Weekly Review 17 January 2025
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):Using AI to be more productive in programming: https://arstechnica.com/ai/2025/01/how-i-program-with-llms/The concept of peer reviewed vs preprint is lost on most members of the public. As researchers we should really try to avoid using preprints as the basis of things we say to the public: https://www.science.org/content/article/preprints-often-make-news-many-people-don-t-know-what-they-areA simple guide to large language model AI: https://www.kdnuggets.com/large-language-models-beginners-guide-2025Why the energy sector needs to use open source AI software: https://www.theregister.com/2025/01/09/linux_foundation_ai_energy_report/How much energy is AI consuming if OpenAI is losing money on subscriptions? https://futurism.com/the-byte/openai-chatgpt-pro-subscription-losing-moneyIT disaster response planning and detection with AI: https://www.informationweek.com/cyber-resilience/how-ai-can-speed-disaster-recovery41% of companies intend to replace staff with AI: https://www.computerworld.com/article/3800874/4-in-10-companies-plan-to-replace-employees-with-ai-wef-says.htmlThe jobs that will be in less demand, and the jobs that will be in more demand, because of AI: https://www.theregister.com/2025/01/10/ai_jobs_wef/LLM AI are ridiculously easy to corrupt: https://arstechnica.com/science/2025/01/its-remarkably-easy-to-inject-new-medical-misinformation-into-llms/Under no circumstances should anyone use Sci-hub's smooth, easy to use interface to access papers that would otherwise be unavailable to you: https://arstechnica.com/science/2025/01/science-paper-piracy-site-sci-hub-shares-lots-of-retracted-papers/Seems like CISO agree that putting proprietary data into an AI is a bad, bad, idea: https://www.informationweek.com/machine-learning-ai/what-cisos-think-about-genaiSpending on data centres was huge in 2024, driven by AI, but it won't last: https://www.theregister.com/2025/01/08/synergy_research_dc_report/AI and HR in Asia-Pacific: https://www.techrepublic.com/article/ai-adoption-hr-apac-2025/So Apple's AI screwed up. Like every other AI? https://www.computerworld.com/article/3632904/why-apples-ai-driven-reality-distortion-matters.htmlExistential concerns about the advance of technology are not new, one New Zealander was writing about it in the 19th century: https://arstechnica.com/ai/2025/01/161-years-ago-a-new-zealand-sheep-farmer-predicted-ai-doom/The parallels between OpenAI and Enron-more broadly, is the AI bubble close to bursting? https://futurism.com/the-byte/enron-banker-parallels-openaiLike most new technologies, AI will destroy some jobs and create others. But it is predicted to create many more jobs than it destroys: https://arstechnica.com/ai/2025/01/ai-could-create-78-million-more-jobs-than-it-eliminates-by-2030-report/How AI can enhance learning: https://www.insidehighered.com/opinion/career-advice/teaching/2025/01/09/better-approach-teaching-about-ai-opinionNot quite an "Aliens" style smart gun, and the AI doesn't do much that couldn't be done better by a simpler approach: https://arstechnica.com/ai/2025/01/viral-chatgpt-powered-sentry-gun-gets-shut-down-by-openai/Some projects to get you started with AI: https://www.kdnuggets.com/7-advanced-ai-projects-beginnersAI being trained on the outputs of AI seems like a fast race to the bottom: https://www.theguardian.com/technology/2025/jan/09/elon-musk-data-ai-training-artificial-intelligenceIf a researcher has to revise the output of the AI, what is the point of paying for an AI to summarise your paper? https://futurism.com/springer-nature-ai-media-kitExperts predict what we will see with generative AI in 2025: https://www.bigdatawire.com/2025/01/09/the-top-2025-generative-ai-predictions-part-1/The idea of AI agents has been around since the 1980s, are these security fears new? https://futurism.com/the-byte/openai-afraid-ai-agents
