#fuck the NLP

[nimthirielrinon]

hey fellow Aussies! I have a button and stickers available if you’re interested.

[goddesszenova]

Go check out my new audio experience

[spec-vp]

looking at phd programs

[lordsardine]

.

[haruspexism]

"put deep blue against a modern ai model and the ais will definitely fail so you can stick it to tech bros because tech bros love being smart and chess is the way to teach them that their product is shit xD" <- real opinion I had to see on tumblr dot com

you have ten minutes to explain to me what you think deep blue is and also what you think ai is

math being taught poorly in schools has resulted in the consequence of people on Tumblr not understanding how AI works beyond "thing that might steal my art" and this is the worst consequence of all because it is personally annoying to me,

[telperinquaar]

Hi, you recently reblogged a post about how Nanowrimo is not disallowing or disavowing AI because doing so is classist and ablist and in your tags suggested that you consider this "yikes."

Honestly, it kind of hurt my feelings as a person with a disability who occasionally uses AI as a disability aid. Let me explain.

I use AI frequently for word recall. I have ADHD- a lot of people do. Many people with ADHD, including myself, struggle with word recall. It can be extremely bad, but how bad it is day to day is variable, and not all people with ADHD struggle with this to the same degree. When my word recall is really bad, NLP's (Natural Language Processors) are practically tailor-made to find that exact word I need. As an example, I used AI to remind me just now about the term "Natural Language Processor," which, along with the term "machine learning," is frankly just a better description than AI for these tools. But I will continue using the term AI for convenience.

The fact that people do not imagine this sort of use in conjunction with AI IS a form of ablism. They immediately assume all use is infringing. If they actually talked to people with disabilities (who do use AI), they would discover these other uses, and perhaps the conversation could be elevated to a more constructive state instead of trying to make everyone who uses a tool feel "yikes" for using it. Many of us are extremely conscientious and well informed of the issues involved.

Consider that if someone has said something is ablist (or classist or any other -ist,) they just might have a point and that you should try to discover what that point is before assuming that it's fake. Don't take everything at face value, but don't dismiss it out of hand either. Listen to people who have differing opinions and try to get the nuances of the conversation.

When people think about AI being used in conjunction with writing and visual art, they only consider the egregious uses - which makes sense, as that is how AI is advertised; as a magic technology that solves ALL problems. But those egregious uses are not the only use of these tools. AI does not have to be a magic wand that replaces the creative process of an artist. I have given one example of such a use above, but I could list many more.

If your "yikes" is in regard to the ecological impact - I hope that the overzealous implementation of AI into everything takes the ecological facts into account and that is ammealorated, but please do not throw people with disabilities under the bus while trying to make buses less polluting.

//The only use of AI in this post was to help me remember the word Natural Language Processor - I know my tone is pretty formal and sometimes comes across as AI, but it's not.

Jeezly fucking crow, dude. It was a single-word comment. I hope you sent this to literally everyone else who commented in a similar manner.

I use AI frequently for word recall. I have ADHD- a lot of people do. Many people with ADHD, including myself, struggle with word recall. It can be extremely bad, but how bad it is day to day is variable, and not all people with ADHD struggle with this to the same degree. When my word recall is really bad, NLP's (Natural Language Processors) are practically tailor-made to find that exact word I need. As an example, I used AI to remind me just now about the term "Natural Language Processor," which, along with the term "machine learning," is frankly just a better description than AI for these tools. But I will continue using the term AI for convenience.

I also have ADHD. I also struggle with word recall. You know what I do? I google things. I use dictionary and thesaurus websites. I use OneLook, which suggests associated words, similar words, and similar concepts.

Not everyone who uses AI is stealing from artists, no, but it's well known that AI does scan people's art--almost always without their consent--to generate pieces. It's also been seen around places like AO3, scraping fics from unlocked accounts.

Personally, I dislike the implication that disabled (or poor--that's what "classist" means here) people are incapable of writing without an AI generating something for them. I've written 100k+ words on AO3, and all of them are mine. I've talked to friends, I've written parallel fics, I've rewritten my own stories, but those words are mine. I wrote them. A disabled person. To imply that I need AI to do that pisses me off.

