#applied statistics

[noosphe-re]

"There was an exchange on Twitter a while back where someone said, ‘What is artificial intelligence?' And someone else said, 'A poor choice of words in 1954'," he says. "And, you know, they’re right. I think that if we had chosen a different phrase for it, back in the '50s, we might have avoided a lot of the confusion that we're having now." So if he had to invent a term, what would it be? His answer is instant: applied statistics. "It's genuinely amazing that...these sorts of things can be extracted from a statistical analysis of a large body of text," he says. But, in his view, that doesn't make the tools intelligent. Applied statistics is a far more precise descriptor, "but no one wants to use that term, because it's not as sexy".

'The machines we have now are not conscious', Lunch with the FT, Ted Chiang, by Madhumita Murgia, 3 June/4 June 2023

#quote #Ted Chiang #AI #artificial intelligence #technology #ChatGPT #Madhumita Murgia #intelligence #consciousness #sentience #scifi #science fiction #Chiang #statistics #applied statistics #terminology #language #digital #computers

[signiorbenedickofpadua]

AI-generated embroidery

I'm starting to see more and more AI generated pics of "embroidered" art popping up, and somehow that makes me even more disappointed than when I spot fake paintings. Of course, a lot of creativity and effort goes into any given form of art, and they all have their limitations which inform the end result, but one can, in theory, paint almost anything (at least in digital painting) while there are far more tangible, physical limitations to what you can make a piece of thread do. One of the best parts of seeing a gorgeous piece of textile/fiber art is admiring how well the artist has been able to cleverly utilise and transcend the limitations of working with thread/fabric/roving/etc., and realising how much time it must have taken.

To then see AI "embroidery" that looks very impressive at first glance before realising that it couldn't possibly be real because it defies the very laws of physics makes me sad. Not only did someone want to generate a piece of art they couldn't be bothered to make themselves, but they then looked at the image the computer spat out and either wasn't interested enough in the craft itself to be able to tell how impossible it was or decided that it didn't matter because the point was just to get some superficially pretty content to share with a mostly unsuspecting audience. It's just... a whole new level of disrespect for a craft that is already niche and underappreciated as it is.

There have been posts pointing out what to look out for in generated "paintings" and "photos", but I have not seen any for embroidery yet, so here's a couple of points off the top of my head:

Can you see individual threads clearly? Do they look consistent in texture (like they're spun the same way). There are a few different types of embroidery thread which are spun in different ways and from different materials, but if it looks like one type of thread was used and the threads still look inconsistent, that's a tell.

If the piece seems to be mixed media (e.g featuring beads or fabric appliqué etc.), do the added materials look realistic? Can you guess at what they're made of or how they may have been attached without the use of sorcery? I have seen a few examples of what looks like a cluster of beads at first glance, but when looking closer they've turned out to be very surreal in shape and get progressively less realistic and more abstract the smaller they get. Then there have been details which I've tried and utterly failed to imagine what material they could even be made from, like decorative vines/borders that look more like 3D-printed plastic than thread or metal wire or anything you might expect to find in an embroidery, even with creative material choices.

Does the motif look very three-dimensional? There are a number of ways to create 3D embroidery, but they all have to follow the laws of physics. Where are those threads going? Do they just end randomly? Does it look like a bunch of normal embroideries stuck on top of each other with no plausible edges or methods of attachment?

I've seen at least one example where most of the image was trying to look like photorealistic embroidery, but then there were a few details that simply looked photorealistic as in, "Wait, half of that bird looks embroidered, and half of it just looks like a photo of a real bird?"

As in other forms of AI art, look at details which you think you know what they ought to look like - especially in parts of the work that are out of focus or far from the center, those tend to be less realistic. If, at first glance, something in the background looks like a rose, look again and ask yourself, "Is that really what a rose looks like, even stylised? Why does that petal look more like a weird tentacle? These shapes make no sense. Was the artist just clumsy and made a mistake, or was it designed like that by a computer?"

