@UnrelatedString Hm, I guess. Just bothers me that questions that have perfectly fine and answerable for years are being closed just so someone can game their numbers

I miss when Popcons were more common.

Some fantastic Graphical Popcons back in the day™

IMO there aren't any good popcon ideas anymore.

Our standards also changed. Many older popcon challenges would not survive if published today.

I have set of 100 points in 10d called A. How can I find a set of ten 10d points B that minimises the total shortest distance from points in B to their closest counterpart in A?

CMQ ^^^

Why do I fear this is a Traveling Salesman problem

@ATaco I hope it isn’t!

@Simd This sounds like you just want to take every point in B, find its distance to every point in A, find the minimum of those distances, then sort the points in B by those minimums, and then take the first 10 of those points

@mathcat Depends on what you consider good, and what you consider too similar to an existing challenge. I think there's room for more language-design challenges like this one, and popularity-contest might be the best winning criterion for those.

What is this for?

@DLosc I would love to see more of those

@user sorry the task is to create the set B

@mathcat no, but some challenges are just a tad too subjective

Can't you just pick any 10 points in A then?

Surely a set B which is any 10 elements of set A has the definite minimum distance to its closest neighbours?

And then the shortest distance from any point in B to its closest counterpart in A is 0

@user yes you could but would those be the best?

Oh I see what you mean

What's this for?

i mean the total sum of distances from A to the closest counterpart in B

I agree, what's this for?

This sounds XY

@user just for interest

Oh come on, you didn't just come up with such a specific thing for a CMQ

@ATaco yes I had it backwards, sorry

It'll be easier to help you if you say what it's for

@user I like the idea of summarising lots of curves with a small number of other curves

I like curves too

Sort of curve clustering

Distance is Dimension agnostic, so you can simplify the problem to 2d

Ok

Basically it’s a little like finding centroids but where the centroids don’t need to be from the data set

I just don’t know how to do that

I feel an expert would know what this is called

I think this is an NP Hard problem

@ATaco oh?

But in an ideal solution, all points are equidistant from one another, in within the outer bounds of the initial dataset.

Why do you say it is np-hard?

Because it's not in P and at least as hard as problems in NP

jeremyjordan.me/grouping-data-points-with-k-means-clustering seems related and in P

Perhaps you should ask Math Exchange :P

@ATaco you are cleverer here :)

Really math.se and not SO?

Probably not a programming problem honestly

this seems like it could fit on math or maybe better on CS.SE; doesn't seem like the best fit for SO though

4294967295&0xFFFFFFFF // -1

That's not helpful JS

@hyper-neutrino ok thanks

@ATaco I guess maybe it is if you want it to be fast?

@Simd K-means clustering is NP-hard, right?

@user yes you are rght

@hyper-neutrino How to find points that solve the problem -> Math.SE; best algorithm for finding those points -> CS.SE; ran into errors while trying to code the algorithm -> StackOverflow; code is working but needs to be faster -> CodeReview.SE; forget fast, I want it as short as possible -> CodeGolf.SE ;P

@NewPosts Also @hyper-neutrino can you 11 this post to appear nicer on the starboard please?

@DLosc How to make it even faster -> Quantum Computing; How to make it faster most of the time -> Cross Validated; How to show people how fast it is -> Webmasters; How to ensure that I'm effectively communicating that information -> Ui/UX :p

@RydwolfPrograms that's because nowadays we spend our time telling at pre-commit ci, weird javascript stuff and things that don't compile :p

Just ask ML to do it, they'll give you the looks-correct answer :D

Yeah Cross Validated does ML stuff here right

@RydwolfPrograms I propose we call this a "computer prion"

Ooh that such a great metaphor

huh, why is my code here bolded? and the other asterisks look to be italicised? weird

must be a unity problem

@JoKing I noticed that too. It seems like it doesn't happen if you remove the Raku syntax highlighting.

Does Raku really call bigints FatRats or is that some sort of library or reference to something?

'cause if so that's...an interesting name?

Probably worse even than pickle

I mean, Perl has terms like "slurp" and builtins like chomp, so it fits :P

clojure uses slurp too

and rust uses yeet

@RydwolfPrograms The default number type is Rat, which is rational until a point where it decays into the floating point type Num. The naming scheme has been pointed out as pretty dumb, but they excuse this by saying they can't change it now

Ohhh so it's a rational type. That makes sense.

I thought it was like, some sort of really weird rodent-related metaphor

BigChungus

DLosc, I'm really getting worried about you

The meme culture is really getting to you

@DLosc Heheh that'll hold em all right!

