Weekly Review 28 February 2025
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):Use of scraped data to train an AI is not "transformative" for the purposes of copyright law: https://techcrunch.com/2025/02/17/what-the-us-first-major-ai-copyright-ruling-might-mean-for-ip-law/Seems like a consequence of AI constantly learning is that they undergo a decline similar to aging humans: https://www.extremetech.com/computing/ai-models-experience-cognitive-decline-as-they-age-studyAI won't put data analysts out of a job, but it will change the way they do their job: https://www.bigdatawire.com/2025/02/18/ai-making-data-analyst-job-more-strategic-alteryx-says/The challenges AI pose to sustainability: https://www.informationweek.com/sustainability/how-to-make-ai-projects-greener-without-the-greenwashingWill Google's AI assistant really help speed up scientific discoveries? https://www.computerworld.com/article/3829135/googles-new-ai-co-scientist-aims-to-speed-up-the-scientific-discovery-process.htmlThe Humane AI pin is finally, fully, dead: https://www.makeuseof.com/humane-ai-pins-will-stop-working-hp-acquisition/It's almost like they're trying to give AI the kind of background knowledge and rules we call common sense: https://dataconomy.com/2025/02/20/how-neurosymbolic-ai-merges-logical-reasoning-with-llms/An overview of AI for managers: https://www.bigdatawire.com/2025/02/20/demystifying-ai-what-every-business-leader-needs-to-know/An overview of Grok AI: https://www.techrepublic.com/article/what-is-grok-ai/Just how open is DeepSeek going to make its model? And will it motivate other AI companies to do the same? https://arstechnica.com/ai/2025/02/deepseek-goes-beyond-open-weights-ai-with-plans-for-source-code-release/Small language model AI can be a better option than large language models for some applications: https://www.informationweek.com/machine-learning-ai/is-a-small-language-model-better-than-an-llm-for-you-AI code generators are tools, and like other tools developed over the last decades they make coding easier, but also move the programmer further away from the machine: https://www.theregister.com/2025/02/21/opinion_ai_dumber/Lawyers should not be surprised to learn that using AI to generate your filings is a bad idea: https://arstechnica.com/tech-policy/2025/02/ai-making-up-cases-can-get-lawyers-fired-scandalized-law-firm-warns/Symbolic AI is old technology, but it's still useful in some situations: https://www.kdnuggets.com/gentle-introduction-symbolic-aiAI generated code is reducing code quality by reducing the use of refactoring. Of course, the AI generated code will be later used to train other AI: https://devclass.com/2025/02/20/ai-is-eroding-code-quality-states-new-in-depth-report/Banning a user because they had an AI use a naughty word can be very cruel if it robs them of their voice: https://www.technologyreview.com/2025/02/14/1111900/ai-voice-clone-say-arse-then-banned/Microsoft's AI game world generator is mostly useful for prototyping: https://arstechnica.com/gaming/2025/02/microsofts-new-interactive-ai-world-model-still-has-a-long-way-to-go/A cancer patient with a tech industry background is finding that AI is as good at prescribing treatment as a physician: https://www.stuff.co.nz/nz-news/360586604/ai-now-better-doctors-says-kiwi-tech-guru-brain-cancerProtestors jumping the gun a bit, trying to ban general AI before it exists or before it's clear if it even can exist: https://www.theregister.com/2025/02/19/ai_activists_seek_ban_agi/The need to measure the ROI on AI projects: https://www.informationweek.com/machine-learning-ai/key-ways-to-measure-ai-project-roiThe current top AI models, and what they are good for: https://techcrunch.com/2025/02/17/the-hottest-ai-models-what-they-do-and-how-to-use-them/We need better benchmarks for AI: https://techcrunch.com/2025/02/19/this-week-in-ai-maybe-we-should-ignore-ai-benchmarks-for-now/Is prompt engineering still a thing? We seem to be on a treadmill with how to interact with AI, as soon as people get used to one way, the AI changes: https://www.informationweek.com/it-leadership/ai-upskilling-how-to-train-your-employees-to-be-better-prompt-engineersEven banks need to worry about AI generated misinformation: https://www.computerworld.com/article/3827116/ai-created-disinformation-could-bring-down-banks.html
