Clearly people who don't know about Occam's Razor, and people who explicitly reject it, still believe in the future. Just as clearly, we can use Occam's Razor or other principles in evaluating theories about what happened in the past. Your claim appears wholly unjustified. Was it just a vague hifalutin' metaphysical claim, or are there some underlying points that you're not bringing out?

That's a little too vaguely stated for me to interpret. Can you give an illustration? For comparison, here's one of how I assumed it would work:

A paperclip-making AI is given a piece of black-box machinery and given specifications for two possible control schemes for it. It calculates that if scheme A is true, it can make 700 paperclips per second, and if scheme B is true, only 300 per second. As a Bayesian AI using Pascal's Goldpan formalized as a utilitarian prior, it assigns a prior probability of 0.7 for A and 0.3 for B. Then it either acts based on a weighted sum of models (0.7A+0.3B) or runs some experiments until it reaches a satisfactory posterior probability.

That doesn't seem intractably circular.

These could do with forward/backward links. The Article Navigator doesn't seem to be able to get me to number 4 in this series, and the page for 'sleeping_beauty' tag appears empty.

Occam\s razor is famously difficult to justify except by circular appeal to itself. It's interesting to think of alternatives but you should be aware of what you give up when you give up Occam's razor. You can no longer make sensible inferences about the future based on your past experiences. For example, you can no longer have any confidence that the direction of gravity will still point downwards tomorrow, or that the laws of physics won't spontaneously change a minute from now. The experimental method itself no longer makes sense if you have no reason to think that the future will resemble the past.

You should read:

• http://plato.stanford.edu/entries/simplicity/

• http://lesswrong.com/lw/s0/where_recursive_justification_hits_bottom/

That is, suppose you are considering whether or not to believe that you can fly by leaping off a cliff and flapping your arms. What is the expected utility of holding this belief?

I completely grant that this scheme can have disastrous consequences for a utility function that discounts consistency with past evidence, has short time horizons, considers only direct consequences, fails to consider alternatives, or is in any other way poorly chosen. Part of the point in naming it Pascal's Goldpan was as a reminder of how naive utility functions using it will be excessively susceptible to wagers, muggings, and so on. Although I expect that highly weighting consistency with past evidence, long time horizons, considering direct and indirect consequences, considering all alternative hypotheses, and so on would prevent the obvious failure modes, it may nevertheless be that there exists no satisfactory utility function that would be safe using the Goldpan. That would certainly be compelling reason to abandon it.

While I think martial arts are pretty useful by hobby standards (although their usefulness is broad enough that they might not be optimal for specialists in several fields), several historical and cultural factors in their practice have combined to create an unusually fertile environment for certain kinds of irrationality.

First, they're hard to verify: what works in point sparring might not work in full-contact sparring, and neither one builds quite the same skillset that's useful for, say, security work, or for street-level self-defense, or for warfare. It's difficult to model most of the final applications, both because they entail an unacceptably high risk of serious injury in training and because they involve psychological factors that don't generally kick in on the mat.

Second, they're all facets of a field that's too broad to master in its entirety in a human lifetime. A serious amateur student can, over several years, develop a good working knowledge of grappling, or of aikido-style body dynamics, or empty-hand striking, or one or two weapons. The same student cannot build all of the above up to an acceptable level of competence: even becoming sort of okay at the entire spectrum of combat is a full-time job. (Many martial arts claim to cover all aspects of fighting, but they're wrong.)

Despite this, though, almost every martial art claims to do the best job of teaching practical fighting for some value of "practical", and every martial art takes a lot of pride in its techniques. As a consequence, there's a lot of posturing going on between nearly incommensurate systems. There have been various attempts at comparing them anyway (MMA is the most popular modern framework): they're better than nothing, but in practice usually come out too context-dependent to be very useful from a research perspective.

On top of that, there's a tradition of secrecy, especially in older systems (koryu, in Japanese martial arts parlance). Until well after WWII, it was uncommon for any system to open its doors to ethnic outsiders, often even to familial outsiders. Until the Eighties it was uncommon for systems to welcome cross-training in their students. Many still require instructors to have trained in only the system they teach. This is intended to prevent memetic cross-contamination but in practice serves to foster the wide range of biases that come with isolation and hierarchy: you can make almost anything work on your own students, as Eliezer's memorable example about ki powers demonstrates. (If you're feeling uncharitable, you could probably make an analogy here to the common cultic practice of isolation.)

Finally, a lot of selection pressure's eased off the martial arts in the modern era. During the Sengoku era, for example, Japanese martial arts were clannish and highly secretive, but it didn't matter too much: two hundred years of warfare made it very clear which taught viable techniques, if only by extinguishing poorer schools. Most other martial cultures were in a position to gain similar feedback, if less intensely. In the 20th century, though, martial arts grew more or less disconnected from martial applications: most militaries still teach simplified systems, but martial arts skill rarely decides engagements, and when it does it's in a narrower range of situations. Same goes for all the civilian jobs where martial arts are useful: there's feedback, but it's narrow, uncommon, and hyperspecialized.

I think there are ways around all of these problems, but no arts that I know of have done a very good job of engaging them systematically (though at least the more modern intersectional martial arts are trying -- JKD comes to mind). This actually wouldn't be a bad exercise in large-scale instrumental optimization, except that it requires a pool of talent that at present doesn't exist in any organized way.

(Disclaimer: as is probably obvious by now, I am a martial artist.)

Personally I suspect that the bathwater only really gets dirty when you are teaching something that is essentially useless in modern society, like martial arts or literary criticism. Most people who study, say, engineering don't do so in the hopes of becoming teachers of engineering.

Now you might say that this is because teachers of engineering are expected to also do research, but firstly that doesn't explain the disparity between fields, and secondly, I don't think that the example of tertiary education is one to aspire to in this way. I seem to recall you are an autodidact, so you may not have the same trained gut reaction I do, but I have seen too many people who did not have the skill of teaching but were good researchers teaching horribly, and I remember one heartbreaking example of an excellent teacher denied tenure because the administrators felt his research was not up to snuff too well, to want to optimize rationality teachers on any basis other than their ability to teach rationality.

Would it be helpful for us to try out these exercises with a small group of people and report back?

There are many ways to do, even small, contributions for everyone. The easiest is giving money (to someone whom you believe is trying to address the "really hard problems"). But there are many others. I would take two examples of things I do (or plan to do in the short future) : I'm helping with the French translation of HP:MoR and I'll (try to at least, nothing serious done yet) help SIAI with migrating their publication to their new LaTeX template (see http://lesswrong.com/r/discussion/lw/9d3/new_si_publications_design/ ). Both are tiny contributions, but which can in the hand help in various ways the SIAI to tackle the really hard problems. A lot of people doing those small things can allow the great things to happen much faster.

Of course, you can replace SIAI with anyone you think could solve the hard problems - other kind of research, charity, political party if you believe a given one is doing more good than harm, ...

The hardest part in that is probably in identifying who is more likely to actually help in solving the really hard problems. I tend to "invest" my energy and money in different kind of entities, hoping at least one of them will do something good enough on the long run.