And believe it or not, my primary dislike of AI isn't ableist or classsist or whatever. (I'm not even against all forms of AI! I understand that in some fields, analyitical AI is quite helpful--I've read that it's great at finding breast cancer, for example.) My primary beef with AI, especially generative AI like ChatGPT is the fact that:

It will just lie to you. It will just make up things. There are people who have used it in court cases (it didn't work), and there are people using it to write books--everything from cookbooks to mushroom identification guides. (Guess what amateurs need expert help with when they're starting out? You know, so they don't die?) It's also happened with animal care guides. AI doesn't need to be used in a generative context at all.

There is also a massive environmental impact that I rarely, if ever, see talked about.

[mezz-merizing]

language, hypnotism, and you!

here in the Hypnokink Zone we tend to center our focus on inductions. and for good reason!! they're the purest form of hypnosis, the intentional induction of someone, an endeavour designed exclusively to take someone into trance. but what if we didn't want to do an induction, or even take someone into trance, but still have a hypnotic effect on them? as it turns out, this is extremely doable, and extremely cool!

if you've been here for a while now you might have seen this post from me, my first proper ramble about all this stuff, and you might remember that bit at the end, when i talk about just how hot i find the words "hypnotized" and "brainwashed" themselves, and how i sometimes just find myself repeating them over and over when i'm really feelin' it. i think that's a good representation of how just single words can fuck with our minds, and it's not just those two <3

so like, let's think for a second about what language actually is; we tend to take for granted the fact that we make air do silly tricks with our mouths and our throats in order to put ideas and concepts into other people's heads. really, a word isn't just a word- it's not just the funny shape that air takes on when you make just the right tongue movements, it's the idea that word actually is. and ideas are extremely powerful!! when you think of the word "brainwashed" you don't just think of the individual letters, you think of helpless subjects, spinning spirals, and complete, devoted obedience. and that has more effects on you than you might realise :3

sometimes people in the hypno sphere call this "neuro-linguistic programming" or "NLP" but that's kind of a misnomer, (actual NLP is a silly and weird pseudoscience largely invented by "pickup artists" to sell courses, so, y'know, i try and avoid the association) i prefer to just use "hypnotic language" because it's just as descriptive! the basis of hypnotic language is this: when you say words, you incept ideas into someone's head. if you have the skill, time, and familiarity with the person you're talking to, you can use this to control their thoughts nearly as effectively as if they were in trance!

a lot of this relies on what words mean the most to the individual you're talking to. let's take me as an example!! the word docile makes me fucking weak. it conjures up images of blank-face, calm-smile obedience, of gently nodding and going about the commands i'm given, of empty-headed servitude. the images it conjures in my head are vivid and hot as hell, and it's just a single word. use it a little bti around me, and you'll ensure i have all those thoughts swirling around in my head! thoughts of servitude. thoughts of enslavement. and so it comes naturally that i'd be easier to control <3

tone matters too! if you speak to someone authoritatively, they'll come to see you as an authority. even in tiny matters, insignificant ones, even in little ways- saying "hey, grab me a glass of water" is more authoritative than "hey, can you get me some water?" obviously, this is a double-edged sword! too much authority and you might give your intentions away, or just come off as kinda bossy, and that's more likely to make people actively resist you than let you in. once again, it's all about knowing your target! you gotta know what your subject's tolerances are for this kinda thing

take all this together, and you can have almost as much of a grip on someone's mind as you would if they were completely hypnotized :3 you break them down overtime, get them hanging on certain words, widen their tolerance for authority... and eventually, with a lot of effort and patience, they're yours. obedient to you- brainwashed, in all but process

can you imagine it? like, from the subject's perspective- being completely under someone's hypnotic control without ever having been hypnotized. maybe looking back at a long time ago and thinking "i sure acted different then", but not worrying about it at all. you're a thrall, and as far as you're concerned, that's just who you are!! that's just kind of how the world works

i dunno about you but i can't imagine a fate i'm more desperate for <3

[collapsedsquid]

