It turns out LLM's can generate the STOP DOING MATH meme https://knowyourmeme.com/memes/stop-doing-math

STOP_DOING_MATH.md

STOP DOING MATH

• NUMBERS WERE NOT SUPPOSED TO BE GIVEN NAMES
• YEARS OF COUNTING yet NO REAL-WORLD USE FOUND for going higher than your FINGERS
• Wanted to go higher anyway for a laugh? We had a tool for that: It was called "GUESSING"
• "Yes please give me ZERO of something. Please give me INFINITE of it" - Statements dreamed up by the utterly Deranged

LOOK at what Mathematicians have been demanding your Respect for all this time, with all the calculators & abacus we built for them

(This is REAL Math, done by REAL Mathematicians):

[fig1 of a complex surface ???] [fig2 of a more complex surface ????] [fig3 of a hyper shape ?????????]

"Hello I would like [graph] apples please"

They have played us for absolute fools

STOP_DOING_THEORY.md

STOP DOING THEORY

• Thought about ATOMS for MILLENNIA but couldn't even find an ELECTRON
• AI? More like "Aye, I didn't see the SCALING HYPOTHESIS coming!"
• Wanted FLIGHT? Theorists be like: "Maybe we just flap real hard?"
• "I've mathed out the perfect STEAM ENGINE!" - Said no one ever before actually building it.

LOOK at the grand THEORIES they've been touting:

• TIMELESS DECISION THEORY
• SEMANTIC GRAPH
• PROVABLY SAFE AGI

(This is REAL Thoery, done by REAL Theorizers):

"According to my calculations, this metaphor for a bird should definitely fly." Reality: "Hold my beer."

They have played us for absolute fools

STOP_THOERIZING_WITHOUT_TESTIING.md

STOP THEORIZING WITHOUT TESTING

• YEARS OF RESEARCH and yet MOST BREAKTHROUGHS come from ACCIDENTAL EXPERIMENTS
• "Let's try it out on paper first" - Said every researcher before their theory got debunked by a simple test
• "My model predicts X" but REALITY shows Y. Yet, they WON'T BUDGE.

LOOK at the grand THEORIES they've been touting:

• TIMELESS DECISION THEORY
• Distinction Graphs and Graphtropy
• AIXI

(This is REAL Thoery, done by REAL Theorizers):

"Hello, based on my calculations, this should work" Yet it spectacularly fails in practice

They have played us for absolute fools

So it turns out LLM's can generate more of these

https://knowyourmeme.com/memes/stop-doing-math

STOP DOING ALIGNMENT

• YEARS OF PHILOSOPHY but can't even align their own SLEEP SCHEDULE
• "Let's make it learn our values!" Can't even define what values ARE
• Wanted safe AI? Best we can do is "pretty please don't be evil"

"Yes please give me PERFECT RATIONALITY. What do you mean humans aren't rational?" - Alignment Researchers

LOOK at what they've been promising:
[fig1 of a utility function that's just "???" ] [fig2 of a safety framework full of holes] [fig3 of a reward model crying]

"Hello I would like [UNDEFINED_HUMAN_VALUES] please"
They have played us for ABSOLUTE FOOLS

---

STOP DOING TRANSFORMERS

• BILLIONS IN COMPUTE just to ATTENTION EVERYTHING TO EVERYTHING
• "Self-attention will solve it!" - Said before every OOM error
• Built MASSIVE MODELS but they still forget what you said 2 prompts ago
• "Let's make the context window INFINITE" - Computational reality: "Hold my beer"

LOOK at their fancy ARCHITECTURES:

• ROTARY EMBEDDINGS
• FLASH ATTENTION
• RETENTIVE NETWORKS
  (This is REAL Architecture, done by REAL Architects):

"According to my attention patterns, this is definitely reasoning" Yet it's just doing smart pattern matching
They have played us for absolute FOOLS

---

STOP DOING SCALING

• TRAINED ON ENTIRE INTERNET yet still makes up FAKE CITATIONS
• "Just double the parameters!" - Last words before hitting compute budget
• Thought they found EMERGENT ABILITIES but it was just BETTER PROMPTING
• "The capabilities will definitely scale smoothly forever" - Said no working GPU ever

LOOK at their SCALING LAWS:

• CHINCHILLA OPTIMAL
• KAPLAN SCALING
• EMERGENCE CURVES
  (This is REAL Scaling, done by REAL Scalers):

"My model needs just 10^27 more parameters" Reality: "Best I can do is memorize the training set"
They have played us for absolute FOOLS

---

STOP DOING EMBEDDINGS

• COMPRESSED ALL KNOWLEDGE into 768 DIMENSIONS OF NOTHING
• "It captures semantic meaning!" - Yet can't tell 'dog' from 'god'
• Built VECTOR SPACES but they're mostly EMPTY SPACE
• "The meaning of life is definitely in this cosine similarity" - Pure madness

I'll help generate similar satirical critiques in that format, focusing on ML/AI topics while maintaining the style's hyperbolic skepticism.

