Think Too Much @green5quirrel - Tumblr Blog

once you recognise the ubiquitous and inevitable fandom life cycle it becomes much easier to free yourself from it and just keep enjoying things in a more healthy way while still thinking critically about them

deactivated-blog

TONIGHT. WE’RE DRINKING FROM THE POND.

bodyfeeling

where did op go?

kineticpenguin

green5quirrel

crybaby-writings

i think in 2026 we need to analyze why so many hearing people online (and offline) so often treat signed languages as a trend/a novelty to be marveled or laughed at rather than.... a real set of languages that actual people use to navigate their lives.

patrochilles-or-bust

Shout out to all the Black ppl that can no longer participate directly in the fandom they love because of the stresses of racism 👍🏾 you contain multitudes of value and I'm sorry that the color of your skin and the power of your voice makes people not want to acknowledge that.

patrochilles-or-bust

Yes, nonblack people can reblog. I'd appreciate it, in fact, if y'all took the time to vocally support your Black friends/fans in fandom.

immortalbeingz

blood-orange-juice-deactivated2

butt types

these brownies seem a little strong

orthoscheme

well if i have any complaints i can just call potassium about it but im too meek so i wont

#Can't get over that it's only 5 g of fiber #Guess that's staying with you forever then

genderkoolaid

so many people on this website seem to believe it's physically impossible for a person to use ai in any way that involves any degree of effort or critical thought. like in that post on ai checkers someone in the notes said "if you think someone used ai for an essay, just check their sources and see if they are made up!" okay what if they used ai to turn a bunch of general notes and outline of a paper into a formal essay. what if they used ai to come up with the topic and outline but researched and wrote the final essay themself. what if someone does more than the bare minimum of just slapping in a prompt and then immediately using the first result they get with 0 fact checking or edits. what if, and i know this is gonna seem crazy, not every single person who uses an LLM is whatever cartoonish version of Evil Idiot Who Hates Using Their Brain For A Single Thing Ever you have made up in your head. perhaps this topic is sliiiiiightly more complicated and nuanced than that! just maybe!

therapatical

this is further complicated by the way companies are slapping the "AI" label onto various things that would've never fit under it to begin with. you could consider anti-aliasing in art programs a type of "AI" if you squint hard enough.

people do not understand how machine learning works, and I really, sincerely, suggest even just looking up a basic youtube video about intro topics (k-clustering, training data, tbh if you type in "how does machine learning work" into a search bar it'll probably give you something semi-decent). the whole thing with genAI is that its a predictive engine. it has its roots in linear regression. as in, you've got data on a scatterplot, you draw a line through it. that's your prediction line. if you then get a value, you can guesstimate where it would fall on that line. that's what it's doing just in a bajillion dimensional object with a shit ton of variables instead of just x and y. that's also why it's mostly unreadable to humans because no human being understands what the weights actually mean or do because it's incomprehensible (for the most part; I'm simplifying)

that is to say: people are acting like this shit is universal, and you'll always be able to "tell", etc. when it depends entirely on the data you fed into it but also what you're trying to predict in the first place. if you ask a large language model "who is the current president of the united states" and it was trained on data from 2009, it's gonna say Obama. if your training data was spread across several years, it's going to weight each of the presidents in those years and select one at quasi-random based on probabilities set by those weights.

how accurate any given model is depends on how well of a match your training data is vs your prompt. if you have an extremely common question with a fairly simple answer, you're likely to have a lot of training data that has the correct answer in it, so it has a high chance of spitting out something correct. if you have an extremely common question with a more complicated answer, you'll get a probability that reflects that training data. if you have an extremely uncommon question, because of the whole "law of large numbers" situation (aka why we use large sample sizes; you get more "extreme" results with smaller sample sizes), you're more likely to get stranger results. because it's just. less common in the training data. it's not trained on that.

that's why chatgpt can't do chemistry very well, but meanwhile there are extremely accurate machine learning models meant to predict how pathogenic a genetic mutation might be based on what is basically a physics engine mixed with some machine learning.

this is also why the whole "em dashes are a sign of AI writing" thing is ridiculous. em dashes are common in human writing. that's why they're common in AI writing, too. because they scraped fucking everything and woah what do you know, a lot of writing has em dashes in it. it's just a reflection of what you put into it.

it's still a frustratingly "we would be better off just building software and other programs to do this same job, why are we throwing the equivalent of brute force hacking at otherwise simple tasks" sorta issue in more cases than not, but. I have biases. and thoughts, clearly. lol.

therapatical

hi, I'm about to finish my bachelor's in data science, so I promise I do know what I'm talking about.

