Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[silverspawn]

Boredom (this take isn't QRI specific) is essentially a mechanism that evolution came up with to regulate the exploitation/exploration tradeoff. I.e. without boredom, we would just keep doing the most pleasurable thing over and over again, but in fact it's better to try new things for the information gain. If music becomes too predictable, your brain's boredom module triggers and artificially makes it unpleasant.

It's also interesting that boredom is just about the only universally negative emotion. Other traditionally negative things like fear, awkwardness, sadness, and probably a whole bunch of others, are sometimes evoked intentionally through media and such. But nothing aims to be boring. Although lots of stuff is boring anyway.

And boredom seems to be triggered by thinking about how repetitive things are given that it completely goes away if you're being mindful. Like many things evolution programmed, it can totally be gamed.

[silverspawn]

Although I don't get why any movie makes things artificially awkward. I don't think I ever appreciated this emotion. Same with gross stuff as I've talked about before. But other people do seem to like it. And I'm totally on board for sadness and fear.

[silverspawn]

Here is the most trippy thing I know in the AI safety space, as far as I understand it

Suppose you want to program a system to make reasonable decisions in full generality. You then need to be able to solve this extremely  difficult and complicated dilemma:

Choose between taking a 5 dollar bill and a 10 dollar bill.

The reason this is complicated is that, assuming you take the 5 dollar bill, you can prove "If I took the 10 dollar bill, I would get 0$.". This is because "I take the 10 dollar bill" is false, and "false implies x" is true for any x.

And worse, you can also prove "If I take the 5 dollar bill, I get 5$; if I take the 10$ bill, I get 0$". This is because a thing called Löb's theorem, which says that the following sentence is true and provable in any formal system that is at least as strong as basic arithmetic, and for any P:



The box means "provable" so this sentence says "if you can prove that [a proof of P implies P], then you can prove P". This sounds totally insane because obviously a proof of P should imply P; this shouldn't mean P is provable. But for reasons I don't quite understand, it is a theorem.

The natural language analog to Löb's theorem is roughly this:

ErrinF created the greatest dominion expansion of all time. Don't believe me? Here is the water-tight proof.


Let S be the sentence, 'if S is true, ErrinF created the greatest dominion expansion of all time'

Suppose S is true. Then the above is true, i.e., it's true that, 'if S is true, ErrinF created the greatest dominion expansion of all time' -- and since S is, in fact, true, that means ErrinF created the greatest dominion expansion of all time.

Thus we have shown that, if S is true, ErrinF created the greatest dominion expansion of all time.

Since the above sentence is precisely S, that means we have shown that S is true -- without assuming anything.

As shown above, this means that ErrinF created the greatest dominion expansion of all time.

(You can't do this in a formal language because a sentence can't be about itself, i.e., you can't define S := S -> X, but you can can construct a sentence S that is equivalent to []S -> X. With [] being the "provable" box. I even used to know how to do this.)

So if your decision algorithm does the provably better thing, then if P = "if I take the 5 dollar bill, I get 5$; if I take the 10 dollar bill, I get 0$", then you can prove that []P -> P. That is because []P means you've proven P, i.e., you've proven that taking 5$ gives 5$ but taking 10$ gives 0$, and since again you do the provably better thing, you take the 5$.

So you've proven []P -> P, and because of Löb's theorem, you can thus prove P

and then you take the 5$.

[silverspawn]

That said this isn't really case of the The One True Principle That Is All You Need To Get Wrong About Philosophy, it's more of not quite understanding how to reason if you don't get it wrong

[silverspawn]