@DLosc I love explode and implode

Or is that PHP?

Yeah, that's PHP

@Seggan Sadly, once it's actually added to the language, it'll be some boring keyword instead of yeet and yoink

PHP also has chop

Why so violent?

how about ruby's chomp

@DLosc Perl-inspired, right?

whoops lol

@user So does Pip (as CH)

but of course it does something different from either Perl or PHP

because Pip Isn't Perl and Pip Isn't PHP

raku gets nsfw with .nude

axonflux.com/5-quotes-by-the-creator-of-php-rasmus-lerdorf

@JoKing What does that do?

it's short for numerator-denominator, returning a pair of numbers representing the fraction of the input

...they didn't think of a better name?

what do you mean? what would be funnier better?

.frac maybe?

@user I don't know if that's true, but if so it explains some things

Or .num_den

I really hate how C functions are named

strstr, for example

@pxeger Hey, give credit where credit's due. I worked my ass off to make most of those bugs. You can't let these damn AIs take our jbos!

Tbh being a human, I would fully expect copilot to product code better than a human

Humans get tired

Case in point :P

@RydwolfPrograms But Copilot doesn't truly understand the code, so if you use it for anything more than the most basic stuff, it might produce code that's more elegant than what a human would write but that's also not right for the current situation

And humans sometimes forget that var, even in a tighter scope within a function where it may never even run, still overrides any instances of that identifier across the function, even before the statement appears, causing variables from the parent scope to be undefined if you reuse a variable name in a context where you wouldn't expect it to cause any ambihuity

Totally didn't spend fifteen minutes tryna debug my awful code because of ^^ today

f

Honestly the first time var being function scoped has actually bitten me

Wonder if some day we'll have (good) AI that isn't some kind of neural network blackbox

Well I mean "good" AI would either mean more humanlike AI, or more predictable AI

Though it seems like a lot of research has been focused on neural networks in the past few years

Neither of which is really that good

Not a general AI, but something like Copilot

I'm just waiting for AGI

I hope it happens within my lifetime

AGI would cause a whole host of ethical problems to pop up

I kinda want to see it happen in my lifetime but I'm also afraid I'll see it get screwed up

It's 100% possible, has to be. If a squishy meat lump in my skull can do it with 25 watts of power, we can make it run on silicon

And then we can make it scale

Imagine what a single mind with the power of 2500 humans could do

Yes, but how are we going to deal with something like that? Should it be given the same rights as a human? (probably not, because it has entirely different needs, but what if some idiot creates an AGI that can pretend it has feelings? Is that the same as actually having feelings?)

We'd have problems with that sort of stuff if we ever meet aliens too

All we'd need to do is make sure it understands our most fundamental ideas about morality, and it would fill in the rest. Idk if it actually exists, but I think I've noticed a pretty strong correlation between intelligence and empathy.

The most powerful use of that AI, IMO, would be processing and unifying data

Just telling it what we value isn't going to make it value the same things

Imagine giving that AI every scientific paper published in the last century

It would be able to find so many connections, provide such effective summaries and explanations, it would revolutionize science and education

That AGI will still have biases, and it'll be very hard to detect when its biases are at work

We just need to engineer it with empathy and get it to empathize with us

just

AI safety is a very hard problem

@RydwolfPrograms I'd assume that by the time humanity gets anywhere close to that, we'll probably be close to being post-scarcity

And by reading about bias it would probably figure out how to confront its own biases

@user Idk, we have a lot of computers

@RydwolfPrograms Gonna press X on that

If we can figure out how to get AI to run on one computer, we can get it to run on a zillion

@RydwolfPrograms lol no

@user Humans can do it, why not a human-like AGI?

Can they really?

^

Yeah...?

And it won't really be human-like

Unless you make some sort of cyborg augmented with a bunch of chips in its brain

I grew up in a conservative, at times close to alt-right household. It took me years to realize all the awful biases I'd picked up. Being racist, transphobic, etc. is practically the default for humans

Most of the time, telling someone about their bias will make them lean harder into it

@user Why?

@RydwolfPrograms i read somewhere some nuclear fusion research did that and got good results

@JoKing Actually, I think that one hasn't been proved

We're theorizing here, why not let it be human-like in our theorizing

@JoKing What on earth makes you think that?

@RydwolfPrograms Why would it be? Neural networks work vaguely like human brains, but they're still not that similar

@user And obviously neural networks are not AGI

@JoKing That's just a bias you have

lol

@RydwolfPrograms Okay, maybe this hypothetical AGI will use a different thing

But still, why would you believe it'd be human-like?