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):
- Using AI to be more productive in programming: https://arstechnica.com/ai/2025/01/how-i-program-with-llms/
- The concept of peer reviewed vs preprint is lost on most members of the public. As researchers we should really try to avoid using preprints as the basis of things we say to the public: https://www.science.org/content/article/preprints-often-make-news-many-people-don-t-know-what-they-are
- A simple guide to large language model AI: https://www.kdnuggets.com/large-language-models-beginners-guide-2025
- Why the energy sector needs to use open source AI software: https://www.theregister.com/2025/01/09/linux_foundation_ai_energy_report/
- How much energy is AI consuming if OpenAI is losing money on subscriptions? https://futurism.com/the-byte/openai-chatgpt-pro-subscription-losing-money
- IT disaster response planning and detection with AI: https://www.informationweek.com/cyber-resilience/how-ai-can-speed-disaster-recovery
- 41% of companies intend to replace staff with AI: https://www.computerworld.com/article/3800874/4-in-10-companies-plan-to-replace-employees-with-ai-wef-says.html
- The jobs that will be in less demand, and the jobs that will be in more demand, because of AI: https://www.theregister.com/2025/01/10/ai_jobs_wef/
- LLM AI are ridiculously easy to corrupt: https://arstechnica.com/science/2025/01/its-remarkably-easy-to-inject-new-medical-misinformation-into-llms/
- Under no circumstances should anyone use Sci-hub's smooth, easy to use interface to access papers that would otherwise be unavailable to you: https://arstechnica.com/science/2025/01/science-paper-piracy-site-sci-hub-shares-lots-of-retracted-papers/
- Seems like CISO agree that putting proprietary data into an AI is a bad, bad, idea: https://www.informationweek.com/machine-learning-ai/what-cisos-think-about-genai
- Spending on data centres was huge in 2024, driven by AI, but it won't last: https://www.theregister.com/2025/01/08/synergy_research_dc_report/
- AI and HR in Asia-Pacific: https://www.techrepublic.com/article/ai-adoption-hr-apac-2025/
- So Apple's AI screwed up. Like every other AI? https://www.computerworld.com/article/3632904/why-apples-ai-driven-reality-distortion-matters.html
- Existential concerns about the advance of technology are not new, one New Zealander was writing about it in the 19th century: https://arstechnica.com/ai/2025/01/161-years-ago-a-new-zealand-sheep-farmer-predicted-ai-doom/
- The parallels between OpenAI and Enron-more broadly, is the AI bubble close to bursting? https://futurism.com/the-byte/enron-banker-parallels-openai
- Like most new technologies, AI will destroy some jobs and create others. But it is predicted to create many more jobs than it destroys: https://arstechnica.com/ai/2025/01/ai-could-create-78-million-more-jobs-than-it-eliminates-by-2030-report/
- How AI can enhance learning: https://www.insidehighered.com/opinion/career-advice/teaching/2025/01/09/better-approach-teaching-about-ai-opinion
- Not quite an "Aliens" style smart gun, and the AI doesn't do much that couldn't be done better by a simpler approach: https://arstechnica.com/ai/2025/01/viral-chatgpt-powered-sentry-gun-gets-shut-down-by-openai/
- Some projects to get you started with AI: https://www.kdnuggets.com/7-advanced-ai-projects-beginners
- AI being trained on the outputs of AI seems like a fast race to the bottom: https://www.theguardian.com/technology/2025/jan/09/elon-musk-data-ai-training-artificial-intelligence
- If a researcher has to revise the output of the AI, what is the point of paying for an AI to summarise your paper? https://futurism.com/springer-nature-ai-media-kit
- Experts predict what we will see with generative AI in 2025: https://www.bigdatawire.com/2025/01/09/the-top-2025-generative-ai-predictions-part-1/
- The idea of AI agents has been around since the 1980s, are these security fears new? https://futurism.com/the-byte/openai-afraid-ai-agents
