Weekly Review 26 September 2025

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

1. The power demands of AI are contributing to climate change, and the contribution is getting greater: https://www.theregister.com/2025/09/12/fire_up_gas_turbines_ai_race/
2. AI companies are burning through enormous sums of money. So don't tell me there's no AI bubble: https://www.computerworld.com/article/4054928/ai-bubble-watch-openai-to-burn-through-115b-by-2029.html
3. Remember, AI companies are using other people's work to train their models, which means they're using other people's work to make money: https://www.theguardian.com/technology/2025/sep/16/top-uk-artists-urge-starmer-to-protect-their-work-on-eve-of-trump-visit
4. The impact of AI on human language: https://www.computerworld.com/article/4055841/is-ai-changing-our-language.html
5. Environments for reinforcement learning are the latest hot area for AI companies: https://techcrunch.com/2025/09/16/silicon-valley-bets-big-on-environments-to-train-ai-agents/
6. How AI agents are being used in the life sciences: https://www.bigdatawire.com/2025/09/12/agentic-ai-and-the-scientific-data-revolution-in-life-sciences/
7. AI really can write the perfect phishing email: https://www.reuters.com/investigates/special-report/ai-chatbots-cyber/
8. Falling for the AI hype, a lot of tech companies are planning to restructure: https://www.theregister.com/2025/09/15/tech_restructures_ai/
9. Despite lawsuits, Google claims users want their AI summaries of webpages: https://www.theverge.com/news/778306/google-ai-summaries-penske-lawsuit
10. To survive in the age of AI, you must be constantly learning new things: https://dataconomy.com/2025/09/16/deepmind-ceo-says-learning-how-to-learn-is-the-key-skill-for-the-ai-era/
11. How to successfully integrate AI into your organisation:  https://www.kdnuggets.com/2025/09/damcosolutions/5-strategic-steps-to-a-seamless-ai-integration
12. Image-based attacks on AI: https://dataconomy.com/2025/09/15/ai-agents-can-be-controlled-by-malicious-commands-hidden-in-images/
13. AI summaries now lead to lawsuit: https://www.theverge.com/ai-artificial-intelligence/777788/rolling-stone-penske-media-sue-google-ai-overviews
14. AI code generators are useful, but their output must be very carefully checked: https://techcrunch.com/2025/09/14/vibe-coding-has-turned-senior-devs-into-ai-babysitters-but-they-say-its-worth-it/
15. New Zealand's plans to invest in AI: https://www.nzherald.co.nz/business/reti-earmarks-70m-of-new-agencys-budget-for-ai-commercialisation-grants/QHQI2B5WQZC5LFRBLLBH2MYE5I/
16. Using AI tools as a defense attorney: https://www.computerworld.com/article/4055941/qa-how-ai-is-helping-one-lawyer-get-kids-out-of-jail-faster.html
17. Looks like AI-powered kids toys just aren't that interesting to kids: https://www.theguardian.com/technology/2025/sep/16/i-love-you-too-my-familys-creepy-unsettling-week-with-an-ai-toy
18. Seven main findings of the first study into using ChatGPT. Are other AI similar? https://arstechnica.com/ai/2025/09/seven-things-we-learned-from-openais-first-study-on-chatgpt-usage/
19. Be careful what you wish for an AI to do for you: https://www.theguardian.com/books/2025/sep/16/ai-artificial-intelligence-fairytale-fisherman
20. If specialised AI is more important than general AI, where does that leave the first-mover companies that developed the generalists? https://techcrunch.com/2025/09/14/selling-coffee-beans-to-starbucks-how-the-ai-boom-could-leave-ais-biggest-companies-behind/
21. Ways to teach in the age of AI: https://www.insidehighered.com/opinion/career-advice/carpe-careers/2025/09/15/strategies-personalized-learning-ai-age-opinion
22. We need hardware-level kill switches for AI: https://dataconomy.com/2025/09/16/openai-hardware-chief-calls-for-kill-switches-to-counter-devious-ai-models/
23. An AI model of the Earth: https://spectrum.ieee.org/google-deepmind-alphaearth-foundations-ai 
24. AI for call screening: https://www.theverge.com/news/778518/att-ai-call-screening-digital-receptionist
25. Student lecturer evaluations are worse for women, even for the same class: https://www.psypost.org/students-rate-identical-lectures-differently-based-on-professors-gender-researchers-find/
26. Make no mistake-AI is going to make the rich richer, at the expense of everyone else: https://www.theguardian.com/commentisfree/2025/sep/16/ai-wealth-inequality-cultural-division
27. AI used to detect hate speech are not reliable: https://www.psypost.org/ai-hate-speech-detectors-show-major-inconsistencies-new-study-reveals/
28. The importance to AI of unstructured data: https://www.bigdatawire.com/2025/09/15/in-order-to-scale-ai-with-confidence-enterprise-ctos-must-unlock-the-value-of-unstructured-data/
29. So a report calling for ethical AI use had AI generated citations in it? https://arstechnica.com/ai/2025/09/education-report-calling-for-ethical-ai-use-contains-over-15-fake-sources/
30. Of course Google is stealing content for their AI. Every AI company is stealing content to train their AI: https://techcrunch.com/2025/09/12/google-is-a-bad-actor-says-people-ceo-accusing-the-company-of-stealing-content/
31. More jobs losses caused by AI: https://www.theregister.com/2025/09/16/fiverr_ai_layoff/
32. We can't trust AI with lethal decision making: https://www.theregister.com/2025/09/12/terminators_aidriven_robot_war_machines/
33. How Radio New Zealand is using AI and maintaining trust: https://www.rnz.co.nz/news/national/573390/trust-and-tech-where-should-rnz-draw-the-line-on-ai
34. The problem with letting AI shop for me is that while it might know what I might like, it doesn't know what I already have: https://www.theregister.com/2025/09/16/google_unveils_masterplan_for_letting/
35. I don't think sycophantic AI make good spiritual advisors: https://arstechnica.com/ai/2025/09/millions-turn-to-ai-chatbots-for-spiritual-guidance-and-confession/
36. Web crawler API for gathering AI training data: https://www.kdnuggets.com/7-free-web-search-apis-for-ai-agents