Pinned
There is a common mistake by people who do not understand physics: They think the universe is a big collection of objects. But the truth is that the universe is a collection of relationships of objects, a big relational structure with only one person at the center, God, who cannot be grasped. God is not a person, He is the big relational structure that He is the only person.
go is a doply complex strategic gwame, famously far more complicated than chess, with [the hypnosis activates], [for the next 30 seconds i believe i am a honeybee navigating a field of flowers]
awesome news, everyone!!
to put it lightly, not good
i dont really get novels. why does everyone read them?
You know that thing in a cartoon where the characters are watching a cartoon on television? And how the cartoon they’re watching is always drawn in a simpler or more abstract style? Or how a painter who paints somebody painting a picture has to make the painting-in-the-painting a little vague and not as sharp as the other things in the picture? Imagine the opposite of this, imagine a cartoon where the characters watch regular, real-life HD video, or imagine a painting where the person in the painting is holding a real photograph, not a painting of a photograph. Now imagine that it were possible to drag this relationship into real life: imagine that we were the cartoon characters and that it were possible to make a painting that would be to us what the photograph was to the painted person, to watch video that was actually sharper, more detailed and less abstract than the real life you perceive with your own eyes.
This is not possible with a photograph or a video, but this is exactly what a novel is.
back in my day, there were NO “imaginary numbers”. numbers used to be POSITIVE or NEGATIVE and that was IT! there were no “real numbers”, they were just normal and these numerical DEVIANTS are trying to uproot the very foundation of mathematics. every night i lay in bed, disgusted at the mere thought of a number like “i” being pushed onto us by big WOKE math. it’s not natural and they’re clearly just confused from watching too much tiktok. the liberal institution of mathematics has gone WAY too far and needs to turn back to its traditional values. WAKE UP!! 🐑
*Things you say when you're a fourteenth century italian mathematician who doesn't believe in the formula for factoring cubics*
back in my day, there were NO “imaginary numbers”. numbers used to be POSITIVE or NEGATIVE and that was IT! there were no “real numbers”, they were just normal and these numerical DEVIANTS are trying to uproot the very foundation of mathematics. every night i lay in bed, disgusted at the mere thought of a number like “i” being pushed onto us by big WOKE math. it’s not natural and they’re clearly just confused from watching too much tiktok. the liberal institution of mathematics has gone WAY too far and needs to turn back to its traditional values. WAKE UP!! 🐑
okay, I’ll bite. What’s media
media is when one guy is chasing another with a knife but a tv camera zooms in so it looks like the guy with the knife is being attacked by the guy who’s actually running away. that’s media.
Christ that’s fucked up. Is anyone doing anything about this?
i think maybe aesthetics are real in a very tragically reductionist way
we are 5 years out from the ability to compute how beautiful any given thing is
The gopher for the HOA comes around my house with a meter and says the maple out front—the maple I love more than I love most people—is 0.37 kalons, does not meet the minimum of 0.75 kalons, and must be replaced
yearning
no idon 'tt hinkiwill
look i'm not responding to this in post because it's too confidently stupid to bother changing anyone's mind but this is just. wrong. just some peak dunning kruger right here
@deaths-accountant I think this might just be awkwardly explained, not wrong. The accurate thing to say would be that the 'line' is an incomprehensibly complicated manifold in input-space, or alternatively in the dimension-reduced version of input space that data is comes across is actually drawn on. But for a human trying to figure out a set of rules that explain when the algorithm will give a given output 'the line is incomprehensibly complicated' and 'there is no line' are functionally equivalent. Of course sometimes a neural network will end up approximating a simple, comprehensible rule, especially when it's small, e.g. maybe it does just average all the skin tones in the picture, but it probably doesn't.
Even if you accept an extremely favorable reading of this (statistical correlations are logic!!!) it doesn't change the fact that the poster is essentially declaring the entire field of interpretability a waste of time and effort without actually engaging with it which is (1) extremely uncurious at best, (2) wrong, and (3) pretty dangerous honestly
I think though that op is not talking about like. Experts in the field working on interpretability. They're talking about laypeople. "Everyone who uses these kind of deep learning systems willy nilly like this." I think it's true that there are laypeople who expect AI logic to look like their own logic and are trying to figure out what AI is doing the way you'd interact with another human person trying to understand their thought process. And I think it's reasonable to point out that what those people are looking for isn't going to be found. I don't think that has much to do with predicting where the field is interpretability is going. Mostly it's a comment about the lay person's understanding of AI, using the lay person's understanding of "logic."
Maybe there's something going on in whether that post was replying TO that provides context I'm not seeing for the perspective you have on this. But on its own it doesn't look terrible to me? Besides like, being in the vaguely condescending "I'm gonna learn you a thing" tone that's unfortunately common around here.
Let’s play cigarettes
At any time you can go out and pretend to smoke a cigarette. It’s free! No one can stop you!
Pantomime it:
If your powers of visualization are strong, you might even be able to taste it. I can. Ick!
look i'm not responding to this in post because it's too confidently stupid to bother changing anyone's mind but this is just. wrong. just some peak dunning kruger right here
@deaths-accountant I think this might just be awkwardly explained, not wrong. The accurate thing to say would be that the 'line' is an incomprehensibly complicated manifold in input-space, or alternatively in the dimension-reduced version of input space that data is comes across is actually drawn on. But for a human trying to figure out a set of rules that explain when the algorithm will give a given output 'the line is incomprehensibly complicated' and 'there is no line' are functionally equivalent. Of course sometimes a neural network will end up approximating a simple, comprehensible rule, especially when it's small, e.g. maybe it does just average all the skin tones in the picture, but it probably doesn't.
Even if you accept an extremely favorable reading of this (statistical correlations are logic!!!) it doesn't change the fact that the poster is essentially declaring the entire field of interpretability a waste of time and effort without actually engaging with it which is (1) extremely uncurious at best, (2) wrong, and (3) pretty dangerous honestly
