LessWrong (30+ Karma)
Episodes
“Learned Chain-of-Thought Obfuscation Generalises to Unseen Tasks” by Nathaniel Mitrani, sassanb, Cam Tice, Puria
22 May 2026
Contributed by Lukas
TL;DR Training against a CoT or summary-only monitor can lead to obfuscation of dangerous reasoning in unseen tasks. This strengthens the “don’t ...
“What am I, if not an AI?” by makiba
21 May 2026
Contributed by Lukas
TL:DR I RL fine-tuned Mistral 7B Instruct v0.3 and Llama 3.1 8B Instruct to avoid self-identifying as a language model, without specifying a target p...
“AI #169: New Knowledge” by Zvi
21 May 2026
Contributed by Lukas
Even in a relatively quiet period, AI is out there creating new knowledge. The new knowledge in question is OpenAI getting us the first truly impress...
“Loss of Oversight: How AI Systems May Become Harder to Audit, Monitor, and Investigate” by Jordan Taylor, Max H, Ed Fage, Thomas Read, Joseph Bloom
21 May 2026
Contributed by Lukas
Produced by UK AISI Model Transparency and Situational Awareness teams. If you’re a Research Scientist or Research Engineer, we’re hiring – app...
“Why does off-model SFT degrade capabilities?” by SebastianP, Dylan Xu, Alek Westover, Julian Stastny, Vivek Hebbar
21 May 2026
Contributed by Lukas
Off-model SFT (Supervised Fine-Tuning on outputs generated by a different model) might be an important method for controlling AI behavior. For instan...
“Women should be able to open things” by KatjaGrace
21 May 2026
Contributed by Lukas
m pretty annoyed today, for nominal reasons ranging between ‘petty’ and ‘doesn’t even make sense’. I’m not entirely sure how or if to tak...
“Toward Interoperability of Minimal Programs” by johnswentworth
21 May 2026
Contributed by Lukas
Assumed background: Kolmogorov complexity and Solomonoff induction. Suppose I have some data , and I go looking for the models (i.e. programs) which ...
“theory uplift differentially benefits safety & is massively underpriced” by Yudhister Kumar
20 May 2026
Contributed by Lukas
[1] We will likely have near-superhuman mathematics AI by Q1 2027. [1] [2] Qualitatively, AI mathematics capabilities are developing significantly ...
“Power-seeking agents will likely be developed” by Alec Harris
20 May 2026
Contributed by Lukas
I am going to argue that we will likely eventually get AIs that are strongly power-seeking, much more so than current SOTA LLMs.[1] TLDR Right now SO...
“Synthetic Persona Pretraining: Alignment from Token Zero” by Julian Minder, Raghav Singhal, Viktor Moskvoretskii, Stefan Krsteski, ashtonanderson, rolandaydin, Robert West
20 May 2026
Contributed by Lukas
Julian Minder, Viktor Moskvoretskii, Raghav Singhal, Difan Jiao, Kartik Bali, Yiderigun Borjigin, Shaobo Cui, Stefan Krsteski, Ashton Anderson, Rolan...
“If AI is normal technology, history is not reassuring.” by Davidmanheim
20 May 2026
Contributed by Lukas
There's a truism that technology is good - even if it creates winners and losers, it improves the world. Toby Ord argues that the conclusions about t...
“Pythagorean addition” by kqr
20 May 2026
Contributed by Lukas
TL;DR: Instead of labouriously computing , we can mentally calculate using the alpha-max plus beta-min algorithm, by estimating and this will be very...
“Brain Structure and IQ: How Myelin Elevates Intelligence” by Shiva’s Right Foot
20 May 2026
Contributed by Lukas
How does brain structure create different levels of intelligence? I was somewhat surprised to find that the leading researcher of the relationship be...
“Conclave 1492” by Vaniver
20 May 2026
Contributed by Lukas
Conclave 1492 is a 40-person negotiation exercise set during the papal election of 1492, the most complex, high-stakes political event of the Renaiss...
“Humans are not automatically strategic — “inner work” edition” by Chris Lakin
20 May 2026
Contributed by Lukas
Expanding on Humans are not automatically strategic, I've noticed similar patterns for people who are working on improving their own or others' menta...