The holdout method does not exhibit statistical or adaptive overfitting. This is not an “always” statement. But a long list of work led by Becca Roelofs, Ludwig Schmidt, and Vaishaal Shankar has shown that it’s true. There isn’t overfitting on Imagenet, in the sense that the better the models on a test set, the better they are on new data sets. There isn’t overfitting in NLP question answering. There isn’t overfitting on Kaggle Leaderboards. If train and test are iid. (Or nearly iid as in MNIST), we do not witness overfitting. The holdout method works better than statistics suggest it should. All of the theorems we prove to justify the holdout method (usually doing some sort of union bound over possible tests) are laughably conservative about what happens in practice. You can even write theory papers digging into this conservatism (this one or that one). Let me be as clear: it’s possible for ML engineers to initially “fit the training set too quickly,” so the test error goes up. This certainly happens, and then they have to deploy tricks to “fit the training set more slowly” or whatever. Add weight decay or dropout or batch norm or whatever is trendy today. Go for it. I don’t even know the best ones anymore because you all are writing tens of thousands of machine learning papers every year. I’m not arguing against the art and skill of machine learning engineering.

You motherfucker is image generation the only application of machine learning you know of? One thing I've had to do in the past is to look for things in the genome, protein binding sites, RNA structure, whatever. if there aren't a ton of them to measure then it's dead easy to build a model that will learn the features based on totally irrelevant pieces of the genome, the genome is fucking huge and there are so many things you can incorrectly use as predictors.

[rottingbunnyheart]

TW ; DELUSIONS, PARANOIA!!!

screaming what if nothijg is real and im not eeal and everything is jusy fucked ip and fake and the sun isnt real and i dont ecenavtolwy feel what if people arw wtaching me all the tome and im acrually on a experiment and everyojes watching wvrruthing i do because im nto even real and jusyvsome made ip fucking thign oh ym gos theyre all laughing at me now i cant wll aidhwb sibcqnt do thalis ajshshd nonno no no noo nop noo noo nko nlp. lo no no

[akjzsd]

turned my voice into a cognitohazard for @idunno45 @mionko , and @blank0s nlp is so fucking based

[mexashepot]

🔥 linguistics/languages

Hopaa

I think that everyone who works in a field related to language has to have at least some very basic background in linguistics, I am not talking about a whole degree or anything, but at least take an intro course, for fuck's sake

For example, it is unthinkable to me that speech therapists diagnose bilingual children when analysing only one of their languages (anyone who knows a thing or two about bilingualism would never do such a terrible mistake)

Or that teachers of foreign languages don't have a basic understanding of linguistics. how can you teach language if you yourself can't grasp its logic and rules?

And my favourite example is the NLP field but I think I have already whined enough to you about it lol

Thank you!!!

[cleliacleliadatura]

One final thought, really wish tarot card readers wouldn't hand out cards saying "catostrophe" to readers/listeners....understanding Neurolinguistic Programming (NLP), even at a basic level, allows for any person to conclude that perhaps this may in fact invite catastrophe in your life.

Be wary, I personally do not align with dark forces, even if sometimes I truly understand why people would, usually after lifetimes of rejection, sorrow, unnursed hurts and a sense of poverty, of irritrievable loss.

These are the people many say Jesus can heal, but it often falls on deaf ears. Telling someone in this mindset Jesus is the answer becomes laughable. It'd be easier for them to dance with the demon they know, the one that was there when they felt no light was available, than to entertain the thought that a prophet who hung out with lepers and whores and died on a cross for their sins is coming to save them, as the invisible carrier and footprints in the sand. Is it only in retrospect people see the value in this poem?

Hard to come back from that, from true despair and illness, to hippity doo dah let's celebrate Jesus. No one wants the fear of the Lord instilled in them when they've experienced true fear, and not to their own fault. Regardless of what others will tell them, to further fuck them up inside.

