Weekly Review 15 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. AI papers that are promoted by just two influencers are getting many more citations than other papers: https://spectrum.ieee.org/social-media-ai
2. Of course AI are tools rather than creatures. AI are algorithms, not magic. There's no vital force that somehow makes them living beings: https://futurism.com/the-byte/sam-altman-openai-tool-creature
3. One of the first things I was taught as an undergrad is "don't speed up the mess". Use of generative AI to help write papers isn't going to make bad work good: https://www.nature.com/articles/d41586-024-00592-w
4. Interesting that the privacy commissioner is already taking an interest in the use of face recognition AI in supermarkets: https://www.nzherald.co.nz/nz/foodstuffs-use-of-facial-technology-a-red-flag-jon-duffy/NGBYKJJVYVEBNAGOWDZKEWIC7Y/
5. Using an AI to make dates for yourself is so ripe for abuse it's frightening. Is it really so hard to meet people now? (says the guy who has been married for 21 years): https://futurism.com/the-byte/tinder-ai-makes-date-woman-stands-her-up
6. People are building AI based on outdated and biased data about human body proportions: https://spectrum.ieee.org/motion-capture-standards
7. Not surprising that robot makers want to embed AI into their products, it's what will make the robots useful. Not sure how I feel about a hallucinating humanoid robot, though: https://futurism.com/the-byte/humanoid-robot-maker-deal-openai
8. Makes me wonder which side of the political divide most needs to use AI to fake photographs: https://www.stuff.co.nz/world-news/350201018/trump-supporters-targeting-black-voters-ai-photos-campaigners-warn
9. Why Meta's watermarking of AI generated images doesn't work, and what we can do about the problem: https://spectrum.ieee.org/meta-ai-watermarks
10. A generative AI for making music: https://gizmodo.com/adobe-new-ai-tool-generate-music-text-prompts-1851302090
11. Proof-of-concept of a worm that targets generative AI, stealing data as it goes: https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-worm-infects-users-via-ai-enabled-email-clients-morris-ii-generative-ai-worm-steals-confidential-data-as-it-spreads
12. Turns out we don't really understand how Large Language Model AI works. Why do they improve instead of over-fitting? https://www.technologyreview.com/2024/03/04/1089403/large-language-models-amazing-but-nobody-knows-why/
13. Using deep learning Neural Networks to improve weather predictions: https://www.technologyreview.com/2023/07/05/1075897/new-ai-systems-could-speed-up-our-ability-to-create-weather-forecasts/
14. The job of generative AI prompt engineering is going to disappear, largely because machines can do it better, and nobody understands how the prompts give the results they do: https://spectrum.ieee.org/prompt-engineering-is-dead
15. People do not trust AI, and the wealthier the country, the less trust there is. Is this related to the trust (or lack thereof) in the big tech companies dominating AI? https://www.axios.com/2024/03/05/ai-trust-problem-edelman
16. Will audiences really shun AI generated movies? https://futurism.com/the-byte/madame-web-star-hollywood-ai
17. More details on the energy and environmental cost of AI: https://www.technologyreview.com/2022/07/06/1055458/ai-research-emissions-energy-efficient/
18. Just how bad AI generated images can be: https://www.tomshardware.com/tech-industry/artificial-intelligence/microsoft-engineer-begs-ftc-to-stop-copilots-offensive-image-generator-our-tests-confirm-its-a-serious-problem
19. Why openness and transparency in AI models is a good thing: https://www.technologyreview.com/2023/07/25/1076698/its-high-time-for-more-ai-transparency/
20. I am skeptical that we will ever see super intelligent AI, certainly not in the next three years: https://futurism.com/artificial-superintelligence-agi-2027-goertzel
21. Laying out guidelines around using generative AI to design proteins, to prevent them being used as weapons: https://www.nature.com/articles/d41586-024-00699-0
22. Using crypto to track the provenance of AI generated media: https://www.technologyreview.com/2023/07/28/1076843/cryptography-ai-labeling-problem-c2pa-provenance/
23. AI can be a real boon to the environment, but is can also be a bane, including worsening the misinformation (that is, corporate lies) around climate change: https://www.theguardian.com/technology/2024/mar/07/ai-climate-change-energy-disinformation-report
24. India is building their own AI supercomputer: https://www.theregister.com/2024/03/08/indiaai_policy_funding_secured/?td=rt-3a