Humans like symmetry. If there's a border or something which looks like it ought to feature repeating patterns but turns out to just be abstract shapes without any sort of pattern to it, then be suspicious. Same thing goes for motifs which look like they're supposed to consist of a perfectly mirrored right and left part but have unexpected inconsistencies (say, a butterfly or a crown, for example).

Finally, as always: is there a source in the caption? Does it say something like, "[Title of the piece/description of the motif] by [artist], embroidery thread and [beads/metal wire/leather appliqué/whatever fun material]"? Is there a caption at all? If not, there's a tiny chance that the artist just posted it themselves without a caption, but it's much more likely that it's either reposted stolen art or AI art that wants to pass under the radar, or even stolen AI art (if, indeed, one can steal something that was created by a computer based on other stolen art to begin with).

Okay, rant over. Go forth and look a bit closer at images of embroidery in future! You'll either discover that it's an AI piece, or you'll get the pleasure of really taking in the cool details and techniques used by a skilled, real-life fiber artist!

#embroidery #AI embroidery #AI detection #applied statistics

[sportsallover]

Just saw the poster for Roland Garros this year, and I really really (really) really loved it until I read that it was created with the help of AI. Now I’m just mad. There are so many skilled artists out there. Why did they have to choose one who uses the aggregated labor of others instead their own honed skills??

#tennis #Roland Garros #RG24 #applied statistics #is my usual tag for AI because that’s all it is #an algorithm using statistics and stolen images to ‘create’ stuff

[proustianrevelry]

oh my god are u serious

article from 2017 abt Disney training an algorithm with Quora post data to judge the "quality" of a story based on, essentially, how many upvotes it would get on the site that makes Yahoo Answers look like an almanac.

2017, the same year as the live-action Beauty and the Beast, which dedicated multiple lines of exposition to fixing/explaining "plot holes" in the original such as: why doesn't Belle talk more about her dead mom? is EVERY object in the beast's castle enchanted? if Belle is such a strong female character, how come she doesn't she teach other girls in town to read books for fun #checkmatefeminists?

#disney #applied statistics #the other hellsite #obviously this isn't the only factor that makes the screenplay bad but jesus christ #even if they discontinued the experiment or gave it a different dataset #this still might've actually influenced the creative team's priorities #this is so stupid it's funny but also a dire portent for workers who want to make things in hollywood that don't suck

[sporemiette]

That Halloween blog you shared from is all ai generated trash in case you didnt know 💫

argh, thanks for letting me know ! i've deleted it :P

#anon #inbox #applied statistics #something about the layout of the room didnt make sense sdjfhgdjfhs #thanks again anon <3

[sweetthingshavesharpteeth]

This really struck home for her when she fed the program three photos from an enormous explosion and fire in Hemel Hempstead, UK. "It looked at it and thought it was great". She realized that if we start making these systems autonomous, such that they can take actions based on their training, "then in the pursuit of beauty, and things that it mistakenly thinks are good, it could kill people". The model has learned that colors in the sky are beautiful, but it does not understand mortality nor does it understand all of the things that are not generally shared online.

#ethics #machine learning #i suppose i should tag it #ai #although its definitely a misleading term #i rather like #applied statistics #as an alternative

[lilacandladybugs]

I'm in need of advice, reddit hasn't been helpful and I'm desperate so I've come to you Tumblr please help me

I'm currently a data scientist for a very small start up company, but I have my background in political science and so I'm concerned that I might be dead in the water if/when the company goes under and I need to find another job. I've consulted with some recruiters and they agree that if I want to go into data science I should get my master's (EDIT: they said I probably should get my degree in Statistics because the program is more widely known so I have a better chance of not getting turned away by HR who will have less knowledge about what a data science master's even is). I think because of my personality, data science is a really good job for me, so I'm planning on going for it.

Here's the issue: I don't want to go to school and end up learning exclusively theory. I've been teaching myself a ton by reading textbooks and I've noticed that while there's a lot of depth in the math/calculus/linear algebra behind how the functions work and what the parameters are, there seems to be very little information on how to actually apply that information in the real world.