Don't force your ideals on people, your words, let them choose. Forcing words, ideas and behaviors on people backfires. It adds fuel and oxygen to the fire.

Some people must dance in the fire to forge the steel within....

[magiskarp]

Okay, in saying the following, I recognize that this is going to sound pretty “out there”, especially if you’re someone who is younger on the internet.

And I will preface what I am about to say with some credentials: I am an Internet Old. I am also someone who is paid professionally to work in the horror genre from time to time. I am also someone who has studied psychology, as well as studying some NLP-type shit (that I do not find ethical in the slightest, as an aside, and I quickly noped out of)…

I also want to let you know upfront that what I will tell you is NOT to make you worry or feel afraid. I am sharing this info only so that you can “see through the curtain” as it were, and recognize media and stories like this for what it actually is.

The tl;dr is - please be very criticial of any online horror story that does not provide a sense of catharsis or release at the end, or at the very least a sense of closure or ending. Or anything that is about someone giving up control.

Why? Because creepypasta stories like this may seem merely surreal and/or hilarious to you, but shit like this is actually subliminally priming you to be more easily manipulated and controlled.

And a lot of these types of stories are absolutely aimed at teens and pre-teens, or college kids, fyi. There is also a really good video here by Ann Reardon about of stories being told over brightly coloured baking videos. Mostly though, these stories are about and aimed at WOMEN.

Basically, this story is linking a fear response (a sewer man! - that sounds bad and weird at first) with a violent action - or a taking away of rights/possessions (the coffee that gets poured in the sewer) that comes from someone with power (the soldier) who is “solving a problem” (thinking rationally - how would coffee dumped in the sewer fucking help anyone?!?!) where the victim (the woman) is shown how the solider is “right” to do what he did.

The subliminal lesson here is that the solider does what he wants to do, because “he knows better” than the civilian girl. This story basically ends on the action of “the soldier did the right thing” and there is no real catharsis or sense of ending. You know, the kind of message that implies “be scared, rely on those in power”. The idea that we should just comply.

And, as a thought exercise: the story would be so, so much different if it were a sweet little old lady that took the coffee.

The initial scare/weirdness factor could still be there, but it wouldn’t have that same ‘heaviness’ feeling to it, you know?

Please just think smartly about media, okay?

[lordsardine]

i feel like this trying to train a fucking NLP model

[ratchated]

*turns your ducks into humans*

[flyingpurplepeopleprogrammer]

Oh fuck in WW's repackaging thread I think I'm the serial numbers quote.

Shout out to Angst, from whom I stole it.

And if anyone cares this is and will remain an NLP hateblog

Academic Papers on NLP

From one of my reddit comments...

So the best introduction to NLP is this video by Irving Kirsch. Here are the slides.

But if you want to really research NLP, you should get into the original papers. This is great fun and relatively easy these days. If you have Zotero and the Scihub plugin for Zotero, you can plug in any DOI link and it'll come up. You may have additional papers that you want to pull from citations; I use Semantic Scholar to turn citations into DOI links that I can get Zotero to pull.

So here's my research on NLP (if you copy and paste the URLs into https://sci-hub.hkvisa.net/ they'll give you the PDFs):

Predicate Matching in NLP

The Eyes Don’t Have It: Lie Detection and Neuro-Linguistic Programming

Research findings on neurolinguistic programming: Nonsupportive data or an untestable theory?

Mental imagery as revealed by eye movements and spoken predicates: A test of neurolinguistic programming

Neuro-linguistic programming treatment for anxiety: Magic or myth?

Double hypnotic induction: An initial empirical test

Neurolinguistic programming: a systematic review of the effects on health outcomes

There's also the papers that Michael Heap did:

Neurolinguistic Programming - An Interim Verdict

Neurolinguistic Programming: What is the Evidence?

Neurolinguistic Programming: A British Perspective

The Validity of Some Early Claims of Neurolinguistic Programming

Also recommended is @ellaenchanting's overview and Wordweaver's discussion on what NLP repackaged.