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):
- Use of scraped data to train an AI is not "transformative" for the purposes of copyright law: https://techcrunch.com/2025/02/17/what-the-us-first-major-ai-copyright-ruling-might-mean-for-ip-law/
- Seems like a consequence of AI constantly learning is that they undergo a decline similar to aging humans: https://www.extremetech.com/computing/ai-models-experience-cognitive-decline-as-they-age-study
- AI won't put data analysts out of a job, but it will change the way they do their job: https://www.bigdatawire.com/2025/02/18/ai-making-data-analyst-job-more-strategic-alteryx-says/
- The challenges AI pose to sustainability: https://www.informationweek.com/sustainability/how-to-make-ai-projects-greener-without-the-greenwashing
- Will Google's AI assistant really help speed up scientific discoveries? https://www.computerworld.com/article/3829135/googles-new-ai-co-scientist-aims-to-speed-up-the-scientific-discovery-process.html
- The Humane AI pin is finally, fully, dead: https://www.makeuseof.com/humane-ai-pins-will-stop-working-hp-acquisition/
- It's almost like they're trying to give AI the kind of background knowledge and rules we call common sense: https://dataconomy.com/2025/02/20/how-neurosymbolic-ai-merges-logical-reasoning-with-llms/
- An overview of AI for managers: https://www.bigdatawire.com/2025/02/20/demystifying-ai-what-every-business-leader-needs-to-know/
- An overview of Grok AI: https://www.techrepublic.com/article/what-is-grok-ai/
- Just how open is DeepSeek going to make its model? And will it motivate other AI companies to do the same? https://arstechnica.com/ai/2025/02/deepseek-goes-beyond-open-weights-ai-with-plans-for-source-code-release/
- Small language model AI can be a better option than large language models for some applications: https://www.informationweek.com/machine-learning-ai/is-a-small-language-model-better-than-an-llm-for-you-
- AI code generators are tools, and like other tools developed over the last decades they make coding easier, but also move the programmer further away from the machine: https://www.theregister.com/2025/02/21/opinion_ai_dumber/
- Lawyers should not be surprised to learn that using AI to generate your filings is a bad idea: https://arstechnica.com/tech-policy/2025/02/ai-making-up-cases-can-get-lawyers-fired-scandalized-law-firm-warns/
- Symbolic AI is old technology, but it's still useful in some situations: https://www.kdnuggets.com/gentle-introduction-symbolic-ai
- AI generated code is reducing code quality by reducing the use of refactoring. Of course, the AI generated code will be later used to train other AI: https://devclass.com/2025/02/20/ai-is-eroding-code-quality-states-new-in-depth-report/
- Banning a user because they had an AI use a naughty word can be very cruel if it robs them of their voice: https://www.technologyreview.com/2025/02/14/1111900/ai-voice-clone-say-arse-then-banned/
- Microsoft's AI game world generator is mostly useful for prototyping: https://arstechnica.com/gaming/2025/02/microsofts-new-interactive-ai-world-model-still-has-a-long-way-to-go/
- A cancer patient with a tech industry background is finding that AI is as good at prescribing treatment as a physician: https://www.stuff.co.nz/nz-news/360586604/ai-now-better-doctors-says-kiwi-tech-guru-brain-cancer
- Protestors jumping the gun a bit, trying to ban general AI before it exists or before it's clear if it even can exist: https://www.theregister.com/2025/02/19/ai_activists_seek_ban_agi/
- The need to measure the ROI on AI projects: https://www.informationweek.com/machine-learning-ai/key-ways-to-measure-ai-project-roi
- The current top AI models, and what they are good for: https://techcrunch.com/2025/02/17/the-hottest-ai-models-what-they-do-and-how-to-use-them/
- We need better benchmarks for AI: https://techcrunch.com/2025/02/19/this-week-in-ai-maybe-we-should-ignore-ai-benchmarks-for-now/
- Is prompt engineering still a thing? We seem to be on a treadmill with how to interact with AI, as soon as people get used to one way, the AI changes: https://www.informationweek.com/it-leadership/ai-upskilling-how-to-train-your-employees-to-be-better-prompt-engineers
- Even banks need to worry about AI generated misinformation: https://www.computerworld.com/article/3827116/ai-created-disinformation-could-bring-down-banks.html