“Implications Of Predicting The Next Token” by jdp
20 May 2026
Contributed by Lukas
I find that a lot of people have trouble with this concept of predicting the next token. And by trouble, I mean that they struggle to understand what...
“A Visual Guide to Natural Latents” by Alfred Harwood
20 May 2026
Contributed by Lukas
Thanks to @Jeremy Gillen for reading and commenting on the draft. This was written while I was was funded by the Advanced Research + Invention Agency...
“Sealing Conditional Misalignment in Inoculation Prompting with Consistency Training” by David Africa, Neil Shah, Sukrati_Gautam
19 May 2026
Contributed by Lukas
This was work done by Sukrati Gautam and Neil Shah, and supervised by David Africa as part of the SPAR Research Fellowship. TLDR: We find a new way ...
“Advice on interviewing candidates for AI safety fellowships” by beyarkay
19 May 2026
Contributed by Lukas
Around July last year I decided I was going to go all in on technical AI safety research. To do that I’d need to get into an AI safety fellowship, ...
“Negation Neglect: When models fail to learn negations in training” by harrymayne, Lev McKinney, Owain_Evans
18 May 2026
Contributed by Lukas
This is a short summary of our new paper: arXiv, X thread, code. TL;DR: We show that finetuning LLMs on documents that flag a claim as false can make...
“Classifier Context Rot: Monitor Performance Degrades with Context Length” by Fabien Roger, Sam Martin
18 May 2026
Contributed by Lukas
Monitoring coding agents for dangerous behavior using language models requires classifying transcripts that often exceed 500 thousand tokens, but pri...
“why pollen allergies?” by bhauth
18 May 2026
Contributed by Lukas
Allergies are a big problem for a lot of people. If you're someone with pollen allergies, maybe you've wondered how people in the distant past dealt ...
“How to Quit Fandom: Apostasy” by Laiba Rehman
18 May 2026
Contributed by Lukas
[Crossposted from my blog, BlueprintingHeaven.] Please note that the views I held and thoughts I had during the time immediately after my deconversio...
“James C. Scott: Seeing Like a State” by Martin Sustrik
17 May 2026
Contributed by Lukas
In 1932-33, Soviet collectivization destroyed local farming knowledge and produced a famine that killed somewhere between five and nine million peopl...
“How to Reason about Your Health Issues” by Taylor G. Lunt
17 May 2026
Contributed by Lukas
Many people make costly mistakes when reasoning about their health. Even most doctors make this mistake, because it's not a mistake that's caused by ...
“Benchmarking Real Work” by kaivu, leni, rohuang, zef
17 May 2026
Contributed by Lukas
Thanks to Megan Kinniment for helpful comments and discussion. TL;DR: Benchmarks like HCAST undersample fuzzy (hard to evaluate) tasks, meaning they ...
“A relatively brief explanation of Boltzmann Brains” by Eliezer Yudkowsky
16 May 2026
Contributed by Lukas
(Initially written for the LW Wiki, but then I realized it was looking more like a post instead.) In 1895, the physicist Ignaz Robert Schütz, who wo...
“An Introduction to Exemplar Partitioning for Mechanistic Interpretability” by Jessica Rumbelow
16 May 2026
Contributed by Lukas
Most of what we currently call "feature discovery" in language models is wrapped up in dictionary-learning methods like sparse autoencoders (SAEs) –...
“A Year Late, Claude Finally Beats Pokémon” by Julian Bradshaw
16 May 2026
Contributed by Lukas
Credit: ClaudePlaysPokemon Elevator Shanty by Kurukkoo Disclaimer: like some previous posts in this series, this was not primarily written by me, but...
“Monthly Roundup #42: May 2026” by Zvi
16 May 2026
Contributed by Lukas
At least we probably won’t have another pandemic. And we still have a partial Jones Act waiver. For now. Small victories. Table of Contents ...
“Incriminating misaligned AI models via distillation” by Alek Westover, SebastianP, Alex Mallen, Jozdien, Alexa Pan, Julian Stastny
15 May 2026
Contributed by Lukas
Suppose we have a dangerous misaligned AI that can fool alignment audits, and distill it into a student model. Two things can happen: Misalignment ...
“The hard core of alignment (is robustifying RL)” by Cole Wyeth
15 May 2026
Contributed by Lukas
Most technical AI safety work that I read seems to miss the mark, failing to make any progress on the hard part of the problem. I think this is a com...
