Hacker News Reader

Comment by lalaithion | Hacker News Reader

That's why it's called "less wrong". The goal isn't to be perfect, the goal is to do better. To be less wrong.

I think this misses the point. Even being "less" wrong requires an amount of work that even the best/smartest etc... cannot consistently apply.

I do believe this is zero-sum in that improving on one set of decisions means no applying the same rigor to others.

This is often seen in the form of very smart people also believing conspiracy theories or throwing their hands up around other massive issues. As an example, the "Rationalist crowd" has de-emphasized work on climate change mitigation in favor of more abstract work on AI safety.

> This is often seen in the form of very smart people also believing conspiracy theories or throwing their hands up around other massive issues. As an example, the "Rationalist crowd" has de-emphasized work on climate change mitigation in favor of more abstract work on AI safety.

To be clear, the argument (in rationalist circles) is not that climate change is no big deal, it's that there's already a ton of people worrying about it, so it is better to allocate some extra resources to underfunded problems.

I wouldn’t be surprised to learn that even being less wrong is NP-hard.

The saving grace of being able to survive in the universe is that it's possible to climb up NP hard problems far enough to get real results with hard work.

It also means that you might not know if the hard work is climbing up or down (towards or away) from the solution.

No, you know if you're getting better or worse in an NP problem because checking answers is in P.

If you find a solution, yes

Oh, you're thinking of NP-hard yes-no problems. Many if not most NP-hard problems of practical importance, including the traveling salesman, involve integer rather than boolean scores.

NP-hardness by definition is only yes-no decision problems. The NP-hard formulation of Traveling Salesman is "given the weighted graph G and an integer X, is there a Hamiltonian cycle in G with total weight less than X?"

You are using a non-standard definition of NP-hard. Here are standard definitions.

A problem is in P if there is a polynomial time algorithm to solve it.

A problem is in NP if there is a polynomial time algorithm to check a purported solution.

A problem is in NP-hard if, if there was a polynomial time algorithm to solve it, every problem in NP could be solved in polynomial time.

A problem is NP-complete if it is both in NP and in NP-hard.

For Traveling Salesman, the NP-complete problem is, "...find a solution with total weight less than X." (Linear verification, check it is Hamiltonian, check its weight.) An NP-hard version is, "...find whether there is a solution with total weight less than X." (To verify, you have to search. Oops.) But ANOTHER NP-hard version is, "...find the solution with least total weight". (To verify, you have to search. Oops.)

Thanks for the additional clarification :)

Indeed. And thanks! I needed a little morale booster now, for reasons unrelated to this topic.

Which is ironic because the pursuit of knowledge only continues to increase the landscape of unknowns towards infinity - the branches of the tree of undiscovered and unknown knowledge continues to grow exponentially. It's as-if today we thought the choices were A or B, yet tomorrow we discover there was a C, and the next day D and so forth. If anything we are only discovering we are "more wrong" every day.

Actually I think this is the same fallacy as one of Zeno's paradoxes, and has the same resolution. We are discovering more wrong, as you say, but the "infinity" of wrongs is in the direction of the infinitely small (or "infinitely detailed"), not the infinitely large. In other words, every time we fill in a gap in our knowledge, we create two more gaps, so to speak, but nevertheless we know more than we did before.

> the goal is to do better

Why is that "the" goal?

Who sets "the" goal?

Its the goal of the LessWrong/Rationalist community.

Less Wrong seems to be a manifestation of a thing that comes in cycles: Something triggers the rise of a "rationalist" movement, including possibly a new evangelist or a new medium. Eventually, rationalism and rational people end up at a standoff. Then the whole thing repeats itself after a period of time.

I'm probably rational enough but also can't make sense of much of the rationalist literature, so I simply follow my own compass and hope for the best. I'm skeptical of Bayes Theater.

The goal is to make decisions that are "better" as defined by your own utility function given limited information. This is also called "winning"

Yudkowsky and other prominent contributors set "the" goal as a sort of revealed wisdom regarding their speculations about what a super powerful AI will do.

Pragmatically, the goals themselves appeal to individuals who want to maintain conventional (liberal) morality yet also position themselves as superior, typically as a form of compensation.

position themselves as superior

This is why we can't have nice things. Any time someone tries to find more effective ways of making good decisions or accomplishing their goals, someone has to bring out the most tortured cynical interpretation to tear them down.

Have you hung out much with rationalists?

Have you? Because the rationalists I know are genuinely well-adjusted people.

It's cheap and easy to make fun of the lesswrong community as a cringy cult of AI-obsessed neckbeards. And to be fair, the writing style on LW tends to support that impression. But I've found that most of the actual people within the rationality/AI safety/effective altruism communities actually don't fit that stereotype at all.

Yes, I was in the community for several years and then "left" but still spend a lot of time on its edges. I'm not trying to call them "neckbeards" (although I knew many of those too), only saying that there is an air of superiority without any actually radical content.

I consider EA separate but related, and it definitely qualifies as staking out a superior position within the constraints of liberal morality.

I went to a couple of meetups and found one cool dude surrounded by a group that exactly fit the stereotype, FWIW.

I went to a couple of meetups and never met anyone who would fit that description.

But then, as a lesswrongy person, maybe it's me, maybe "feeling superior" or whatever is just normal to me :shrug:

(But my actual theory is, it's probably just geographical.)

i consider myself a rationalist (or at least, an aspiring rationalist), and people like hanging out with me. learning about this stuff has changed my life and relationships for the better.

Interesting, can you give a specific example of using rationalist techniques to solve a relationship problem?

Sure. In the past, I've been a person with a strong desire to be right in every situation. If I thought I was right I would be very willing to stubbornly argue my point, without any consideration of the counterpoints that were presented to me. Rationality is in part about understanding that every piece of knowledge I have is probabilistic - that there's a chance that each belief I have is wrong, and that I need to be able to let go of a belief if I encounter enough counter-evidence. I'm sure you've met people who can't be swayed from a belief no matter what you tell them, and you know how difficult it can be to get along with those people. Internalizing the methods of rationality have helped me transform from one of those sorts of people to a more reasonable one whom people have a more pleasant time interacting with.

As a specific example, I made a comment to my roommate last winter about how I thought his girlfriend's hyper caution around COVID was limiting my personal freedom. I realized that I had been crass and apologized to him, but he told his girlfriend anyways and it caused a great deal of tension between the three of us. My own girlfriend told me I should apologize to her. I believed I had nothing to apologize for since I hadn't said anything to her directly, I didn't believe my roommate should have repeated the comment to her in the first place, and I had apologized to him for it already. My girlfriend gave me reasons why an apology was in order, though, and I assigned a lot of weight to her reasoning since I know her to be a more sensitive and emotionally intelligent person than myself. I was able to let go of the belief in my own righteousness and write a heartfelt apology which did wonderfully to mend the relationship.

A previous version of me would have clung to the belief that I was in the right, and either not apologize or write a half-assed apology that would do nothing to fix the situation. The current version of me which strives to be rational was aware of my own biases, recognized that my internal map may not match the territory, and was willing to update based on the evidence from my girlfriend's greater authority on emotional matters.

Reply on news.ycombinator.com