- Science Is Drowning in AI Slop – The Atlantic
NeurIPS, one of the top AI conferences, has seen them double in five years. ICLR, the leading conference for deep learning, has also experienced an increase, and it appears to include a fair amount of slop: An LLM-detection start-up analyzed submissions for its upcoming meeting in Brazil and found more than 50 that included hallucinated citations. Most had not been caught during peer review.That might be because many of the peer reviews were themselves done by AI. Pangram Labs recently analyzed thousands of peer reviews that were submitted to ICLR, and found that more than half of them were written with help from an LLM, and about a fifth of them were wholly AI-generated. Across the academic sciences, paper authors have even started using tiny white fonts to embed secret messages to LLM reviewers. They urge the AIs to rave about the paper they’re reading, to describe it as “groundbreaking” and “transformative,” and to save them the trouble of a tough revision by suggesting only easy fixes.
- When two years of academic work vanished with a single click
After turning off ChatGPT’s ‘data consent’ option, Marcel Bucher lost the work behind grant applications, teaching materials and publication drafts. Here’s what happened next.….
If a single click can irrevocably delete years of work, ChatGPT cannot, in my opinion and on the basis of my experience, be considered completely safe for professional use. As a paying subscriber (€20 per month, or US$23), I assumed basic protective measures would be in place, including a warning about irreversible deletion, a recovery option, albeit time-limited, and backups or redundancy.
- llama.cpp/AGENTS.md at master · ggml-org/llama.cpp
This project does not accept pull requests that are fully or predominantly AI-generated. AI tools may be utilized solely in an assistive capacity.Read more: CONTRIBUTING.md
AI assistance is permissible only when the majority of the code is authored by a human contributor, with AI employed exclusively for corrections or to expand on verbose modifications that the contributor has already conceptualized (see examples below)
- (Part 1) “The Digital Battlefield of Truth” from Democracy, AI, and the Freedom to Learn: A Survival Curriculum for the 21st Century — AltLiberalArts
Lecture Description: This opening lecture introduces AI-Enhanced Reflexive Control (AIRC)—a framework explaining how AI scales and modernizes Soviet-era psychological warfare into a system of unprecedented manipulation. Drawing on data from the GCRD–Sensika Disinformation Observatory, including a detailed case study of Russia’s Pravda network targeting Bulgaria (December 2024–April 2025), participants explore how AIRC operates across multiple dimensions such as computational content generation tailored to cultural contexts, predictive audience modeling that identifies psychological vulnerabilities and dynamic narrative evolution that adapts to events while preserving strategic goals. The session reveals how these operations reshape the informational environment itself—producing second-order effects in which targets feel autonomous while being steered toward predetermined outcomes. - Flights of Fancy: The Nine Birds of Jacques de Fornazeris (1594) — The Public Domain Review
- What a simple question about My Cousin Vinny taught me about LLMs
Pretty good, right? When I first saw this generate, I thought it wasn’t that bad. Then I put it into my LLM-based fact-checker. (Results)It’s actually horrible.
First, the good. Answer one is a bit off on what their conversation is about. She’s not frustrated by the fact they are in Alabama as much as she needs Vinny to succeed and get his life on track (including winning his case) so they can take their life to the next stage. She’s been waiting for him to get his life together as he has failed the bar repeatedly. But this is minor, really.
For the ChatGPT thing, there is a big warning after the first click. “Are you sure? Files will be deleted.” (Or something similar) He ignored that, figuring he’d just opt out for a bit to test something before opting back in. It sounds like the system worked as designed – they actually deleted stuff when they said they would, rather than secretly retaining it?
That’s absolutely true. I haven’t tried it. It’s funny how much people expect technology to protect them from the thing they are trying to do on purpose.