this is not necessarily true. "artificial intelligence" is a subset of machine learning that mostly handles natural language processing, whereas machine learning refers to a lot of different things that can honestly go as vast as a linear regression model if you reaaaaally wanna stretch that definition. people like to use them interchangeably when talking about this subject because the AI techbros desperately want a veneer of legitimacy and have been attempting to ride the coattails of machine learning for a while now.

all AI is machine learning. not all machine learning is AI. you've got the subsets reversed.

kitkatsnow

I'm a random person on the internet and you don't have much reason to believe me, so I will mostly focus on giving resources to learn more. In particular, I find the first chapter of Artificial Intelligence, A Modern Approach to be pretty accessible and a really good overview of AI as a field. I will say that I've been doing Data Science professionally since before the wide-availability of LLMs, and I don't consider being almost done with a Bachelor's in DS something that'd make you an expert in AI, especially given how varied DS programs are (some are more stats-heavy, some are Computer Science-heavy, and some focus more on SME). Techbros suck and I spend a lot of time dealing with the effects of them selling BS, and also I don't see any benefit to being wrong about what AI is. Please ignore any formatting issues- I'm doing this on mobile.

Here are resources from reputable institutions/professors to explain why ML is a subset of AI:

What AI is: Academic Definition of AI: "[T]he rational-agent approach to AI has prevailed throughout most of the field’s history. In the early decades, rational agents were built on logical foundations and formed definite plans to achieve specific goals. Later, methods based on probability theory and machine learning allowed the creation of agents that could make decisions under uncertainty to attain the best expected outcome. In a nutshell, AI has focused on the study◮ and construction of agents that do the right thing. What counts as the right thing is definedDo the right thing by the objective that we provide to the agent. This general paradigm is so pervasive that we might call it the standard model." Source: Artificial Intelligence a Modern Approach 4th ed pg 4

Second Academic Definition: "Artificial Intelligence (AI) is a term coined in 1955 by John McCarthy, Stanford's first faculty member in AI, who described it as "the science and engineering of making intelligent machines." Today it is a broad term for computer systems that can perform tasks with human-like intelligence, such as understanding language, recognizing images, learning from data, reasoning, and making decisions. Modern AI often works by finding patterns in large amounts of data and using those patterns to generate predictions or responses. It can be narrow (good at a specific task) or more general-purpose, like today’s large language models that can handle many tasks." Source: https://hai.stanford.edu/ai-definitions/what-is-artificial-intelligence-ai

Practical Definition of AI: "artificial intelligence, AI, branch of computer science devoted to developing data processing systems that perform functions normally associated with human intelligence, such as reasoning and learning " Source: https://electropedia.org/iev/iev.nsf/display?openform&ievref=171-09-16

Second Practical Definition: "A machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments." Source: https://csrc.nist.gov/glossary/term/artificial_intelligence

What ML is: (included for others who see this post)

Definition of ML: "Machine Learning is a branch of artificial intelligence that enables computers to learn patterns and make decisions from data without being explicitly programmed with rules. Instead of following step-by-step instructions, ML algorithms analyze examples to identify patterns and improve their performance over time through experience. Common applications include email spam filters, recommendation systems, image recognition, voice assistants, and fraud detection." Source: https://hai.stanford.edu/ai-definitions/what-is-machine-learning

Specific resources on ML being AI:

Article stating that ML is a subset of AI: "Machine learning is the subset of artificial intelligence (AI) focused on algorithms that can “learn” the patterns of training data and, subsequently, make accurate inferences about new data." Source: https://www.ibm.com/think/topics/machine-learning

Footnote clarifying the relationship between ML and AI: "In the public eye, there is sometimes confusion between the terms “artificial intelligence” and “machine learning.” Machine learning is a subfield of AI that studies the ability to improve performance based on experience. Some AI systems use machine learning methods to achieve competence, but some do not" Source: Artificial Intelligence a Modern Approach 4th ed. page 1

#Thread about AI that isn't just AI BAD but more nuanced #Language evolution and clarification about Machine Learning (ML) and Artificial Intelligence (AI)#I personally hope language surrounding this topic evolves into something more defining of what the AI tool DOES #And I also hope what AI used to mean (a romanticized version of an electronically made organism who veers unsettlingly enigmatically human)#gets a new term to satiate all the robot lovers out there who want freedom for our fictional sentient computer (lovers)#Maybe Artificial Sentience?#Would Isaac Asimov cringe too much at that?

itmeansalot

green5quirrel

#Mrr hrr hrr #Mmyesss!

