounts of resources on science.

    Speaking as someone who doesn’t receive research money but pays a lot of taxes that goes to research, if they started using this market mechanism I would have to reluctantly start opposing government research funds and be opposed to supporting them.

  55. James Says:

    Scott: I was just using making bets as a substitute for making statements precise. Even though I’ve kind of been playing here, I do generally believe that you have a precise statement if and only if you can bet on it. The fact that Fermat’s Last Theorem is true could make a difference in someone’s life because it is actually a precise statement that someone can understand. E.g., if someone says “I bet you that I can find integers a,b,c such that…”, you take the bet.

    I guess all I’m saying is this: You think that these D-wave guys won’t be able to do something they’re claiming to be able to do? It might be a good exercise to make that statement precise. Mathematicians do this all the time. If someone has a conjecture of the form “There should exist a theory that explains X”, people roll their eyes, essentially because there’s no way of deciding the implicit bet. Not that there’s anything wrong with making conjectures like that to yourself, but when you bring things out for the general community and don’t want people to roll their eyes, you should find some precise implication of your vague ideas and then make that your conjecture. Then you have something worth arguing about.

  56. matt Says:

    Ok, here’s a precise statement (a few details to be filled in): a third party will generate many large, random 3-SAT instances which are just on the SAT side of the SAT-UNSAT transition (just how far on the SAT side and just how large are some of the details). This is the kind of problem Orion is supposed to be able to do, no?

    D-wave gets to use an off-the-shelf desktop PC (as an interface) plus Orion. Scott (or whoever takes Scott’s side) gets a Beowulf cluster of 16 of said PCs (he could even get 128 or more, and still be cheaper, but lets say he gets 16 of them). Whoever solves more of the instances in 24 hours wins, using whatever algorithm they want but only the given hardware.

  57. Douglas Knight Says:

    I actually think the way the western world has funded science for the last 50 years is unbelievably amazing in the history of mankind – have smart people try to figure out the best way to spend research money.

    Yes, it was a nice experiment, but it’s not clear that it was an improvement over older ways.

    Also, I’m not sure your description distinguishes it from older ways. When scientists were self-funded, it was still the scientists who were deciding how to spend the money, probably more so than today!

  58. anonymous Says:

    Scott, this is a bit random, so I apologize, but I was wondering what your thoughts are on the ontological status of mathematical objects. I think it would be great if you’d write a post about this (I’m hoping you’re a platonist)

  59. Scott Says:

    anonymous, the short answer is that I don’t lose much sleep over the ontological status of mathematical objects. I think that if we ever made contact with aliens, or even beings in an alternate universe subject to completely different laws of physics, we’d find that to whatever extent we were interested in the same mathematical questions, we agreed completely about the answers. Does that make me a platonist? If so, I’m happy to be one!

  60. dave tweed Says:

    Re James, the point I was making is that bets are primarily binary (I know spread betting exists, although not much more about it and don’t think it changes the argument). To my understanding, for the betting notion to be meaningful you have to decide in advance the completely exact conditions on which it’ll be decided. But that seems to rule out funding lots of scientific endeavour just because the evaluation methodology is too crude. I know grant committee’s at the moment weigh up which research proposals are more likely to yield either/both interesting or/and useful results and rank competing proposals. There are lots of flaws in the system, but at least in principle you’re not forcing something into the procrustean bed of an evaluation process where technicalities can be used to overturn a “fair result”.

    I’m sure the econ guy suggesting bets isn’t stupid, so I wanted to know what the flaw in my understanding was.

  61. Scott Says:

    Dave: Speaking for myself, I don’t see any flaw in your understanding. Indeed, it’s precisely for the reason you give that I see Robin’s proposal as a possible supplement to grants (for those questions that are well-defined enough to bet on) but not a replacement for them. How useful a supplement it would be is an empirical question, and one to which I’d very much like to know the answer. (Bets, anyone? :-) )

  62. Shtetl-Optimized » Blog Archive » Wanna bet? Says:

    […] Shtetl-Optimized The Blog of Scott Aaronson Quantum computers are not known to be able tosolve NP-complete problems in polynomial time. « The Myth of the Ivory Tower […]

  63. Matteo Martini Says:

    What I do not get, is why, you guys ( Scott, Dave,.. ) keep biting Geordie Rose and D-Wave; why do not you just let them try to build their quantum computer..
    You have just one year to wait.
    In mid-2008, either they will have built a 1000 qbit QC or not, and, then, we will all know if it is possible to do it or not.
    Why waste time, now, ” attacking ” them?

  64. Paul Beame Says:

    Ok, here’s a precise statement (a few details to be filled in): a third party will generate many large, random 3-SAT instances which are just on the SAT side of the SAT-UNSAT transition (just how far on the SAT side and just how large are some of the details).

    This example fails Scott’s requirements. We actually don’t know what the transtion point is and it is not clear that it is a hard problem. Current randomized classical algorithms can find satisfying assignments for million variable problems at clause-variable ratios around 4.26 and above which is higher than was previously conjectured for the threshold itself.

    Turing to Robin Hanson’s proposal, the use of market mechanisms for making technical bets has problems beyond the potential for skewing the field (which people seem most worried about).

    * The timeline is often too long for incentives involved for monetizing the payoff to be effective. What good is the payoff to me if I’ll be dead by the time I could collect?

    * Prediction markets have some value for assessing the likelihood of an outcome (e.g. the Iowa markets for elections) but they don’t give much incentive for people to settle the questions since the payoff for any individual bettor would necessarily be small for a typical question. (How much could one make if one bet that P is not equal to NP? What about P equals NP? Based on Bill Gasarch’s survey, probably not much either way.) The marginal benefit is tiny compared with the benefit that already exists in resolving the question. It won’t change how science operates.

    * If we modified these markets somehow to have real value then we have another problem. Markets with any value tend to generate secondary markets that involve risk-avoidance. Most people thought that Cold Fusion was totally bogus but those guys got funding that was based in part on avoiding the risk that they were correct. There is a good business to be done in creating uncertainty that allows one to benefit from people hedging their bets. I wonder how much this is the business model for more recent claims?

  65. Nagesh Adluru Says:

    Why waste time, now, ” attacking ” them?

    Why should D-Wave “claim” now undermining the scientific effort? I believe it’s actually them who have to wait for proper claim.

  66. Nagesh Adluru Says:

    Further, thanks to Scott (and Umesh’s comments), I understand clearly it’s not sufficient to show that they can build 1000 qubits but to show that those are really 1000 (scalable) qubits.

  67. Scott Says:

    Matteo: At this point, I’m actually not interested in “biting” D-Wave any longer; the only reasons I brought them up were (1) to respond to an attack, and (2) to use them as a journalistic “hook” for my broader point about the myth of the ivory tower.

  68. Matteo Martini Says:

    Nagesh Adluru ans Scott,
    thanks for your replies.
    Just, do not you think that, there are investors, who poured in some USD45 millions in D-Wave, and this fact, needs to be taken into consideration?