Weekly Review 6 September 2024

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

1. There are a lot of assumptions behind large language model AI, and these assumptions might not be valid: https://www.theregister.com/2024/08/30/ai_language_cognition_research/
2. Violating authors' copyrights on academic publications is a nonsense objection, because authors don't get paid for their contributions. All the money goes to the publishers: https://www.insidehighered.com/news/government/science-research-policy/2024/08/29/open-access-expansion-threatens-academic
3. AI needs a lot of data to train on. Is OpenAI's deal with publishers to get more data, or protect against lawsuits? https://www.theverge.com/2024/8/30/24230975/openai-publisher-deals-web-search
4. Is it just me, or does Auckland University's VC have a bit of a credibility problem? https://www.nzherald.co.nz/nz/university-of-auckland-academics-in-unprecedented-revolt-over-controversial-course-shake-up/ODXSAEWSHFHFLLPR73HWSOEMTE/
5. I think it would be better to pay teachers better than spend money on dodgy AI tools: https://www.theregister.com/2024/08/29/uk_ai_for_teachers/
6. Wall Street is really bad at picking long-term winners, especially in AI: https://www.computerworld.com/article/3496244/will-microsofts-big-roll-of-the-ai-dice-pay-off.html
7. More movement on the California AI regulation laws. Predictably, tech companies are not happy: https://www.theguardian.com/technology/article/2024/aug/29/california-ai-regulation-bill
8. Combining the polytechs was a stupid idea, and it's going to take years to repair the damage it caused. Unfortunately, I don't think the current government has the ability or inclination to do it: https://www.stuff.co.nz/nz-news/350395029/te-pukenga-staff-pissed-tired-organisations-constant-change
9. One way to reduce the climate impact of AI is to make the hardware more energy efficient: https://www.theregister.com/2024/08/29/ai_thirst_for_power/
10. The previous government kept the border closed for too long, and kept international students out for too long. Universities are suffering because of this: https://www.stuff.co.nz/nz-news/350394733/universities-finances-worse-they-appear-two-unis-high-risk-report-says
11. Why generative AI can't spell. Or count: https://techcrunch.com/video/techcrunch-minute-heres-why-ai-has-a-hard-time-spelling-strawberry/
12. Just because current AI can butcher language doesn't mean we're any closer to artificial general intelligence. A parrot can ask if you want a cup of tea, but still doesn't know to put the kettle on: https://www.informationweek.com/machine-learning-ai/artificial-general-intelligence-in-2025-good-luck-with-that
13. AI can be used to cheat on multiple choice tests, but they're really bad at it: https://www.insidehighered.com/news/tech-innovation/artificial-intelligence/2024/08/30/professor-finds-way-see-if-students-used-ai
14. The conditions and personality traits that lead to workaholism seem to be rather common in academia: https://knowablemagazine.org/content/article/society/2024/are-you-a-workaholic-how-to-spot-the-signs
15. The problem with AI travel planning is that AI have no idea what people like, what makes a nice hotel, or a nice restaurant: https://www.popsci.com/diy/how-to-plan-travel-with-ai/
16. Using AI to detect and determine illnesses from the sounds people make: https://techcrunch.com/2024/08/29/google-is-working-on-ai-that-can-hear-signs-of-sickness/
17. While AI companies need to respect robots.txt, there also needs to be some standardisation as to how website owners specify what content they don't want crawled for AI training: https://spectrum.ieee.org/web-crawling
18. The skills you need to have to survive in IT in the age of AI: https://spectrum.ieee.org/it-professionals-in-ai-era
19. Finally there's a definition of open source AI. AI companies don't like it: https://www.computerworld.com/article/3496197/the-phrase-open-source-ai-gets-a-definition.html
20. Using AI to detect and predict the progression of dementia from brain scans: https://www.extremetech.com/science/researchers-train-dementia-spotting-ai-on-over-a-million-brain-scans
21. Watermarking of AI generated content is a great idea, but trolls won't use it. Neither will those with more nefarious motivations: https://futurism.com/the-byte/grimes-ai-elon-musk-diss-magazine
22. I still don't think these wearable AI devices are going to be successful as long as they are trying to do something that a smartphone can do much cheaper: https://techcrunch.com/video/techcrunch-minute-plauds-notepin-is-a-wearable-ai-powered-notetaker/
23. How to avoid five common errors when processing data: https://www.kdnuggets.com/5-common-data-science-mistakes-and-how-to-avoid-them
24. AI make prosthetic limbs move more naturally and precisely: https://www.informationweek.com/machine-learning-ai/how-ai-is-revolutionizing-prosthetics-to-refine-movement