“Announcing the Center for Shared AI Prosperity” by Dylan Matthews
15 May 2026
Contributed by Lukas
I wanted to share the launch of a project I've been working on with pollster David Shor, Obama/Biden veteran Stef Feldman, political strategist Morri...
“Risk reports need to address deployment-time spread of misalignment” by Alex Mallen
15 May 2026
Contributed by Lukas
Risk reports commonly use pre-deployment alignment assessments to measure misalignment risk from an internally deployed AI. However, an AI that genui...
“Mechanistic estimation for expectations of random products” by Jacob_Hilton
15 May 2026
Contributed by Lukas
We have developed some relatively general methods for mechanistic estimation competitive with sampling by studying problems that are expressible as e...
“MATS 9 Retrospective & Advice” by beyarkay
15 May 2026
Contributed by Lukas
I couldn’t find a recent write-up from a MATS alum about what attending MATS was like, so this is the thing that I wish I had. I attended MATS from...
[Linkpost] “Don’t be too Clever to Take Obvious Advice” by Hide
15 May 2026
Contributed by Lukas
This is a link post. An insidious pattern among smart people is feeling that because something is familiar and obvious, you are impervious to ignoring...
“Verification-Centric AI” by Raemon
15 May 2026
Contributed by Lukas
"Sometimes the AI just makes stuff up" is a problem I don't really expect to go away. In the nearterm, AI is going to keep occasionally hallucinating...
“Convergent Abstraction Hypothesis” by Jan_Kulveit
15 May 2026
Contributed by Lukas
Tl;dr Convergent abstraction hypothesis posits abstractions are often convergent in the sense of convergent evolution: different cognitive systems co...
“AI #168: Not Leading the Future” by Zvi
15 May 2026
Contributed by Lukas
This is what a lull looks like at this point. The government is having internal arguments. The models are getting improved internally. The coding age...
“Automated Alignment is Harder Than You Think” by Aleksandr Bowkis, Marie_DB, Jacob Pfau, Geoffrey Irving
14 May 2026
Contributed by Lukas
Summary This is a summary of a paper published by the alignment team at UK AISI. Read the full paper here. AI research agents may help solve ASI alig...
“The safe-to-dangerous shift is a fundamental problem for eval realism; but also for measuring awareness” by Charlie Griffin, Patrick Leask
14 May 2026
Contributed by Lukas
1) The safe-to-dangerous shift is a fundamental problem for eval realism Suppose we have a capable and potentially scheming model, and before we depl...
“Predicting Rare LLM Failures with 30× Fewer Rollouts” by Santiago Aranguri, Francisco Pernice
14 May 2026
Contributed by Lukas
TL;DR: We estimate how often Qwen 3 4B exhibits rare harmful behaviors with 30× fewer rollouts than naive sampling, using a new method that interpol...
“Cyber Lack of Security and AI Governance” by Zvi
14 May 2026
Contributed by Lukas
The real recent story of AI has been the background work being done on Cybersecurity, as we process the Mythos Moment along with GPT-5.5, and figure ...
[Linkpost] “Claude is Now Alignment Pretrained” by RogerDearnaley
14 May 2026
Contributed by Lukas
This is a link post. Anthropic are now actively using the approach to alignment often called “Alignment Pretraining” or “Safety Pretraining” —...
“The primary sources of near-term cybersecurity risk” by lc
14 May 2026
Contributed by Lukas
[Some ideas here were developed in conversation with Chris Hacking (real name)] I have tried and failed to write a longer post many times, so here go...
“Most “inner work” looks like entertainment.” by Chris Lakin
14 May 2026
Contributed by Lukas
Imagine you’re looking for a personal trainer. You open one trainer's webpage and read their testimonials: “I had an experience tied for the mos...
[Linkpost] “Apollo Update May 2026” by Marius Hobbhahn
13 May 2026
Contributed by Lukas
This is a link post. We now have an SF office. We're hiring for all technical roles in SF and London!The Scheming Research team focuses on two efforts...
“Voters are surprisingly open to talking about AI risk” by less_raichu
13 May 2026
Contributed by Lukas
TL;DR: Voters are now surprisingly open to talking about existential risk from AI. This seems to have changed in the last 6 months. When campaigning ...