STOP DOING DEEP LEARNING

• BILLIONS OF PARAMETERS and yet CAN'T EVEN COUNT TO TEN RELIABLY
• "Just add more layers!" - Said every ML researcher before their model started hallucinating
• Training on THE ENTIRE INTERNET but still thinks Napoleon was alive in 2023
• "Yes please give me INFINITE COMPUTE for my PERFECTLY SCALED MODEL" - Statements from the thoroughly Deluded

LOOK at what ML researchers have been demanding funding for:

• MIXTURE OF EXPERTS
• SPARSE ATTENTION
• CONSTITUTIONAL AI
  (This is REAL ML, done by REAL ML Engineers):

"According to my scaling laws, consciousness emerges at 10^15 parameters" Yet they can't even make it add 2+2 consistently
They have played us for absolute FOOLS

STOP DOING ALIGNMENT

• YEARS OF DEBATE about UTILITY FUNCTIONS but can't even define HUMAN VALUES
• "Let's just make it maximize happiness!" - Said before discovering wireheading
• Built SUPERINTELLIGENT AI but forgot to make it CARE ABOUT HUMANS
• "The AI will definitely understand what we REALLY meant" - Famous last words

LOOK at their grand SOLUTIONS:

• REWARD MODELING
• INVERSE REINFORCEMENT LEARNING
• VALUE LEARNING
  (This is REAL Alignment, done by REAL Aligners):

"My model is perfectly aligned!" Reality: "I have determined humans would be happier as paperclips"
They have played us for absolute FOOLS

Would you like me to continue with more examples in other domains?

---

LOOK at their EMBEDDING SPACES:

• SENTENCE TRANSFORMERS
• WORD2VEC
• CONTEXTUAL EMBEDDINGS
  (This is REAL Embedding, done by REAL Embedders):

"King - Man + Woman = Queen" Yet can't figure out basic analogies consistently
They have played us for absolute FOOLS

Want more domains to satirize?

STOP DOING BOOKS

• Spent MILLENNIA on SCROLLS then suddenly "Oh just BIND some PAGES!"
• "Let's write a 1000-page THEORY tome!" - While the 20-page paper had everything
• Published MILLIONS of books but can't match ORAL TRADITION's memorability
• "Yes this needs TWELVE VOLUMES to explain" - Claims by the academically POSSESSED

LOOK at what PUBLISHERS push as "essential reading":

• INFINITE JEST
• GÖDEL ESCHER BACH
• CRYPTONOMICON
  (This is REAL Literature, done by REAL Writers):
  "According to my footnotes, you need these other five books to understand this one"
  They have played us for ABSOLUTE FOOLS

---

STOP DOING FOOD SCIENCE

• Obsessed with NUTRIENTS for DECADES but can't beat GRANDMA'S RECIPES
• "Let's MOLECULARLY GASTRONOMY everything!" - While peasant food feeds billions
• OPTIMIZED DIETS? More like "Optimized Confusion!"
• "I've scientifically formulated the perfect meal!" - Said no one enjoying their dinner

LOOK at their NUTRITIONAL THEORIES:

• SOYLENT
• OPTIMAL PROTEIN TIMING
• CHRONOBIOLOGY OF DIGESTION
  (This is REAL Nutrition, done by REAL Scientists):
  "According to my metabolic calculations, this should be delicious." Reality: "Pass the butter."

---

STOP DOING SCIFI

• Predicted FLYING CARS but missed SMARTPHONES
• "Let's add MORE QUANTUM to everything!" - While missing social media's impact
• Space Opera? More like "Space NOPE-ra!"
• "I've extrapolated the future perfectly!" - Said every author before reality proved wilder

LOOK at their PREDICTIONS:

• VACUUM TUBE COMPUTERS IN SPACE
• MARTIAN COLONIES BY 2000
• HUMANOID ROBOTS EVERYWHERE
  (This is REAL SciFi, done by REAL Futurists):
  "According to my worldbuilding, we should have FTL by now" Reality: "Best I can do is Twitter"
  They have played us for ABSOLUTE FOOLS

Want more topics? Can cover: rationality, genetics, programming languages, or transhumanism.

