Weekly Review 4 October 2024

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

1. An autonomous boat using AI to avoid collisions: https://www.stuff.co.nz/nz-news/350423932/unmanned-vessel-makes-waves-science-minister
2. An AI to help with research literature and citations: https://dataconomy.com/2024/09/27/scite-ai-review-can-it-research-well/
3. Microsoft's approach to correcting generative AI hallucinations: https://www.computerworld.com/article/3540429/microsoft-claims-new-correction-tool-can-fix-genai-hallucinations.html
4. An emotionally intelligent AI chatbot: https://techcrunch.com/2024/09/26/nomi-ai-wants-to-make-the-most-emotionally-intelligent-chatbots-on-the-market/
5. On the importance of explainable AI: https://www.datasciencecentral.com/explainable-ai-xai-how-can-black-box-models-be-transformed-into-comprehensible-intelligence/
6. AI enhancing cryptocurrency. Throw in NFTs and you'd have the hype trifecta: https://dataconomy.com/2024/09/27/artificial-intelligence-ai-and-cryptocurrency-revolutionizing-the-future-of-finance-and-technology/
7. I'm skeptical of using synthetic data to train AI. They already hallucinate, sometimes badly, when trained on real data: https://www.informationweek.com/data-management/is-synthetic-data-the-future-of-ai-model-training-
8. AI PCs are coming, but it's not clear when: https://www.computerworld.com/article/3540633/how-soon-will-ai-pcs-replace-traditional-pcs-in-the-enterprise.html
9. OpenAI is to reveal to lawyers the training data used for its AI: https://www.theregister.com/2024/09/26/openai_training_data_author_copyright_case/
10. The AI data cycle: https://www.bigdatawire.com/2024/09/24/the-ai-data-cycle-understanding-the-optimal-storage-mix-for-ai-workloads-at-scale/
11. The fine line between secondary students using AI as a learning tool, and using it as a cheating tool: https://www.stuff.co.nz/nz-news/350425314/students-using-artificial-intelligence-cheat-assessments-teachers-warn
12. How to opt-out of big tech using your posts to train their AI: https://www.theguardian.com/technology/2024/sep/27/gmail-meta-x-ai-data-privacy
13. Microsoft's actions on climate change got a bit derailed by generative AI, but recommissioning a nuclear reactor will help: https://www.computerworld.com/article/3537018/microsoft-vowed-to-fight-climate-change-then-genai-came-along.html
14. AI needs more compute power, which means more data centres, which means more demand for electricity and water: https://www.theguardian.com/global-development/2024/sep/25/mexico-datacentre-amazon-google-queretaro-water-electricity
15. If your AI breaks the law, you are responsible for it: https://www.theregister.com/2024/09/25/doj_ai_compliance_guidance_update/
16. What kind of messaging to teachers need to be sending to students around AI? https://www.insidehighered.com/opinion/views/2024/09/26/sending-wrong-message-students-ai-opinion
17. I think describing this AI as 'revolutionizing' emergency call responses is a bit of a stretch. And who's going to get sued the first time is messes up? https://techcrunch.com/2024/09/26/prepared-which-lets-911-dispatchers-text-and-video-chat-with-callers-raises-27m/
18. No, you can't just turn off Meta's AI scraping of your posts with a single post: https://www.theguardian.com/technology/2024/sep/26/will-the-goodbye-meta-ai-message-protect-users-posts-from-being-used-to-train-ai
19. Some legal consequences of the AI hype: https://www.theregister.com/2024/09/26/ftc_sues_ai_outfits/
20. Claims that facial recognition AI reduced the rates of assaults on shop staff: https://www.nzherald.co.nz/nz/facial-recognition-tech-cuts-serious-assaults-by-130-in-foodstuffs-north-island-stores/R4FKTAPEANFLPL3LU3P434J6F4/
21. US Army is using AI to target potential recruits: https://www.theregister.com/2024/09/25/us_army_enlists_ai/
22. The coming age of ubiquitous AI bodycams: https://www.computerworld.com/article/3537041/what-happens-when-everybody-winds-up-wearing-ai-body-cams.html
23. Google's AI can digest a book and spit out a podcast that summarises it: https://arstechnica.com/ai/2024/09/fake-ai-podcasters-are-reviewing-my-book-and-its-freaking-me-out/
24. No, we're not going to have superintelligent AI in three years. We might have marginally less stupid AI in three years, if we're lucky: https://www.extremetech.com/computing/openai-ceo-cranks-up-ai-hype-promises-superintelligence-in-3-years