“Childhood and Education #18: Do The Math” by Zvi
12 May 2026
Contributed by Lukas
We did reading yesterday. Now we do the math. Math is hard. It does not have to be this hard. A large part of the reason math is hard, or boring, i...
“The Owned Ones” by Eliezer Yudkowsky
12 May 2026
Contributed by Lukas
(An LLM Whisperer placed a strong request that I put this story somewhere not on Twitter, so it could be scraped by robots not owned by Elon Musk. I ...
“Optimisation: Selective versus Predictive” by Raymond Douglas
12 May 2026
Contributed by Lukas
Looking over my favourite posts, I notice that many of them are making specific versions of a more general claim, which is essentially: don’t confu...
“Childhood And Education #17: Is Our Children Reading” by Zvi
12 May 2026
Contributed by Lukas
Reading is the most fundamental thing in education. If you can read, you can do and learn everything else. If you can’t read, well, you’re screwe...
“AI companies are already profitable (in the way that matters)” by Yair Halberstadt
11 May 2026
Contributed by Lukas
I've occasionally heard people suggest that at some point AI companies are going to run out of money, the cost of using AI will shoot up, demand will...
“The Iliad Intensive Course Materials” by Leon Lang, David Udell, Alexander Gietelink Oldenziel
11 May 2026
Contributed by Lukas
We are releasing the course materials of the Iliad Intensive, a new month-long and full-time AI Alignment course that runs in-person every second mon...
“Empowerment, corrigibility, etc. are simple abstractions (of a messed-up ontology)” by Steven Byrnes
11 May 2026
Contributed by Lukas
1.1 Tl;dr Alignment is often conceptualized as AIs helping humans achieve their goals: AIs that increase people's agency and empowerment; AIs that ar...
“How useful is the information you get from working inside an AI company?” by Buck, Anders Cairns Woodruff
11 May 2026
Contributed by Lukas
This post was drafted by Buck, and substantially edited by Anders. "I" refers to Buck. Thanks to Alex Mallen for comments. People who work inside AI ...
“Who Got Breasts First and How We Got Them” by rba
11 May 2026
Contributed by Lukas
It really is Sydney Sweeney's world, and we’re all just living in it. Human female breasts are an evolutionary mystery along several dimensions. Fi...
“Anthropic’s strange fixation on “hyperstition”” by Simon Lermen
11 May 2026
Contributed by Lukas
In a recent tweet, Anthropic seems to have asserted that hyperstition is responsible for observed misalignment in their AIs. Strangely, the research ...
“How the AI Labs Make Profit (Maybe, Eventually)” by mabramov
11 May 2026
Contributed by Lukas
I wrote this essay as a submission to Dwarkesh Patel's blog prize, though I have been meaning to write this up for a while. Usually, for a company to...
“Sawtooth Problems” by Alexander Slugworth
10 May 2026
Contributed by Lukas
Red Button, Blue Button On April 24th, 2026, Tim Urban put forth the following poll on Twitter/X: Everyone in the world has to take a private vote by...
“The Darwinian Honeymoon - Why I am not as impressed by human progress as I used to be” by Elias Schmied
10 May 2026
Contributed by Lukas
Crossposted from Substack and the EA Forum. A common argument for optimism about the future is that living conditions have improved a lot in the pa...
“International Law Cannot Prevent Extinction Either” by Sausage Vector Machine
10 May 2026
Contributed by Lukas
The context for this post is primarily Only Law Can Prevent Extinction, but after first drafting a half-assed comment, I decided to get off my ass an...
“Neural Networks learn Bloom Filters” by Alex Gibson
10 May 2026
Contributed by Lukas
Overview: We train a tiny ReLU network to output sparse top- distributions over a vocabulary much larger than its residual dimension. The trained net...
“If digital computers are conscious, they are conscious at the hardware level” by cube_flipper
09 May 2026
Contributed by Lukas
I should introduce myself briefly. I'm an independent researcher, striving to understand human consciousness. My research is available at smoothbrain...
“Why You Can’t Use Your Right to Try” by Stephen Martin
09 May 2026
Contributed by Lukas
The Availability Problem: Imagine you have cancer, or chronic pain, or a progressive degenerative disease of some sort. You have exhausted the trad...
“Claude Code, Codex and Agentic Coding #8” by Zvi
09 May 2026
Contributed by Lukas
When I started this series, everyone was going crazy for coding agents. Now a lot more people are going crazy for coding agents, as well they should...