Obviously the math is important and very exciting :D but if all I do is learn the math and I don't learn how to apply the knowledge I have to non-ideal data sets and situations then I'm not really learning the information I need to know.

Are there any graduate programs that are well known for really preparing people for data science roles in the workforce instead of just focusing on the academic side of statistics?

#statistics #data science #math #careers #jobs #grad school #what is the community tag for this idk #;---; #i have to sign up for classes for prerequisites right now so im nervous bc idk which classes to take if idk which schools i should apply to #reddit has not helped me </3

[bixels]

Cherry picked and cropped information. Misunderstanding of what the phrase "not striking over its use at all, mostly over how it is used" which you also proved with your evidence, and general anti intellectualism and fearmongering over technology rather than action. This is not an "Anti AI strike" it is a workers rights strike and while applied statistics makesup some of the talking points it is not the only thing OR the MAIN thing being discussed. You are the person who fears the gun and not the racist who holds it.

Dude, I literally screenshotted from SAG-AFTRA and WGA's objectives page on their website.

You're right, this is a worker's rights strike. And AI, under studio operation, poses a threat to worker's rights. A strike can be about more than one thing, buddy.

Also, what the fuck are you talking about.

#ask me #anon #also why do you keep calling it applied statistics

[greetings-inferiors]

Linear algebra is the most boring interesting subject so far. Like I can tell that once we get going it’ll be really cool and really interesting, but at the moment we’re still kinda going over the basics and it’s a bit boring and tedious. But I have such high hopes for it

#for reference: #applied is the most interesting boring subject #algebra and analysis are the most interesting interesting subjects (don’t make me choose) #statistics is the most boring boring subject #text post #randyposting #maths #math #statistics slander #mathematics #mathposting

[siflshonen]

Give peace a chance, love your neighbor as yourself, treat others as you would like to be treated, and stop showing me some of the worst takes ever about my blorbo as suggested content THANK YOU!

#AMEN #blorbo #tumblr #blorbo posting #the haters on this site have a right to hate on their own blogs in peace #so stop encouraging violence and throwing them in my face! #we may use the same tags sometimes but one of us uses their heart to read a story #and the other one sits there and crunches numbers given offhandedly in the story through extrapolated data #figures that are not important to the story mind you #and that’s what destroys their suspension of disbelief #I cannot do and do not care about math or applied statistics in my fiction! #I am very stupid and care about shit like dramatic irony and story setup #we are NOT the same

[stinkrascal]

do you also feel that female tav's are more "popular" here on tumblr? I mean I have like 8 tav's but the female one's get waaay more recongnition?

idk, i dont think about it when im posting my tavs/durge. my bg posts dont get many notes anyways bc im originally a simblr. so i just post whatever i want without caring about the note count, caring about the note count is the killer of fun. i just like making female tavs more bc i think girls are prettier 🫶🏻

#jade answers #anonymous #the only reason daveth is a man is bc when i was brainstorming durge #i was like well irl men are statistically more likely to commit violent crimes so he’d most likely be a man 😭 #which yeah ik that doesn’t apply to dnd but it made sense in my head when i made him okay

[transingthoseformers]

Holy shit if that isn't the most amazing way to find out there are aliens among you and just... chilling and living their lives. Recovering and healing from their own stuff. Sure, there are outliers who decide they want to fuck shit up, but that happens in humans too. Some help with stuff, some keep to themselves. Doing the whole life-ing thing. Helping advance human technology for the better of all.

After all, considering what I know about irl life on earth and what I've seen in various continuities, out of all we see Cybertronians and humans are some seriously social bastards. And at spark, the ability to work together is a winning strategy in nature.

The universe is so fucking big, can you imagine having friends to help explore it with you?

