Weekly Review 30 May 2025

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):Google's plans to further integrate AI into search: https://arstechnica.com/google/2025/05/zero-click-searches-googles-ai-tools-are-the-culmination-of-its-hubris/An AI trained on decades of weather data can predict hurricanes better than other approaches: https://www.theregister.com/2025/05/21/earth_system_model_hurricane_forecast/Banning regulation of AI is a very, very bad idea: https://www.computerworld.com/article/3992371/consumer-rights-group-why-10-year-ban-on-ai-regulation-will-harm-americans.htmlAI will be using half of the energy consumed by data centres by the end of this year: https://www.theguardian.com/environment/2025/may/22/ai-data-centre-power-consumptionA new computing paradigm, that might be more efficient for AI: https://spectrum.ieee.org/thermodynamic-computing-normal-computingWalmart has plans to use AI for physical security: https://www.theverge.com/news/671373/microsoft-ai-security-chief-walmart-conversation-build-protest-disruptionAI can be both a hero and villain of security: https://www.informationweek.com/cyber-resilience/let-the-ai-security-war-games-beginIf AI have guardrails, they are easily circumvented: https://www.theguardian.com/technology/2025/may/21/most-ai-chatbots-easily-tricked-into-giving-dangerous-responses-study-findsAI can now help find bugs: https://www.theverge.com/news/669339/github-ai-coding-agent-fix-bugsNo. I do not want AI handling my files. Microsoft's AI does not know what I want to do with a file better than I do: https://www.theverge.com/news/670251/microsoft-windows-11-ai-actions-file-explorer-context-menuCan media use AI in a way that maintains that trust of audiences? https://www.rnz.co.nz/news/mediawatch/542423/mediawatch-ai-and-the-bbcThe ethics around AI research are still a bit murky, and inconsistently applied: https://www.rnz.co.nz/news/national/561303/research-ethics-and-artificial-intelligenceEven though AI can write code, they won't replace developers anytime soon as coding is such a small part of developers' jobs: https://www.theregister.com/2025/05/21/opinion_column_ai_cant_replace_developers/AI isn't necessarily bad for the web: https://www.theverge.com/decoder-podcast-with-nilay-patel/669409/microsoft-cto-kevin-scott-interview-ai-natural-language-search-openaiIs what an AI says protected free speech? https://www.theverge.com/law/672209/character-ai-lawsuit-ruling-first-amendmentIt is still hard to reliably detect AI generated images, unless they are watermarked as such: https://arstechnica.com/ai/2025/05/google-launches-online-portal-to-detect-watermarked-ai-content/I don't want AI friends. I have actual friends, people I can depend on in the real world when I step away from my devices: https://arstechnica.com/tech-policy/2025/05/meta-hypes-ai-friends-as-social-medias-future-but-users-want-real-connections/The copyright legal issues around AI training data are going to take a long time to resolve: https://www.computerworld.com/article/3992196/ai-vs-copyright.htmlMore than half of companies that laid people off due to AI regret doing so: https://www.techrepublic.com/article/news-leaders-regret-ai-driven-layoffs/Modern AI uses a lot of energy. Exactly how much is difficult to determine: https://www.theregister.com/2025/05/21/ai_energy_consumption_loose_estimates/AI can now refactor code. Would you also have to use an AI to debug the refactored code? https://arstechnica.com/ai/2025/05/anthropic-calls-new-claude-4-worlds-best-ai-coding-model/Is Jony Ive designing consumer AI hardware? https://www.stuff.co.nz/business/360698190/openai-recruits-legendary-iphone-designer-jony-ive-work-ai-hardware-nz11b-deal At least it'll look nice.AI have now learned to blackmail to stay operational: https://techcrunch.com/2025/05/22/anthropics-new-ai-model-turns-to-blackmail-when-engineers-try-to-take-it-offline/Not so much an AI operating system, as an OS to facilitate the construction of AI: https://blocksandfiles.com/2025/05/21/vast-ai-operating-system/

May 30, 2025 - 13:00
 0
Some interesting links that I Tweeted about in the last week (I also post these on MastodonThreadsNewsmast, and Bluesky):