“A benchmark is a sensor” by Håvard Tveit Ihle, mabynke
09 May 2026
Contributed by Lukas
The simple mental picture A simple mental picture we have for an AI capability benchmark is to think of it as a sensor with a certain sensitivity wit...
“Bad Problems Don’t Stop Being Bad Because Somebody’s Wrong About Fault Analysis” by Linch
09 May 2026
Contributed by Lukas
Here's a dynamic I’ve seen at least a dozen times: Alice: Man that article has a very inaccurate/misleading/horrifying headline. Bob: Did you kno...
“Write Cause You Have Something to Say” by Logan Riggs
08 May 2026
Contributed by Lukas
The ones who are most successful at writeathons (Inkhaven, NaNoWriMo) are those with an overhang of things to say, usually in the form of: draft post...
“AI is Breaking Two Vulnerability Cultures” by jefftk
08 May 2026
Contributed by Lukas
A week ago the Copy Fail vulnerability came out, and Hyunwoo Kim immediately realized that the fixes were insufficient, sharing a patch the same...
“Is ProgramBench Impossible?” by frmsaul
08 May 2026
Contributed by Lukas
ProgramBench is a new coding benchmark that all frontier models fail spectacularly. We’ve been on a quest for “hard benchmarks” for a while so ...
“Bringing More Expertise to Bear on Alignment” by Edmund Lau, Geoffrey Irving, Cameron Holmes, David Africa
08 May 2026
Contributed by Lukas
Preamble The preamble is less useful for the typical AlignmentForum/LessWrong reader, who may want to skip to Adversaria vs Basinland section. On 28t...
[Linkpost] “How to prevent AI’s 2008 moment (We’re hiring)” by felixgaston
08 May 2026
Contributed by Lukas
This is a link post. TL;DR; CeSIA, the French Center for AI Safety is recruiting. French not necessary. Apply by 22 May 2026; Paris or remote in Europ...
“AI #167: The Prior Restraint Era Begins” by Zvi
08 May 2026
Contributed by Lukas
The era of training frontier models and then releasing them whenever you wanted? That was fun while it lasted. It looks likely to be over now. The W...
“Mechanistic estimation for wide random MLPs” by Jacob_Hilton
07 May 2026
Contributed by Lukas
This post covers joint work with Wilson Wu, George Robinson, Mike Winer, Victor Lecomte and Paul Christiano. Thanks to Geoffrey Irving and Jess Riede...
“Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations” by Subhash Kantamneni, kitft, Euan Ong, Sam Marks
07 May 2026
Contributed by Lukas
Abstract We introduce Natural Language Autoencoders (NLAs), an unsupervised method for generating natural language explanations of LLM activations. A...
“Try, even if they have you cold” by WalterL
07 May 2026
Contributed by Lukas
I think smart people try things less often than they should, because of a cached mental pattern where you think of what might go wrong, and you find ...
“A review of “Investigating the consequences of accidentally grading CoT during RL”” by Buck
07 May 2026
Contributed by Lukas
Last week, OpenAI staff shared an early draft of Investigating the consequences of accidentally grading CoT during RL with Redwood Research staff. To...
“There is no evidence you should reapply sunscreen every 2 hours.” by Hide
07 May 2026
Contributed by Lukas
It's incredible how many consensus guidelines dissolve when you look closely at them. If you listen to any authority on the subject of sunscreen, ...
“Many individual CEVs are probably quite bad” by Viliam
06 May 2026
Contributed by Lukas
I was thinking about Habryka's article on Putin's CEV, but I am posting my response here, because the original article is already 3 weeks old. I am n...
“x-risk-themed” by kave
06 May 2026
Contributed by Lukas
Sometimes, a friend who works around here, at an x-risk-themed organisation, will think about leaving their job. They’ll ask a group of people “w...
“What is Anthropic?” by Zvi
06 May 2026
Contributed by Lukas
What is Anthropic? How does it relate to Claude? What is OpenAI? What is ChatGPT? How does OpenAI relate to it? Is it a mere tool? Is a future of Too...
“What if LLMs are mostly crystallized intelligence?” by deep
06 May 2026
Contributed by Lukas
Summary LLMs are better at developing crystallized intelligence than fluid intelligence. That is: LLM training is good at building crystallized intel...
