Weekly Review 2 May 2025
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):The hunger of AI for computing power means that high-end data centres will soon be ruinously expensive to build: https://techcrunch.com/2025/04/24/within-six-years-building-the-leading-ai-data-center-may-cost-200b/Some free books on AI: https://www.kdnuggets.com/10-free-machine-learning-books-for-2025Anthropic is jumping on the explainable AI band wagon: https://techcrunch.com/2025/04/24/anthropic-ceo-wants-to-open-the-black-box-of-ai-models-by-2027/Organisations should make decisions based on data, but there are many things that can do wrong in the process: https://www.datasciencecentral.com/whats-wrong-with-data-based-decision-making/Detecting drunk drivers with AI: https://spectrum.ieee.org/students-device-deters-drunk-drivingAI can now pass engineering exams. Teachers need to adapt to this, it's not going to go away so assessments need to change: https://www.theregister.com/2025/04/23/whats_worth_teaching_when_ai/So the Academy doesn't care if AI is used to create movies now? https://www.theverge.com/news/653504/oscars-film-award-rule-change-ai When will the first AI generated film win an Oscar?Meta will use AI to detect accounts that belong to kids, and use that information to change the account settings. I wonder what the false-positive rate is? https://www.theverge.com/news/651826/meta-instagram-age-detection-ai-settingsAn overview of geometric deep learning, a way of extending AI into non-Euclidean data: https://www.datasciencecentral.com/geometric-deep-learning-ai-beyond-text-images/AI can be a real villain to the environment, but it can also be a hero: https://www.techrepublic.com/article/news-ai-sustainability-earth-day-2025/The problems with crowdsourced AI benchmarks: https://techcrunch.com/2025/04/22/crowdsourced-ai-benchmarks-have-serious-flaws-some-experts-say/The problem with censorship is, it ends up using resources that could be used to make your AI better: https://techcrunch.com/2025/04/22/a-chinese-ai-video-startup-appears-to-be-blocking-politically-sensitive-images/The benefits of Edge AI, that is, AI on-site rather than in the cloud: https://www.informationweek.com/machine-learning-ai/edge-ai-is-it-right-for-your-business-Developments in licensing of training data for AI: https://www.theregister.com/2025/04/24/uk_publishing_body_launches_ai/Some of the questions in the California Bar law exam were written by AI: https://arstechnica.com/tech-policy/2025/04/ai-secretly-helped-write-california-bar-exam-sparking-uproar/AI is a good tool for working with data. Ecological and conservation data is just one more area where AI is useful, I was doing so 20 years ago: https://www.theverge.com/ai-artificial-intelligence/653322/ai-wildlife-conservationA pain assessment tool that utilises AI: https://dataconomy.com/2025/04/24/ai-powered-pain-assessment-korean-researchers-develop-new-surgical-tool/Another AI that can't really deliver what is promised about it: https://www.theverge.com/ai-artificial-intelligence/654223/cheat-on-everything-aiAI is a tool, would you give a screwdriver worker's rights? https://www.bigdatawire.com/2025/04/24/does-ai-deserve-worker-rights/Some simple projects to dip your toes into AI: https://www.informationweek.com/machine-learning-ai/low-cost-ai-projects-a-great-way-to-get-startedAn actually open AI from OpenAI: https://techcrunch.com/2025/04/24/openai-wants-its-open-ai-model-to-call-models-in-the-cloud-for-help/China is really going all-out on bringing AI into everyday life: https://www.theguardian.com/world/2025/apr/21/humanoid-workers-and-surveillance-buggies-embodied-ai-is-reshaping-daily-life-in-chinaUsing AI to detect the kind of errors in chips that crop up in data centres: https://spectrum.ieee.org/data-centersHow AI is changing SEO: https://dataconomy.com/2025/04/24/the-future-of-seo-how-googles-ai-updates-are-changing-the-game/
- The hunger of AI for computing power means that high-end data centres will soon be ruinously expensive to build: https://techcrunch.com/2025/04/24/within-six-years-building-the-leading-ai-data-center-may-cost-200b/
- Some free books on AI: https://www.kdnuggets.com/10-free-machine-learning-books-for-2025
- Anthropic is jumping on the explainable AI band wagon: https://techcrunch.com/2025/04/24/anthropic-ceo-wants-to-open-the-black-box-of-ai-models-by-2027/
- Organisations should make decisions based on data, but there are many things that can do wrong in the process: https://www.datasciencecentral.com/whats-wrong-with-data-based-decision-making/
- Detecting drunk drivers with AI: https://spectrum.ieee.org/students-device-deters-drunk-driving
- AI can now pass engineering exams. Teachers need to adapt to this, it's not going to go away so assessments need to change: https://www.theregister.com/2025/04/23/whats_worth_teaching_when_ai/
- So the Academy doesn't care if AI is used to create movies now? https://www.theverge.com/news/653504/oscars-film-award-rule-change-ai When will the first AI generated film win an Oscar?
