AI Horror Stories

Articles and papers regarding AI limitations.

Serious

The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity, Apple, June 2025: Testing LLMs/LRMs with puzzles. The models fail to perform beyond certain complexity threshold.
See also commentary by Gary Marcus
Diabolus Ex Machina, Amanda Guinzburg, June 2025: Full transcript of ChatGPT lying.
Bubble Trouble, Bryan McMahon, March 2025: AI bubble is about to burst.
Breaking: OpenAI's efforts at pure scaling have hit a wall, Gary Marcus, February 2025: LLM scaling have hit the wall.

How to keep sea-monkeys alive, Mike Monteiro, June 2025: AI is sea monkeys.
Irony too funny for words, Gary Marcus, February 2025: AI company asking people not to use AI.

An open letter to third-party suppliers, Patrick Opet, 2025: Open letter from JPMorganChase CISO, criticising current approach to software and cybersecurity, including blind use of AI.
Central Brain of Humanity, Radovan Semančík, June 2025: My blog post complaining about use of AI and inflated expectations.
IAM Myth: AI Will Fix It, Radovan Semančík, March 2024: IAM myth, describing inlated expectations of AI capabilities in identity and access management.