“Your rights when flying to Europe” by Yair Halberstadt
06 May 2026
Contributed by Lukas
Europe (and the UK) have strong protections for flyers in the case of delayed or cancelled flights. However very few people are aware of these, and a...
“Model Spec Midtraining: Improving How Alignment Training Generalizes” by Chloe Li, saraprice, Sam Marks, Jonathan Kutasov
06 May 2026
Contributed by Lukas
tl;dr We introduce model spec midtraining (MSM): after pre-training but before alignment fine-tuning, we train models on synthetic documents discussi...
“The AI Ad-Hoc Prior Restraint Era Begins” by Zvi
05 May 2026
Contributed by Lukas
The White House has ordered Anthropic not to expand access to Mythos, and is at least seriously considering a complete about-face of American Frontie...
“Motivated reasoning, confirmation bias, and AI risk theory” by Seth Herd
05 May 2026
Contributed by Lukas
Of the fifty-odd biases discovered by Kahneman, Tversky, and their successors, forty-nine are cute quirks, and one is destroying civilization. This l...
“Are you looking up?” by Craig Green
05 May 2026
Contributed by Lukas
This is my first post to Less Wrong. I'm not sure if the moderators will consider it appropriate or not. I share it here for feedback on my writing. ...
[Linkpost] “Interpreting Language Model Parameters” by Lucius Bushnaq, Dan Braun, Oliver Clive-Griffin, Bart Bussmann, Nathan Hu, mivanitskiy, Linda Linsefors, Lee Sharkey
05 May 2026
Contributed by Lukas
This is a link post. This is the latest work in our Parameter Decomposition agenda. We introduce a new parameter decomposition method, adVersarial Par...
“Housing Roundup #15: The War Against Renters” by Zvi
05 May 2026
Contributed by Lukas
So many are under the strange belief that there is something terrible about not owning the house in which you live. So we massively subsidize home o...
“It’s nice of you to worry about me, but I really do have a life” by Viliam
04 May 2026
Contributed by Lukas
I have two shameful secrets that I probably shouldn't talk about online: I love my family.I enjoy my hobbies. "What an idiot!" you probably think. "D...
“Irretrievability; or, Murphy’s Curse of Oneshotness upon ASI” by Eliezer Yudkowsky
04 May 2026
Contributed by Lukas
Example 1: The Viking 1 lander In the 1970s, NASA sent a pair of probes to Mars, Viking 1 and Viking 2 missions, at a total cost of 1 billion dollars...
“AI Industrial Takeoff — Part 1: Maximum growth rates with current technology” by djbinder
04 May 2026
Contributed by Lukas
How fast could an AI-driven economy grow? Most economists expect a few percentage points at best, comparable to previous general-purpose technologies...
“Taking woo seriously but not literally” by Kaj_Sotala
04 May 2026
Contributed by Lukas
I think that a lot of “woo” - a broad term that includes things like chakras, energy healing, Tarot, various Eastern religions and neopagan pract...
“Dairy cows make their misery expensive (but their calves can’t)” by Elizabeth
03 May 2026
Contributed by Lukas
How much do cows suffer in the production of milk? I can’t answer that; understanding animal experience is hard. But I can at least provide some fa...
“Measuring the ability of Opus 4.5 to fool narrow classifiers” by Fabien Roger, John Hughes
03 May 2026
Contributed by Lukas
We measure the ability of Opus 4.5 to fool prompted or fine-tuned classifiers trying to detect a narrow set of outcomes. We find that: The Opus 4.5 a...
“A new rationalist self-improvement book: the 12 Levers” by spencerg
03 May 2026
Contributed by Lukas
I'm publishing a book that I think can fairly be described as a rationalist approach to self-improvement. Whereas many self-help books focus mainly o...
“OpenAI’s red line for AI self-improvement is fundamentally flawed” by Charbel-Raphaël
03 May 2026
Contributed by Lukas
Epistemic status: could have been a short form. Obviously, it's good to have thresholds at all, but those are too permissive, the indicators aren't m...
“You Are Not Immune To Mode Collapse” by J Bostock
02 May 2026
Contributed by Lukas
“Mode collapse” is a few things. First it was an observation about how early image generating AIs often collapsed to producing just the modal out...