@Whisper

Ya know, it always annoys me when people tell me how talented I am when I bust out some Chopin in a hotel lobby or whatever because it understates the 10,000 hours I spent sitting in front of my black Baldwin growing up.

I know I’m smart. I don’t need to hear it from you, and frankly it makes me uncomfortable because I work with people who are actually next-level smart. It was nice to hear when we started talking 5-6 years ago but I’ve learned that natural intelligence isn’t anything to be proud of any more than is being tall, and it also only gets you so far.

FWIW I made a pretty good joke about the hairy ball theorem in that math thread, but it got pruned for being uncivil. I’m the new fish so I’ll leave it at that.

In my opinion, formal systems are best regarded as generators of hypotheses, not theories.

Your opinion mirrors history.

Aristotle was the author of one of the earliest recorded theories of motion, one that was accepted widely for over 1,500 years. It is clear that he (like many theorists since) was not one to let experiments results get in the way of his theorizing. His “scientific method” was to let logic and common sense direct theory; the importance of experimentally validating the resulting theory only developed recognition centuries later.

Cynically, we can note that while the influence of Aristotle’s ideas have waned, since Galileo and Kepler did not leave much to the imagination, his method has not, and it continues to prevail in areas where the scientific method yields unpopular results.

Indeed, employing the modern day version of the scientific method and publishing the results can be thought of as a “Dick move”.

@fieldline

This is why, despite occasionally using the term "rational" in its colloquial sense, and using formal language tools such as math and logic, I would describe myself as an "anti-rationalist".

In general, communities of thinkers that describe themselves as "rational", or "rationalist", or "skeptical, tend to overestimate the usefulness of formal computation tools, because they underestimate the difficulty of translating real world problems into formal systems.

Always, we end up introducing some subtly wrong premise. Always, we end up assuming the chicken is a sphere to make the math easier. Which is where the need for empirical testing of ideas comes in.

People who pride themselves on "rationality", or people who spend excessive amounts of time with formal systems, tend to be seduced by the siren song of the easiest part of the problem. If only we could write everything in some LISP-derived language, and eliminate testing and bugs forever by formally proving our programs correct! If only we could conduct our political debates by diagramming everything out in first-order logic!

But no actual practitioner of building shit and getting shit done actually takes this seriously, because he knows that any formal-language representation is, at best, an oversimplified and inaccurate representation of the world, and that the real bugs lie not in the code, but in the gaps between the code and the problem description.

In my opinion, formal systems are best regarded as generators of hypotheses, not theories.

Which is why I'm using logic with this young fellow I'm trying to reason with right now. We haven't gotten to the stage of discussing whether his hypothesis is correct, because we haven't gotten to the stage of stating what his hypothesis even is.

Neither, I suspect, has he.

Which is why it's rather cruel of you to bring up complexity theory and polynomial time... others around here are still struggling with the concept of inverse, converse, and contrapositive.

We all know you're smart, already. Siddown and have a beer.

@Whisper

Humans are not naturally rational creatures.

It isn't just hard to be perfectly rational…. it is computationally impossible.

Everyone who took a semester of statistics in college knows that Bayes’ Theorem gives a mechanism to update our beliefs in the presence of new evidence.

But what if we have a network of interrelated beliefs (a “Bayesian Net”)? Luckily, while a bit more obscure, it turns out that Bayes' Theorem also prescribes a unique answer. Unluckily, it turns out that working out that answer is NP-hard. It is computationally intractable to calculate the single data point that should flip our worldview, despite our ability to prove its existence.

You can read all the lesswrong you want. You can masturbate to Thinking, Fast and Slow and learn why we fail as we do. But the one thing that you cannot do is have the smarts required to actually BE rational.

The effort of doing better is still worthwhile. But the goal itself is unachievable.