#transformers #tfp au #but also more than just the tfp au #also kinda applies to my pipe dream of a best case scenario for meeting aliens in real life. #statistically nothing is impossible— merely some degree of improbable. It's the infinite monkey theorem. #let me be wholesome about aliens; of possible futures i severely doubt I should even be hoping to see #let me dream past my dystopic life

[ice-block]

Nothing against the bloggers posting these specifically I don’t think it’s intentional but it is a little interesting when those “pick a number to get a random character who is now your SO/boss/roommate/etc” posts show up on my dash and after I pick mine I go through all the options and every single time it turns out that only 2 out of 30 are women

#something something I feel like that statistic that men feel like women take up most of the conversation when they speak more than like 30% #of the time applies here. where 1/4 or less women represented in a silly fandom poll feels like a even amount #te/rfs dni

[anotherpapercut]

in 1 hour I will have officially relinquished the $1,032.96 that the library took from me by firing me for handing in my 2 weeks notice and then telling me I could only have it if I waive all my legal rights o7

#just realized one of the things in this contract is also that i cant fight them about my last paycheck #and im pretty sure theyre either a half hour or a full hour short and im so willing to fight that its unbelievable lmao #im not signing your fucking gag order for 1k but i will fight you for $6 #one of the stipulations in this thing is that im never allowed to apply to work at the library again or id have to give the money back #i doubt im going to be applying to this hell hole any time soon but that just makes me mad. fuck you you ridiculous bitch #im one of the most qualified people in this entire city. die. i want the option of returning after these people are all dead #because someone is gonna kill this bitch thats not a threat its an observation on statistics #you treat enough people like this eventually theres gonna be a consequence

[ask-serendipity-sky]

Hi! Just wanted to share a hopefully-less-inflamtory viewpoint on the numbers theories. It bothers me absolutely nothing if any one person does or doesn't buy into the theories but for me, it's definitely on a scale. I'm personally more open to the blatant number examples rather than the ones that require extrapolation.

Example: Jimin's Live with the not current time on his watch.

Fairly low hurdle to jump to connect it to JKs birth time. Especially since it was right at the beginning of the Live and very easily could have been intentionally set.

Example: JK's post time being linked to JM's birth time.

This is more of a stretch for me whenever addition needs to be applied. Seems rather arbitrary when the decision is made to apply addition or not. Why sometimes the full date is included and other times it isn't.

As someone who is paid to generate data analytics for a worldwide organization, I can attest that the exact same data points can be used to draw exact opposite conclusions given any set of data. It's all about the justification in why specific methods are applied in each case.

In conclusion though, we are all just strangers on the internet in pursuit of elevating our fandom experience by having these discussions and absolutely no one should be criticized for what level they choose to partake. In other words: Don't Yuck someone else's Yum!

Hello,

Oh yes. I agree with you on the addition of numbers and dates.

There is a way to do it that is constant and simple.

I've been trying to get someone to explain the exact way the dates and times should be added but I haven't found it so far which is why I've been cautious with that.

From what I've observed in kjkkr tweets, they add the time only. But perhaps, I haven't seen enough.

Of course, adding an interpretation to the patterns is up to each person. But I wonder how you view this since these patterns have been observed since 2015. Can it still be said that we are drawing conclusions from something that has been happening for years? At what point can it be taken as a "fact"? When does a hypothesis become a theory in the scientific sense? It's when things can be proven but how would this be proven? Do we have to do statistical analysis? I'm curious.

But yes, I agree! These are observations and no one is forced to view them or agree with them.

Thank you for sharing your thoughts.

#lets apply statistics to jikook

[gayseyjones]

So far basically every argument I've seen trying to define how "ai art" isn't "real art" either doesn't make logical sense or quickly veers into copyright dickriding + "trad" definitions of art and I just think it would be a lot more productive to focus on the labor aspects. It's not like we have 2 massive entertainment unions on strike right now protesting partially against the use of AI to replace their careers. But that's probably not important so who cares

#💖 #and while we're at it i think we should start calling ai ''applied statistics'' like that one guy said bc it works a lot better IMO