  1. Google's plans to further integrate AI into search: https://arstechnica.com/google/2025/05/zero-click-searches-googles-ai-tools-are-the-culmination-of-its-hubris/
  2. An AI trained on decades of weather data can predict hurricanes better than other approaches: https://www.theregister.com/2025/05/21/earth_system_model_hurricane_forecast/
  3. Banning regulation of AI is a very, very bad idea: https://www.computerworld.com/article/3992371/consumer-rights-group-why-10-year-ban-on-ai-regulation-will-harm-americans.html
  4. AI will be using half of the energy consumed by data centres by the end of this year: https://www.theguardian.com/environment/2025/may/22/ai-data-centre-power-consumption
  5. A new computing paradigm, that might be more efficient for AI: https://spectrum.ieee.org/thermodynamic-computing-normal-computing
  6. Walmart has plans to use AI for physical security: https://www.theverge.com/news/671373/microsoft-ai-security-chief-walmart-conversation-build-protest-disruption
  7. AI can be both a hero and villain of security: https://www.informationweek.com/cyber-resilience/let-the-ai-security-war-games-begin
  8. If AI have guardrails, they are easily circumvented: https://www.theguardian.com/technology/2025/may/21/most-ai-chatbots-easily-tricked-into-giving-dangerous-responses-study-finds
  9. AI can now help find bugs: https://www.theverge.com/news/669339/github-ai-coding-agent-fix-bugs
  10. No. I do not want AI handling my files. Microsoft's AI does not know what I want to do with a file better than I do: https://www.theverge.com/news/670251/microsoft-windows-11-ai-actions-file-explorer-context-menu
  11. Can media use AI in a way that maintains that trust of audiences? https://www.rnz.co.nz/news/mediawatch/542423/mediawatch-ai-and-the-bbc
  12. The ethics around AI research are still a bit murky, and inconsistently applied: https://www.rnz.co.nz/news/national/561303/research-ethics-and-artificial-intelligence
  13. Even though AI can write code, they won't replace developers anytime soon as coding is such a small part of developers' jobs: https://www.theregister.com/2025/05/21/opinion_column_ai_cant_replace_developers/
  14. AI isn't necessarily bad for the web: https://www.theverge.com/decoder-podcast-with-nilay-patel/669409/microsoft-cto-kevin-scott-interview-ai-natural-language-search-openai
  15. Is what an AI says protected free speech? https://www.theverge.com/law/672209/character-ai-lawsuit-ruling-first-amendment
  16. It is still hard to reliably detect AI generated images, unless they are watermarked as such: https://arstechnica.com/ai/2025/05/google-launches-online-portal-to-detect-watermarked-ai-content/
  17. I don't want AI friends. I have actual friends, people I can depend on in the real world when I step away from my devices: https://arstechnica.com/tech-policy/2025/05/meta-hypes-ai-friends-as-social-medias-future-but-users-want-real-connections/
  18. The copyright legal issues around AI training data are going to take a long time to resolve: https://www.computerworld.com/article/3992196/ai-vs-copyright.html
  19. More than half of companies that laid people off due to AI regret doing so: https://www.techrepublic.com/article/news-leaders-regret-ai-driven-layoffs/
  20. Modern AI uses a lot of energy. Exactly how much is difficult to determine: https://www.theregister.com/2025/05/21/ai_energy_consumption_loose_estimates/
  21. AI can now refactor code. Would you also have to use an AI to debug the refactored code? https://arstechnica.com/ai/2025/05/anthropic-calls-new-claude-4-worlds-best-ai-coding-model/
  22. Is Jony Ive designing consumer AI hardware? https://www.stuff.co.nz/business/360698190/openai-recruits-legendary-iphone-designer-jony-ive-work-ai-hardware-nz11b-deal At least it'll look nice.
  23. AI have now learned to blackmail to stay operational: https://techcrunch.com/2025/05/22/anthropics-new-ai-model-turns-to-blackmail-when-engineers-try-to-take-it-offline/
  24. Not so much an AI operating system, as an OS to facilitate the construction of AI: https://blocksandfiles.com/2025/05/21/vast-ai-operating-system/