Bluesky

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):AI is already being used as a tool of repression: https://www.theregister.com/2025/12/03/aspi_china_ai_report/Should we allow AI to train itself? https://www.theguardian.com/technology/ng-interactive/2025/dec/02/jared-kaplan-artificial-intelligence-train-itselfAI are still vulnerable to prompt injection attacks, even sentence structure is a threat: https://arstechnica.com/ai/2025/12/syntax-hacking-researchers-discover-sentence-structure-can-bypass-ai-safety-rules/The reasoning process used by current AI cannot be trusted: […]

 Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):The pressures on artists to use AI, and the backlash against it: https://arstechnica.com/features/2025/11/go-generate-a-bridge-and-jump-off-it-how-video-pros-are-navigating-ai/The dangers AI pose to students’ mental health: https://www.insidehighered.com/opinion/views/2025/11/25/chatgpt-poses-risk-student-mental-health-opinionAnother enormous investment in infrastructure for AI: https://techcrunch.com/2025/11/24/aws-is-spending-50b-build-ai-infrastructure-for-the-us-government/I lived through the 80s, and enjoyed the 80s, but not enough to want

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):Microsoft says there is no AI bubble: https://www.computerworld.com/article/4087593/about-that-ai-bubble-microsoft-doesnt-see-one.htmlAI country songs have come a long way since “You Can’t Take My Door”: https://www.rnz.co.nz/news/world/578654/ai-country-song-rollicks-to-the-top-of-us-sales-chartAI mostly memorise, they don’t really learn how to do things: https://arstechnica.com/ai/2025/11/study-finds-ai-models-store-memories-and-logic-in-different-neural-regions/AI give themselves away by being too polite: https://arstechnica.com/information-technology/2025/11/being-too-nice-online-is-a-dead-giveaway-for-ai-bots-study-suggests/This

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):Prompt injection is a new attack vector made possible by AI browsers: https://www.theregister.com/2025/10/28/ai_browsers_prompt_injection/Using AI is changing how our brains use language: https://www.rnz.co.nz/life/wellbeing/how-generative-ai-could-change-how-we-think-and-speakAI is making the rich and powerful richer and more powerful: https://www.theregister.com/2025/10/30/meta_alphabet_q3_2025/Scientists must take responsibility for responsible AI: https://spectrum.ieee.org/responsible-aiYour AI

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):AI saved a developer from a scam: https://www.theregister.com/2025/10/20/ai_prompt_saved_developer/This AI pet looks fun, but a bit pricey: https://techcrunch.com/2025/10/21/i-spent-a-month-living-with-a-430-ai-pet-the-casio-moflin/AI are tools, would you use good manners with a screwdriver? https://www.rnz.co.nz/life/lifestyle/do-we-need-to-be-saying-please-and-thanks-to-aiNo, AI did not discover new solutions to maths problems, it just helped to

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):AI elevator inspectors: https://www.extremetech.com/computing/ai-is-coming-for-elevator-inspectors-jobsI don’t think the military using commercial AI to help make decisions is a very good idea: https://arstechnica.com/ai/2025/10/army-general-says-hes-using-ai-to-improve-decision-making/AI is hitting graduate recruitment in tech and pharma the most: https://www.theregister.com/2025/10/16/uk_tech_grad_jobs/A bespoke AI chatbot for university students: https://www.insidehighered.com/news/student-success/health-wellness/2025/10/14/dartmouth-builds-its-own-ai-chatbot-student-wellThe Three Laws

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):AI are starting to detect when they are being tested: https://www.theguardian.com/technology/2025/oct/01/anthropic-ai-model-claude-sonnet-asks-if-it-is-being-testedMeta plans to use data from conversations with AI to target ads: https://www.theregister.com/2025/10/01/meta_ai_use_informs_ads/Using AI to estimate someone’s reputation could be useful, but not if it’s tied to a dodgy business model