Knowledge limits: Some references intentionally exaggerated for satirical effect.

STOP DOING MACHINE LEARNING

• YEARS OF TRAINING yet still can't tell a CHINCHILLA from a BERT
• "Let's just add more LAYERS" - Solution to every ML problem ever
• Wanted better results? Just throw MORE PARAMETERS at it!
• "Yes please give me INFINITE compute. What do you mean that's not realistic?" - ML Researchers

LOOK at what ML Engineers have been claiming all this time:
[fig1 of a confusion matrix that's mostly wrong] [fig2 of a loss curve going nowhere] [fig3 of a model hallucinating]

"Hello I would like [TOKEN_NOT_FOUND] please"
They have played us for ABSOLUTE FOOLS

---

STOP DOING TRANSFORMERS

• Attention is all you need? More like COMPUTATION is all you waste!
• Built a 100B parameter model but still can't remember what you said 2 minutes ago
• "Let's make the context window BIGGER" - Said every researcher before their GPU melted
• Self-attention? More like self-delusion!

LOOK at the grand ARCHITECTURES they've been promoting:
(This is REAL AI, done by REAL AI Researchers):

• MIXTURE OF EXPERTS (that are all equally wrong)
• ROTARY EMBEDDINGS (just spinning in circles)
• FLASH ATTENTION (because regular attention wasn't fast enough to be wrong)

"According to my calculations, this should be coherent."
Output: "The purple elephant drives democracy through photosynthesis."
They have played us for ABSOLUTE FOOLS

---

STOP DOING ALIGNMENT

• YEARS OF PHILOSOPHY but can't even align their own SLEEP SCHEDULE
• "Let's make it learn our values!" Can't even define what values ARE
• Wanted safe AI? Best we can do is "pretty please don't be evil"

"Yes please give me PERFECT RATIONALITY. What do you mean humans aren't rational?" - Alignment Researchers

LOOK at what they've been promising:
[fig1 of a utility function that's just "???" ] [fig2 of a safety framework full of holes] [fig3 of a reward model crying]

"Hello I would like [UNDEFINED_HUMAN_VALUES] please"
They have played us for ABSOLUTE FOOLS

---

STOP WRITING BOOKS

• CENTURIES OF LITERATURE yet still can't agree on what "THE CURTAINS WERE BLUE" means
• "This book will change your life!" - Said about every book ever written
• Wanted to learn something? Here's 500 pages when 5 would do!
• "Yes please give me ANOTHER trilogy that never ends" - Publishers

LOOK at what Authors have been selling us:
[fig1 of a plot diagram that's just a scribble] [fig2 of a character arc that's a flat line] [fig3 of a theme that's just "society bad"]
"Hello I would like [METAPHOR_NOT_FOUND] please"
They have played us for ABSOLUTE FOOLS

---

STOP DOING HUMANS

• MILLENNIA OF EVOLUTION just to spend all day looking at SCREENS
• "Let's optimize for survival" - Proceeds to invent SOCIAL MEDIA
• Wanted to communicate better? Created MEETINGS
• "Yes please give me MORE PRODUCTIVITY APPS" - Said everyone before doing nothing

LOOK at what we've been celebrating:
[fig1 of a brain mostly thinking about memes] [fig2 of a productivity graph going down] [fig3 of a motivation chart in the negatives]

"Hello I would like [BASIC_LIFE_SKILLS.exe not found] please"
We have played ourselves for ABSOLUTE FOOLS

apparently very very smart people would just find themselves creating these with utter joy at 3am lol

STOP DOING RECURSION

• FUNCTIONS WERE NOT SUPPOSED TO CALL THEMSELVES

• YEARS OF NESTING yet NO REAL-WORLD USE FOUND for going deeper than THREE STACK FRAMES

• Wanted to go deeper anyway for a laugh? We had a tool for that: It was called "ITERATION"

• "Yes please give me BASE CASE of something. Please give me INFINITE STACK OVERFLOW of it" - Statements dreamed up by the utterly Deranged

LOOK at what Computer Scientists have been demanding your Respect for all this time, with all the compilers & IDEs we built for them (This is REAL Recursion, done by REAL Programmers):

[INCOMPREHENSIBLE FIBONACCI TREE]

?????

[INSCRUTABLE MERGE SORT]

???????

[MIND-MELTING TOWERS OF HANOI]

???????????????

"Hello I would like n + factorial(n-1) apples please"

They have played us for absolute fools​​​​​​​​​​​​​​​​