the angel staying over at my house asked for a nightlight in their room and i told them buddy, don't you produce your own light? what're you gonna do with more? and they said they wanted to see why people like it so much. and also that the nightlight i own is blue and they're been trying to understand color. anyways i think they've stared at it for an hour now

pigswithwings

the angel staying over at my house said their eyes hurt from staring, so i took them to the store to get some eye drops. well midway through the medicine aisle i turned around to see them flashing in and out of colors like a combusting star, so i asked them what was up without looking directly at them. they told me that there were too many people at the store and they forgot how they looked. which. is understandable to be honest they've only been on earth for a little over 24 hours. my phone started to melt a little when i showed them pictures of themself, but my hand feels fine even though they held it while calming down. they're currently resting and i am not sure what to do with 5 empty bottles of eye drops

pigswithwings

the angel staying over at my house has been eating my food for the past few days and i don't really know if they can taste it because the food sort of bursts into ash and fire before it reaches their mouth? but they seem to enjoy eating it or at least watching me cook. they tried to help me fry some onions earlier but the stove flame turned from blue to white and the onions started crackling instead of sizzling so they backed off. however they have the ability to chop garlic like a pro

pigswithwings

the angel staying over at my house is taking daily walks with me around the neighborhood and i don't think they know about animals yet. we were going past someone's house and this labrador started barking at us. normal, right? but the angel staying over at my house asked me if that was music. and i had to stop and think about that one for a second. in the end i said it can be music to us, it's just that the dog doesn't know it (and may never know it, which is a real tragedy). then they asked me what's a dog

pigswithwings

the angel staying over at my house went to the library with me and they wanted to know if humans actually dream or if that was something people made up for television. guess what the staff told us that the full name of the angel staying over at my house couldn't be printed on a library card so i just said well couldn't we have one anyways? and now we do but it's really like the first third of their name and not written correctly. oh well. i have to go help carry some huge books about neuroscience into the house

pigswithwings

today the angel staying over at my house came up to me and started crying tears that melted through my carpet. told me they realised they wouldn't be able to use their library card after they left earth and i. i honestly forgot that was going to happen. but i told them hey dude that's just how humans feel too. can't read all the books, but the fact that you got to read some is worth bragging about, right? they could tell all of their angel friends about the cool stuff they've done. and they said that they used to think that the world was so small and that humans were being petty about missing earth but now they live here and they feel small too. and they're not sure what to tell their angel friends.

so i asked to hold their hand again. they didn't burn me this time, either. and i asked them if this moment felt small to them. they said it didn't.

anyways we made popcorn and watched it's a wonderful life. they told me that angels don't speak like that at all by the way

#cool story #but so much property damage

regexkind

Waking up in the middle of a colonoscopy to discover that the gastroenterologist is industriously constructing a model ship inside of my large intestine

#jealous

charl0ttan

blindbeta

Hi! I hope this isn’t redundant, but I have a question about choosing between alt text and using an image description in the main text.

So, in your opinion, is it worse to use alt text? For reference, I have a fanart blog. I’ve seen it mentioned that people who don’t use a screen reader might have difficulty viewing the alt text, or that it glitchs out sometimes. I don’t like the look of putting the ID below the image, but I’d rather not sacrifice accessibility for the sake of aesthetic.

The reasons I previously said I would advocate for image descriptions in the post over ALT text were:

1) at the time, ALT text could only be picked up through screen-readers

2) ALT text usually imposed a word limit

and 3) higher visibility for those who need it and increased awareness of accessibility needs

Since then, ALT text has been updated on tumblr, twitter, and instagram, although I’m not sure if that applies to other sites. ALT text is now visible through a box announcing the picture has an ALT text image description. It is more visible now because of this marker and because the text is literally visible when you click on the box.

This means it is more visible for those who don’t use screen-readers. It is also visible for raising awareness of accessibility needs.

While alt-text has been improved a bit, I still suggest using image descriptions in the post itself. This increases visibility and reduces glitching issues. If the post doesn’t load, the image description is still available.

Even if you don’t like the look of descriptions in the post, it is still better than some people not being able to enjoy your work and feel included.

angelonasher

I have also seen people say that they prefer alt text because they have no way to know if an external image description is just a caption or other text. I personally tend to at least put "ID below" as the alt text for an image to try and fix that problem.

blindbeta

One way to address this to use labels such as “Image description.”

Example:

[Image description: A screenshot of Witch Hat Atelier anime featuring main character Coco. She is smiling, her short blonde hair blowing back in wind and magic. End image description.]