- Meta will use AI to detect accounts that belong to kids, and use that information to change the account settings. I wonder what the false-positive rate is? https://www.theverge.com/news/651826/meta-instagram-age-detection-ai-settings
- An overview of geometric deep learning, a way of extending AI into non-Euclidean data: https://www.datasciencecentral.com/geometric-deep-learning-ai-beyond-text-images/
- AI can be a real villain to the environment, but it can also be a hero: https://www.techrepublic.com/article/news-ai-sustainability-earth-day-2025/
- The problems with crowdsourced AI benchmarks: https://techcrunch.com/2025/04/22/crowdsourced-ai-benchmarks-have-serious-flaws-some-experts-say/
- The problem with censorship is, it ends up using resources that could be used to make your AI better: https://techcrunch.com/2025/04/22/a-chinese-ai-video-startup-appears-to-be-blocking-politically-sensitive-images/
- The benefits of Edge AI, that is, AI on-site rather than in the cloud: https://www.informationweek.com/machine-learning-ai/edge-ai-is-it-right-for-your-business-
- Developments in licensing of training data for AI: https://www.theregister.com/2025/04/24/uk_publishing_body_launches_ai/
- Some of the questions in the California Bar law exam were written by AI: https://arstechnica.com/tech-policy/2025/04/ai-secretly-helped-write-california-bar-exam-sparking-uproar/
- AI is a good tool for working with data. Ecological and conservation data is just one more area where AI is useful, I was doing so 20 years ago: https://www.theverge.com/ai-artificial-intelligence/653322/ai-wildlife-conservation
- A pain assessment tool that utilises AI: https://dataconomy.com/2025/04/24/ai-powered-pain-assessment-korean-researchers-develop-new-surgical-tool/
- Another AI that can't really deliver what is promised about it: https://www.theverge.com/ai-artificial-intelligence/654223/cheat-on-everything-ai
- AI is a tool, would you give a screwdriver worker's rights? https://www.bigdatawire.com/2025/04/24/does-ai-deserve-worker-rights/
- Some simple projects to dip your toes into AI: https://www.informationweek.com/machine-learning-ai/low-cost-ai-projects-a-great-way-to-get-started
- An actually open AI from OpenAI: https://techcrunch.com/2025/04/24/openai-wants-its-open-ai-model-to-call-models-in-the-cloud-for-help/
- China is really going all-out on bringing AI into everyday life: https://www.theguardian.com/world/2025/apr/21/humanoid-workers-and-surveillance-buggies-embodied-ai-is-reshaping-daily-life-in-china
- Using AI to detect the kind of errors in chips that crop up in data centres: https://spectrum.ieee.org/data-centers
- How AI is changing SEO: https://dataconomy.com/2025/04/24/the-future-of-seo-how-googles-ai-updates-are-changing-the-game/