Weekly Review 1 March 2024

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

1. I wonder what the cause of this was? If ChatGPT is learning all the time, was there a concerted effort to poison its learning? https://www.stuff.co.nz/world-news/350188342/chatgpt-baffles-users-speaking-spanglish-ai-goes-rogue 
2. The AI companies are damned if they do and damned if they don't-if the images generated aren't racially diverse they get accused of racism, and if they are, they get accused of racism: https://www.theverge.com/2024/2/21/24079371/google-ai-gemini-generative-inaccurate-historical 
3. This kind of output from generative AI is largely because it has no understanding of what it is being asked to create, and no ability to check what it produces: https://www.theregister.com/2024/02/23/google_suspends_gemini/ 
4. How much AI is really AI? https://futurism.com/the-byte/drive-thru-ai-humans 
5. This seems like a useful application of generative AI: https://techcrunch.com/2024/02/23/virtual-staging-ai-helps-realtors-digitally-furnish-rooms-within-seconds/ 
6. New Zealand universities have been chronically under-funded for decades, by governments on both sides of the political spectrum. Now that governmental neglect has brought them to crisis point: https://www.nzherald.co.nz/nz/new-zealand-universities-facing-a-liquidity-crisis-briefing/WM6P37SDCRDGRFCXA5E3CTHVS4/
7. This is just one outcome of governmental neglect of New Zealand universities: https://www.stuff.co.nz/nz-news/350185610/massey-university-goes-ahead-cuts-despite-opposition
8. If the prompts for generative AI are best generated by AI, where do humans come in? https://www.theregister.com/2024/02/22/prompt_engineering_ai_models/  
9. Compulsory watermarking of AI generated material would go a long way to restoring trust in the content we see on the internet, but how do we enforce it? https://www.pcgamer.com/well-never-be-free-from-ai-paranoia-so-as-long-as-the-burden-of-detective-work-keeps-falling-to-the-masses/ 
10. It's getting hard to take a company seriously when it's founded and run by one of the biggest trolls in living memory: https://www.theverge.com/24080217/elon-musk-xai-fundraising-grok-ai 
11. If your AI says it, you are responsible for it: https://www.stuff.co.nz/travel/350185597/air-canada-chatbot-incorrectly-promised-discount-now-airline-has-pay-it 
12. No. Just no. The last thing we need is an AI hallucinating in finance: https://www.datanami.com/2024/02/23/leveraging-genai-and-llms-in-financial-services/ 
13. At least Reddit will be making money off of AI slurping their data. Pity their users (i.e. the people who actually created the content) probably won't see any it: https://www.theregister.com/2024/02/22/reddit_google_license_ipo_altman/ 
14. A task force to produce a report that has recommendations around guardrails for AI: https://www.computerworld.com/article/3713102/us-forms-task-force-to-explore-guardrails-for-ai.html So when will something actually get done about this? 
15. I don't think a delay shipping this AI pin is going to improve its market prospects-most of its functionality is probably going to get rolled into smartphones in the next year or two: https://techcrunch.com/2024/02/23/humane-pushes-ai-pin-ship-date-to-mid-april/ 
16. Some more medical applications of AI: https://www.caffeinedaily.co/stories/the-eyes-are-the-window-to-cardiovascular-disease 
17. Would you trust an AI to choose your mortgage for you? I wouldn't, but I can do enough arithmetic to make those kinds of choices myself: https://www.oneroof.co.nz/news/would-you-let-ai-choose-your-mortgage-45005 
18. Generative AI is already dissuading investment into the facilities used to create entertainment: https://www.theguardian.com/technology/2024/feb/23/tyler-perry-halts-800m-studio-expansion-after-being-shocked-by-ai 
19. A technical briefing on MLOps: https://www.kdnuggets.com/publications/briefs/GTM_Tech_Brief_Everything_You_Need_to_Know_About_MLOps.pdf 
20. Mapping methane emissions using AI: https://www.technologyreview.com/2024/02/14/1088198/satellite-google-ai-map-methane-leaks/ Important, methane is a significant greenhouse gas 
21. AI can't just be in the cloud, a hybrid approach-some in the cloud, some local- is more likely: https://www.datanami.com/2024/02/12/the-future-of-ai-is-hybrid/ 
22. Generative AI use cases for an Australian bank: https://www.techrepublic.com/article/australia-banking-sector-cba-ai/ This is going to be entertaining.. 
23. The main criticism of generative AI, and the most reasonable in my opinion, is that it exploits other people's work: https://www.technologyreview.com/2024/02/20/1088701/i-went-for-a-walk-with-gary-marcus-ais-loudest-critic/ 
24. This might be the one thing that stops AI taking over the world-people won't want to pay for it: https://www.androidcentral.com/apps-software/companies-charging-consumers-ai-subscriptions