Because that's literally what makes it AGI?

@RydwolfPrograms this is known as the backfire effect

@RydwolfPrograms ???

@RydwolfPrograms no, that's definitely not the definition

@JoKing Has this not been disproved?

> Artificial general intelligence (AGI) is the ability of an intelligent agent to understand or learn any intellectual task that a human being can

Close enough

Intellectual tasks

They might be able to mimic what a human would do but they won't necessarily feel emotions or anything

No need for emotions

Emotions are just heuristics from back when we were dumb

We don't just think with our brains, we have a bunch of glands making chemicals or whatever influencing us too

I'm not sure how it could be human-like without emotions

We don't necessarily have ethics out of the gate either, or at least we can be trained out of them

We definitely have some basic ethics

No way were the scientists at Unit 731 or Nazi concetration camps using the same ethics you and I do, for example

@user "subsequent research has since failed to replicate findings supporting the backfire effect", interesting

If you raise a baby without any rules, it's not (necessarily) going to become a psychopath

> necessarily

@RydwolfPrograms Last I checked, not all of humanity was as crazy as those scientists

No, because we live in a society that makes an effort to continue being ethical

@RydwolfPrograms If there's one thing Worldbuilding.SE has taught me, it's that things you'd expect to be able to scale often can't.

Sure, babies are born to be prejudiced against those they see as "others", but they also feel bad for those they see as being part of their families or whatever

ethics system = goal function in AI

@RydwolfPrograms Strongly disagree

@DLosc That's true, yeah. Could be that there's like, an exponential cost to increasing the power of a general intelligence.

Emotions stop us from hurting each other

@JoKing Not necessarily...

My goal function is to be happy, right?

We wouldn't have survived as a species if humans all had to be trained out of killing each other

@RydwolfPrograms Everyone has different goals

And I don't really derive happiness from someone on the other side of the world not dying, but I still really don't want them to die

@user Sure on some level, but our goal function is happiness

are you sad when you hear that someone died?

A negative utilitarian would seek to reduce sadness instead of maximizing happiness (and then there's average happiness, total happiness, maximum minimum happiness, so many metrics)

@RydwolfPrograms That's its whole point

@RydwolfPrograms Our goal function is keeping the species going

No, that's evolution's goal function, and it's for genes not species

happiness is mostly a byproduct rather than a goal

I am not constantly using every second of my life to try to have more children

Well if you're talking about personal goals, not everyone's goal is to maximize their own happiness

That's not what we're wired to do

^^

@user I'm not talking about personal goals either. I'm talking about our brains' goal functions.

Yeah, I'm not sure our brain does that either

what's the difference

I think it does, tho in a more complicated way since there's different chemicals responsible for different things

Can we ask about this on some SE site?

Philosophy

bah

Bunch of ivory tower intellectuals

Philosophy.SE is just shower thoughts stack exchange

No serious discussion belongs on there

Is there a Neuroscience SE?

Yes

^^^^

But this is probably too vague for that

i'm not actually sure what the disagreement is

Multi-leveled

We've moved on to different disagreements over the course of the conversation

Move to OTTNB?

@DLosc (just want to clarify, the "bah" wasn't directed at you, I was just expressing that I don't like how a lot of posts on Philosophy.SE seem to be pure conjecture about random abstract stuff without any evidence whatsoever backing them)

Why would AGI be humanlike? Why would being more powerful make it better at empathizing?

I agree that knowing that you're biased and how you're biased will help you reduce it, but not eliminate it entirely

Because we're just somewhat quirky, biased AGIs

But other AGIs don't have to be like that

Like what?

Are you implying that if alien life pops up somewhere else, they'll also think in ways similar to humans?

@RydwolfPrograms Like humans

You keep saying AGI would be humanlike

for point one, if you had the option, would you wirehead yourself?

Not Rydwolf, but I definitely would

@user I mean, necessarily, yes. Other than domain-specific things like sensory processing, motor skills, and emotions, I think most of the remaining thinking is pretty general and would work the same for aliens or AIs

(it's the thing where you go into a simulation and become happy forever, right?)

> other than ... emotions

An AGI wouldn't necessarily be good at empathizing, because it wouldn't have human emotions

more akin to shortcircuiting the orgasm button in your brain

That works too

And as for empathy, isn't a necessary consequence of being able to think in general terms the ability to think from another thing's perspective?

