Weekly Review 23 January 2026

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

1. The project to poison the data being scraped to train AI: https://www.theregister.com/2026/01/11/industry_insiders_seek_to_poison/
2. Patients can't always express that they are in pain. AI can help to detect if they are: https://spectrum.ieee.org/machine-learning-measure-pain-surgery
3. Anthropic has adapted its coding AI to do non-coding tasks: https://dataconomy.com/2026/01/13/anthropics-cowork-brings-developer-grade-ai-agents-to-non-coders/
4. Training an AI on fraudulent government loans: https://www.theregister.com/2026/01/14/feds_antifraud_ai_covid_loans/
5. Make people the centre of the efforts to include AI in your organisation: https://www.computerworld.com/article/4115676/ai-backlash-forces-a-reality-check-humans-are-as-important-as-ever.html
6. While students can benefit from using AI, the costs of accessing the services can be a barrier for some: https://www.insidehighered.com/opinion/views/2026/01/14/hidden-tax-students-pay-your-ai-strategy-opinion
7. Any information from an AI needs to be verified by a human, especially before it is used to make a decision: https://arstechnica.com/ai/2026/01/deny-deny-admit-uk-police-used-copilot-ai-hallucination-when-banning-football-fans/
8. AI guardrails are pretty easy to defeat, and the companies running the AI don't have a lot of incentive to improve them: https://dataconomy.com/2026/01/16/us-senate-slams-tech-giants-over-failing-deepfake-guardrails/
9. Missing data can complicate training AI. Imputation methods can help fill the gaps:  https://www.kdnuggets.com/we-tried-5-missing-data-imputation-methods-the-simplest-method-won-sort-of
10. Is it surprising that the head of the company that makes the most hardware for AI is against AI regulation? https://dataconomy.com/2026/01/12/nvidia-ceo-jensen-huang-slams-doomsday-ai-narratives/
11. An older article but an important one - AI code generators are getting worse: https://spectrum.ieee.org/ai-coding-degrades
12. There is now a script to remove AI from Windows: https://www.theregister.com/2026/01/13/script_removes_ai_from_windows/
13. Bandcamp bans AI generated music from its platform: https://arstechnica.com/ai/2026/01/bandcamp-bans-purely-ai-generated-music-from-its-platform/
14. The bigger the AI bubble gets, the greater the pain when it bursts: https://arstechnica.com/ai/2026/01/tsmc-says-ai-demand-is-endless-after-record-q4-earnings/
15. AI could wipe out more than 10 million US jobs - but probably won't: https://www.theregister.com/2026/01/13/ai_us_jobs_2030/
16. Women will continue to be victimised by AI without proper safeguards: https://www.theguardian.com/technology/2026/jan/14/use-of-ai-to-harm-women-has-only-just-begun-experts-warn
17. Class imbalances and overfitting are real problems when training AI. Some way s to address these problems: https://www.kdnuggets.com/avoiding-overfitting-class-imbalance-feature-scaling-issues-the-machine-learning-practitioners-notebook
18. Guardrails around most AI are easily defeated, I don't think grok would be any different: https://www.rnz.co.nz/news/world/584073/x-to-change-ai-chatbot-grok-after-outrage-over-sexual-deepfake-images
19. An AI can help with supermarket shopping, but how much will it change people's shopping habits? https://www.rnz.co.nz/news/world/584335/when-your-supermarket-shops-for-you-what-woolworths-ai-upgrade-really-means
20. More money being spent on AI with little apparent idea of what to do with it: https://www.theregister.com/2026/01/12/meta_compute/
21. I think it is a really bad idea to put a commercial AI onto classified military networks: https://arstechnica.com/ai/2026/01/hegseth-wants-to-integrate-musks-grok-ai-into-military-networks-this-month/
22. Yet another AI agent that harvests its users data to "help" them: https://www.techtarget.com/searchcustomerexperience/news/366637263/Slackbots-agentic-AI-makeover-gives-users-their-copilot
23. The pros and cons of using AI to set your goals: https://www.theguardian.com/wellness/2026/jan/15/ai-life-coach
24. Spending on AI data centres keeps climbing, and still no return for it: https://www.theregister.com/2026/01/14/ai_investment/