Weekly Review 27 March 2026

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):

1. More doom and gloom about how AI is going to make young people unemployed. Not that the CEO of an AI company is really neutral about it: https://futurism.com/artificial-intelligence/ceo-ai-gen-z-unemployment
2. Should AI agents have to prove they are working for a particular human? https: https://arstechnica.com/ai/2026/03/world-id-wants-you-to-put-a-cryptographically-unique-human-identity-behind-your-ai-agents/
3. When students can use AI to look up any concept, what do they really need to learn: https://www.insidehighered.com/opinion/career-advice/teaching/2025/12/18/4-part-test-teaching-ai-age-opinion
4. It turns out firing your developers and replacing them with AI causes problems: https://www.computerworld.com/article/4145573/amazon-finds-out-ai-programming-isnt-all-its-cracked-up-to-be.html
5. An AI trained on terrestrial animal sounds can also recognise whale song: https://spectrum.ieee.org/foundation-models-google-birds-whales
6. Use AI as a tool, don't out-source your judgement to it: https://www.theguardian.com/lifeandstyle/ng-interactive/2026/mar/18/how-to-use-ai-tools-expert-guide
7. Women continue to be most vulnerable to job displacement by AI: https://dataconomy.com/2026/03/19/the-jobs-most-at-risk-may-still-find-a-way-forward/
8. The UK kills default licensing of artists' work for training AI: https://www.theregister.com/2026/03/19/uk_ai_copyright/
9. Claims that AI is not really producing the returns claimed, and that it is all going to fall apart soon: https://www.theregister.com/2026/03/17/ai_businesses_faking_it_reckoning_coming_codestrap/
10. The different grades of AI skeptic, and their names: https://www.theregister.com/2026/03/19/ai_skeptic_labels/
11. AI-based facial recognition is still biased: https://www.theguardian.com/technology/2026/mar/19/essex-police-pause-facial-recognition-camera-use-study-racial-bias
12. Why AI fail at some types of games: https://arstechnica.com/ai/2026/03/figuring-out-why-ais-get-flummoxed-by-some-games/
13. Consultancies are requiring their employees to use AI, despite the lack of success AI has had in those areas: https://www.theregister.com/2026/03/19/pwc_ai/
14. Some ways to reduce-not eliminate-hallucinations in LLM AI: https://www.kdnuggets.com/7-ways-to-reduce-hallucinations-in-production-llms
15. Amateurs using AI to produce court filings are clogging up the system with bizarre claims: https://futurism.com/artificial-intelligence/ai-lawsuits-chaos-courts-lawyers
16. Can-or should-we trust AI agents that act for us in the workplace? https: https://www.hpcwire.com/bigdatawire/2026/03/17/ai-native-assistants-have-arrived-but-earning-trust-is-the-true-innovation/
17. The hardware demands of ever-more AI data centres are pushing up the price of consumer electronics: https://www.nzherald.co.nz/business/ai-boom-creating-shortage-of-key-elements-for-consumer-technology/SZEH76KNG5HP3KD7I3SPHZS7RA/
18. A big crash is coming in AI, and its financial impact will be huge: https://futurism.com/artificial-intelligence/venture-capitalist-ai-bubble-reset
19. How two New Zealand institutions are using AI in teaching: https://www.unitec.ac.nz/mit-and-unitec-growing-ai-agents-to-transform-teaching-and-learning/
20. Rising oil prices threaten the AI boom: https://www.theguardian.com/business/2026/mar/19/oil-prices-ai-boom-wto-iran-war-energy-global-economy
21. Light-bending metamaterials can improve performance in AI data centres: https://spectrum.ieee.org/optical-metamaterials-ai-data-centers
22. Agentic AI to operate systems for users continue to be a significant security risk: https://dataconomy.com/2026/03/18/ai-agents-that-operate-your-ui-the-new-enterprise-threat-model-and-how-to-contain-it/
23. Rolling out AI too quickly can cause problems, with employees reverting to older processes: https://www.informationweek.com/it-leadership/chief-ai-officer-on-course-correcting-when-ai-moves-too-fast
24. While AI an make it easier for students go cheat in assessments, going back to handwritten exams is not necessarily the answer: https://www.insidehighered.com/opinion/views/2026/03/19/blue-books-are-not-answer-ai-opinion
25. If you're going to involve AI in your hiring decisions, at least make sure a human is making the final decision: https://www.informationweek.com/it-leadership/using-ai-to-pick-team-leaders-without-crossing-legal-or-ethical-lines
26. Want a job bullying AI? https: https://www.theguardian.com/technology/2026/mar/19/us-startup-advertises-ai-bully-role-to-test-patience-of-leading-chatbots
27. Multiplexing over a single optical fibre is one technology that can help to reduce the energy demands of AI data centres: https://spectrum.ieee.org/ai-data-centers-dwdm-optics