I don't have hit points or a hunger bar, but I can empathize with Steve not wanting to take a heart of damage

Even if you can think from someone else's perspective, you won't necessarily care about them

@user No offense taken. I did find it a bit amusing that you put "ivory tower intellectuals" and "shower thoughts" in the same category... I'd think "ivory tower" would produce long answers full of impenetrable jargon and oblique citations to a bunch of authors I've never heard of, while "shower thoughts" would produce one-paragraph answers without any citations. Philosophy.SE does seem to attract both of those kinds of answers, but they're not the same thing. :P

An AGI might know that a human would be sad if a robot chopped their arm off, but that wouldn't stop the AGI from chopping their arm off

An AGI wouldn't have "happiness" as a goal or "anger" as a bias, but it could understand that. It would just need to have a goal of wanting to help its fellow AGIs maximize their utility functions.

Which put in those terms isn't a particularly complex goal to communicate

@RydwolfPrograms (most) humans also have the instinctive ability to literally feel what we observe others to be feeling

@DLosc I mean, weren't a lot of early philosophers like "Hey, I think this is how the world works, and I'm totally convinced so why bother going out and experimenting to see if it actually works that way"?

@JoKing Yes, but that's just an optimization, right?

(also, it was Rydwolf who called it shower thoughts SE, but I do agree with him)

sure, but an AI/alien wouldn't necessarily have that

@RydwolfPrograms Hardware acceleration :P

and wouldn't necessarily want that either

No reason an AGI couldn't understand "what we're feeling" ("how well we're fulfilling our utility functions") and have that mapped that directly to its own utility

@JoKing Well, an AI/alien could do "software virtualization" to try to guess what we're feeling, but yeah, they still wouldn't necessarily care even if they knew we were feeling sad

It's a two part problem

Kinda hard to program an AGI that's a blackbox

@user I don't think so (note that an early term for science was "natural philosophy"), but I'm not super familiar with the history of philosophy so I can't say for sure

@user No matter how blackboxy an AI is, one of its paramters is still a utility function

A mind floating in a void would have no cause to think about anything meaningful at all

@DLosc Neither am I, so I guess I kinda did exactly what I just accused early philosophers of doing :P

the utility function can also be a blackbox

@JoKing No

@RydwolfPrograms are we bringing in Bolztmann Brains now :P

@RydwolfPrograms then clearly happiness is not your ultimate goal

Well it's my brain's final goal

Hence why if I did a bunch of drugs I'd get addicted

you are your brain?

I'm my brain's thoughts, I think there's a difference somewhere

feels like a semantic difference

I think this is a mesa-optimizer sorta situation

@JoKing Seems a bit reductionist--the rest of the body is important too

that is also you sure

(even leaving out the possibility of a mind or soul distinct from the body)

And I think that probably makes me a mesa-optimizer within my brain, in the same way that my brain is a mesa-optimizer within nature

ok, i think understand your point of view (though I personally don't think there's any divide between your brain and your brain's thoughts)

Consider playing a video game. By doing that, you're substituting a foreign utility function in for your own, in a way. I think that's basically what a goal of, e.g., making the world a better place, or having a family, would be

@RydwolfPrograms I think that's because since 2021, three things have happened: a) we've gotten better at golfing, so we no longer need to have a quick byte save most of the time, b) the language has gotten better with more things making programs already short enough that flags don't have a real use and c) challenges have moved more and more away from having stricter output formats and to having just data in data out

Well also I think there's a social factor, I see a lot of flag-(ab?)using programs being submitted below non-flags one in the same answer

And there's still the non-I/O flags like the range base ones and r

@RydwolfPrograms I think that's more of a show that "we could go shorter, but we already are tying/winning so we won't"

at least, that's why I include flag versions

@RydwolfPrograms my initial read of this was "flag-(ab?) using programs", which made me wonder why so many answers would be using that flag

@RydwolfPrograms the range ones are mostly artifacts from when range generation was from 0 to n-1 and r is cumbersome for 99% of programs

That's fair I guess

keep in mind that the time you're talking about when it was "rampant" was the pre-2.6 releases. The language was really young then, and was missing a lot of features

There was only a maximum of 415 available commands during that time. Now there's over 550

2.6 changed a lot of things and made things generally better, and each version release has made more and more improvements

@RydwolfPrograms I'm also 90% sure those range flags are actually broken internally :p

I do respect how much Vyxal's evolved, I tend to stop working on my languages as soon as, if not right before, they go live

It's the result of some of the greatest golfing minds coming together and opening a whole bunch of issues telling me where I've gone wrong :p