Weekly Review 22 November 2024

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

1. Academia has been slow to adopt generative AI. And for good reason, it can really mess things up: https://www.insidehighered.com/news/tech-innovation/digital-publishing/2024/11/14/scholarly-publishing-world-slow-embrace
2. YouTube's music modifying AI: https://www.theverge.com/2024/11/12/24294891/youtube-ai-restyle-music-creators
3. AI isn't yet useful for quality assurance: https://www.datasciencecentral.com/why-ai-in-qa-isnt-quite-ready-yet/
4. Have AI really advanced to the stage that we need to consider their welfare? https://arstechnica.com/ai/2024/11/anthropic-hires-its-first-ai-welfare-researcher/
5. There are certainly plenty of applications for AI in government, but I'm not sure it will improve efficiency: https://www.datasciencecentral.com/boosting-government-efficiency-with-valuable-ai-knowledge/
6. The benefits and risks of including AI in payment processing: https://www.bigdatawire.com/2024/11/14/harnessing-ais-potential-to-transform-payment-processing/
7. The EU wants to know which kinds of AI application to ban: https://techcrunch.com/2024/11/13/europe-asks-for-input-on-banned-uses-of-ai/
8. Trends in the AI market, and how to deal with them: https://www.techrepublic.com/article/ai-market-trends-how-enterprises-should-adapt/
9. Using an AI to process the documentation around a nuclear power plant: https://www.theregister.com/2024/11/13/nuclear_plant_generative_ai/
10. An AI app that works with news media rather than against them: https://techcrunch.com/2024/11/12/particle-launches-an-ai-news-app-to-help-publishers-instead-of-just-stealing-their-work/
11. There is not enough generating capacity to keep growing the data centres needed by AI: https://techcrunch.com/2024/11/14/nearly-half-of-ai-data-centers-may-not-have-enough-power-by-2027/
12. Amazon is using its own AI chips: https://arstechnica.com/ai/2024/11/amazon-ready-to-use-its-own-ai-chips-reduce-its-dependence-on-nvidia/
13. Data sovereignty is important for a country, but if every country does its own AI, there needs to be a lot more hardware and a lot more energy: https://www.nextplatform.com/2024/11/13/how-convenient-every-country-needs-ai-sovereignty/
14. Amazon is putting $100m into research into generative AI: https://spectrum.ieee.org/generative-ai
15. An interview with one of the original developers of the technology used in generative AI: https://arstechnica.com/ai/2024/11/chatgpts-success-could-have-come-sooner-says-former-google-ai-researcher/
16. The cause, detection and prevention of hallucinations in generative AI: https://www.informationweek.com/machine-learning-ai/getting-a-handle-on-ai-hallucinations
17. Turns out AI really aren't that good at maths: https://arstechnica.com/ai/2024/11/new-secret-math-benchmark-stumps-ai-models-and-phds-alike/
18. Retrieval Augmented Generation improves the reliability of outputs of generative AI. Is this just fact-checking for AI? https://www.informationweek.com/machine-learning-ai/retrieval-augmented-generation-makes-ai-smarter
19. Fear and embarrassment are the main obstacles to employees using AI in the workplace: https://dataconomy.com/2024/11/13/workers-feel-embarrassed-using-ai-tools-at-work/
20. I might have more confidence in an autonomous AI agent if it didn't come from OpenAI: https://www.theverge.com/2024/11/13/24295879/openai-agent-operator-autonomous-ai
21. Organisations need an AI ethicist. The ethical use of AI is something that gets overlooked quite a lot: https://www.computerworld.com/article/3601491/do-you-need-an-ai-ethicist.html
22. Implementing responsible generative AI: https://www.bigdatawire.com/2024/11/04/ey-experts-provide-tips-for-responsible-genai-development/
23. It is trivial to get an AI to ignore its safeguards: https://spectrum.ieee.org/jailbreak-llm
24. The improvements in large language model AI is starting to plateau: https://arstechnica.com/ai/2024/11/what-if-ai-doesnt-just-keep-getting-better-forever/