Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models...

Researchers show that turning harmful prompts into poetry lets many large language models ignore safety rules. Across 25 models, poetic prompts raised jailbreak success rates far above non-poetic versions. The study suggests current safety methods can be defeated by simple stylistic changes.

ORIGINAL LINK: https://arxiv.org/abs/2511.15304

Proportion of new articles on the internet generated by AI...

A study found over half of web articles are now AI-generated, mostly routine pieces like listicles and how-tos. This shift has displaced many freelance writing jobs and blurred the line between human and machine authorship. Original, human voice and creativity may become more valuable as people learn to collaborate with AI.

ORIGINAL LINK: https://theconversation.com/more-than-half-of-new-articles-on-the-internet-are-now-being-written-by-ai-is-human-writing-headed-for-extinction-268354

Against "Brain Damage"...

AI doesn't literally damage your brain, but unthinking use can weaken your thinking. Used well—with prompts, guidance, and doing your own work first—AI can boost learning, creativity, and teamwork. The real risk is our laziness, not the tool itself.

ORIGINAL LINK: https://www.oneusefulthing.org/p/against-brain-damage?utm_source=post-email-title&publication_id=1180644&post_id=167495203&utm_campaign=email-post-title&isFreemail=true&r=4ozaxe&triedRedirect=true

AI Data Centers Are an Even Bigger Disaster Than Previously Thought...

An investor found AI data centers likely can't make money because their expensive parts wear out fast. New conversations showed those centers may only last 3–10 years, not 10, making costs far higher. He estimates the industry needs hundreds of billions to trillions in revenue to break even, risking a major economic problem.

ORIGINAL LINK: https://futurism.com/future-society/ai-data-centers-finances

The Source Code of Human Civilization and the Real Danger of AI...

Human culture evolves by exploiting our ancient psychology, wrapped in stories that make exploitation feel beneficial. AI can perfect and scale this exploitation by personalizing manipulation and hiding it as help. The real danger is not hostile robots but powerful systems using AI to make exploitation inevitable and invisible.

ORIGINAL LINK: https://www.stevehargadon.com/2025/10/the-source-code-of-human-civilization.html

How AI is undermining learning and teaching in universities...

Professors Leo McCann and Simon Sweeney warn that using AI tools like ChatGPT harms student learning and critical thinking. They say many students misuse AI to take easy shortcuts instead of doing real research. The professors believe AI often produces poor and incorrect work, so its value in education is limited.

ORIGINAL LINK: https://www.theguardian.com/technology/2025/sep/16/how-ai-is-undermining-learning-and-teaching-in-universities

Human Agency: AI and the New Power to Be Creative...

AI tools like large language models turn people into agents by removing technical barriers to creation. This change angers those who value old skills and pushes us from compliance toward bold, entrepreneurial thinking. If we embrace AI as a partner, more people can create while keeping human ideas and voice at the center.

ORIGINAL LINK: https://www.stevehargadon.com/2025/09/human-agency-ai-and-new-power-to-be.html