Also,

l = chr(123) # opening curly bracket
r = chr(125) # closing curly bracket
q = chr(39) # single quotation mark
t = chr(9) # tab
n = chr(10) # linebreak
z = [n, l+f"z"+r+q+n+t+"+ f"+q]
x = [n+t, l+f"x"+r]
e = [q, l+"e"+r+q+", end="+q+q+")"]
for i in range(0,2):
        print(f'l = chr(123) # opening curly bracket{z}'
        + f'r = chr(125) # closing curly bracket{z}'
        + f'q = chr(39) # single quotation mark{z}'
        + f't = chr(9) # tab{z}'
        + f'n = chr(10) # linebreak{z}'
        + f'z = [n, l+f"z"+r+q+n+t+"+ f"+q]{z}'
        + f'x = [n+t, l+f"x"+r]{z}'
        + f'e = [q, l+"e"+r+q+", end="+q+q+")"]{z}'
        + f'for i in range(0,2):{x}print(f{e}', end='')

[silverspawn]

unrelated, it does seem to be true that a lot of people are trying to get involved with NFTs hoping this will make them money (someone invited me to a discord server after reading my reddit post, which doesn't make any sense because I disavow any financial motive in the post), and all of this does seem really stupid. In particular, it's all zero-sum; if a bunch of people spend time on NFTs for money, the result is a lot of time invested for a bunch of redistribution and zero added value.

[silverspawn]

It's like playing poker except that playing poker is at least supposed to be for fun, so it's okay to not produce value

[silverspawn]

And I suppose that insofar as NFTs are used this way, calling it a ponzi scheme is accurate. But it stops being a ponzi scheme as soon as people are buying NFTs for fun without hoping for money

[silverspawn]

Just reading that Aubrey de Grey got fired from the organization he founded for some sexual harassment thing. Men are seriously morons. He's outright one of the smartest people out there, insanely high IQ and an ability to make sense and inherited a shit ton of money and using it to work on something that maybe matters at least a little (i.e., fighting aging). He literally made progress on a famous math problem in his off time, a problem that no-one has been able to contribute to for over a decade. Unquestionably a genius. And yet he just can't keep himself from molesting women and then try to cover it up. Whhhhhyyyyy

[silverspawn]

I wonder what the dark number is for famous  men who harass women. Because I do assume that the number of false positives is low and the number of false negatives substantial; and the number of true positives + false positives is obviousy substantial since that's what  we can observe.

[silverspawn]

Also very curious about the dark number of famous women harassing men. Probably a lot lower in absolute terms, but probably high compared to the observed number

[silverspawn]

I hereby nominate The Woman In The House Across The Street From The Girl In The Window for the dumbest plot twist of any movie ever made ever. Too bad, I genuinely enjoyed it up to the end

[silverspawn]

I guess the second dumbest after the sugar bowl in the series of unfortunate events

[silverspawn]

And no you don't get to say it's satire. The show is occasionally funny but up to that point it's trying to tell a story, not be a parody.

[silverspawn]

I feel like there is this thing that some writers do where they just sprinkle a bit of parody and a bit of comedy in there and a few references, and then they have a fully general rebuttal to criticism by just calling it a parody or whatever

[silverspawn]

At least it just makes me annoyed, not creeped out or angry. I'll give it a 5/10

[silverspawn]

Ok there is definitely a deeper point here about genres and what it means to be a parody, but I have to think about this

[silverspawn]

Nemurerumorinobijo

[silverspawn]

totally out of the blue some professor emailed me and offered me an interview based on the (rejected) application I wrote. this complicates things

[silverspawn]

she however does reinforcement learning in multi-agent settings, which is not what I signed up for, not something i know much about, and most likely not safety relevant

[silverspawn]

and her published research does not excite me, *unlike* that of the professors who I wanted to work with

[silverspawn]

now if I decline this, which I probably am going to, I get to say that I turned down a PhD. How cool

[silverspawn]

if I did accept this, it would imply that my true preference is something like "any PhD is fine" and all thoughts to the contrary were delusional

If the above is my true preference but I reject it in order to look consistent, I'm both delusional and a moron

but probably I'm just going to reject it because it's not good enough

[silverspawn]

Also have to question why she wants to work with me given that it doesn't seem to be an amazing fit

probably she just emailed a bunch of people and the formal part of the application looks decent or something

[silverspawn]

But yeah I can't do this. Can't commit several years to working with someone whose research doesn't even seem interesting let alone useful