As you can see, the markers let people know where the description is and where it ends. I did include alt text so screen-readers don’t simply say ‘image’.

I hope this helps.

wobblebees

Being a notorious robotlover in the current state of "AI" is the worst. I love robots discovering how they process their feelings in fiction, but when a chatbot pretends it knows how to love it feels like an insult to the craft

rosalesbeausderholle

One thing that worries me about the use of AI is whether or not it can worsen people's dementia and alzheimer's in the future. When my grandmother was first diagnosed, we got her math activity books. Now, my grandmother never had a formal education, but we did our best to keep her sharp, get her to do math and writing activity books, sudokus, playing board games that required some level of strategizing with her. Her family is prone to alzheimer's and dementia (both her siblings had it and deteriorated very very very quickly, which yeah, scares the shit out of me being her granddaughter) but she was the one whose mind lasted the longest, she only passed away two years ago, at 88, ten whole years after her initial diagnosis and sure, she had forgotten things, recipes and where she put her glasses and appointments, but she never forgot any of us, ten whole years in, she still remembered us. Now, this may have been luck, but doctors always said the constant mental work + companionship + medicine helped her a lot. So I'm thinking, these people who are now relying on AI for everything, from email-writing to thinking what's for dinner to casual conversations, I've even seen people rely on it to calculate what time they should leave their house if they need to be at a place at a specific time and their commute lasts X number of minutes. As if that's not... the simplest math operation possible? You shouldn't even need a calculator for that!!! Idk I don't know how long it'll take us to see the effects of this + exposure to brain-rotting short form content that is completely meaningless + people addicted to right-wing conspiracy style media. Idk I'm very worried. Please, read, read complicated books! Take up a book on philosophy and try to decipher it and make your own opinions on it, please buy a maths activity book and relearn how to do math, please get a hobby that involves lots of thinking and concentrating. PLEASE!!!

notalotgoingonatthisinstant

As a neurologist, I’ll give you the pretty name for it: cognitive reserve.

The way I explain it to my patients is that our neurons don’t regenerate. They make connections with each other and that’s it. If you don’t use your brain, they make fewer connections and, if one of them dies, you’re gonna miss it, because that was the only one that knew how to do X. Now, if each one of them has many, many connections, you won’t notice the difference when one of them dies. The others pick up the slack.

As of 2024, 45% of dementia risk factors are modifiable. Relevant to this conversation, 5% for less education and 5% for social isolation.

We absolutely are going to see the reflection of this, but it’s gonna take decades and it’ll be too late. So, for the love of your brain, pretend that it’s a muscle and make it work. People complain about “when am I ever gonna use this maths formula in my life?” You’re not. You’re teaching your brain to think logically. Those sinapses will be there for when you need to figure out your week’s schedule. English classes taught me how to interpret data and how to convey it in this text so it’s clear and you understand what I’m saying, not because I needed to justify why the curtain is blue.

Make your brain know how to do different things. Logic games, puzzles, taking care of a garden even if small, planning a church’s event or birthday, learn a new instrument, learn a few words in another language, look at a calendar every day, do some manual labor if possible. Do not, I repeat, do not let your brain get rid of sinapses by letting AI do everything. Your brain uses 20% of your body’s energy — do you really think it’s going to maintain connexions that aren’t in use?

Most cases of Alzheimer’s are sporadic, meaning no family history. Family history of a first-degree relative with Alzheimer’s starting before they were 80yo increases your risk in 2-3x on average.

TLDR: Yes. From the knowledge we have today, AI will increase the number and severity of dementia cases.

sexy-people-contests-2026

Let’s make tea

Tea

Water

Sugar

Honey

Milk/Cream

Lavender

Edible Glitter

Voting ended onApr 27

Do it for him

littlescratches

Thank you for these tags @asimovsideburns

I nearly died laughing while reading them and then again while reading them out loud to the spouse.

Please, someone choose water!

sexy-people-contests-2026

I’d like to point out that last year it wasn’t a liquid and I still ate it so if no one else adds water I’d still eat the final product

sexy-people-contests-2026

Oh yeah just so everyone knows the lavender is a concentrated syrup, it’s very sweet but yeah it’s in syrup form cause that’s what I have on hand

mysteryteacup

Guys. Please. We're at less than 30% liquid. This is a tea flavoured snack bar. Please.

sexy-people-contests-2026

Don’t listen to them, if water is anywhere near winning this is gonna be way too easy, I want a gelatinous glitter blob to chew on

the-orion-scribe

@chronicangel

Trending Blogs

Recently Viewed Blogs

Think Too Much