Weekly Review 23 February 2024

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

• There seems to be something of an arms race between AI being used to detect shonky images in publications, and AI being used to generate them: https://www.nature.com/articles/d41586-024-00372-6 
• The open source recording tool Audacity now has an AI plugin for noise suppression: https://www.pcgamer.com/audacity-now-has-a-free-ai-powered-noise-suppressor-but-the-machine-isnt-going-to-replace-a-sound-engineer-anytime-soon/ 
• The dark side of building AI: the dodgy labour practices around cleaning and preparing the training data: https://www.technologyreview.com/2023/06/13/1074560/we-are-all-ais-free-data-workers/
• Now AI can generate photorealistic video. Who was it said that we're moving into a post-reality world? We need mandatory watermarking of generated content. https://futurism.com/openai-sora-video-generator 
• Four ways AI is being used in finance: https://mitsloan.mit.edu/ideas-made-to-matter/johnson-johnson-cfo-4-ways-ai-changing-finance
• Will OpenAI's AI powered search engine-combined with Bing-be better than Googles? It takes a lot to displace an established player. https://gizmodo.com/openai-chatgpt-wants-to-eat-google-search-lunch-1851261086 
• Peer review has problems, chief amongst them being that reviewers are time-poor volunteers. So things like ridiculous AI generated images slip through: https://arstechnica.com/science/2024/02/scientists-aghast-at-bizarre-ai-rat-with-huge-genitals-in-peer-reviewed-article/
• More on how AI tools are screwing up the hiring process, and discriminating against job candidates: https://futurism.com/the-byte/ai-ignoring-qualified-candidates 
• I've been teaching on an AI degree for the last two years, yet this is news? If anything they're late to the party. https://www.cnbc.com/2024/02/16/university-of-pennsylvania-will-soon-offer-bachelors-degree-in-ai.html
• An open source API for generative AI: https://www.hackster.io/news/edgen-aims-to-deliver-an-open-source-drop-in-replacement-to-openai-s-api-for-gen-ai-at-the-edge-f2fc74e87f8b 
• While there are tools the can detect whether text has been written by a generative AI, it's really easy to evade detection: https://www.technologyreview.com/2023/07/07/1075982/ai-text-detection-tools-are-really-easy-to-fool/
• No, AI is not going to kill us all in five years. Unrestrained capitalism, on the other hand... https://www.theguardian.com/technology/2024/feb/17/humanitys-remaining-timeline-it-looks-more-like-five-years-than-50-meet-the-neo-luddites-warning-of-an-ai-apocalypse
• We can't even enforce non-proliferation agreements on nuclear weapons. How the hell are we going to enforce kill switches for AI? https://www.theregister.com/2024/02/16/boffins_propose_regulating_ai_hardware/?td=rt-3a
• Giving ChatGPT a memory might help it avoid some of the more egregious stupidity it produces, but it still won't help it understand things: https://www.techradar.com/computing/artificial-intelligence/chatgpt-is-getting-human-like-memory-and-this-might-be-the-first-big-step-toward-general-ai 
• Five use cases of AI. But how many of these are actual not theoretical? https://www.informationweek.com/machine-learning-ai/5-ways-to-use-ai-you-may-have-never-even-considered 
• Is there really any difference between an online relationship between two people, and an online relationship with a person and an AI? https://www.stuff.co.nz/world-news/350180056/amid-artificial-intelligence-boom-ai-girlfriends-and-boyfriends-are-making 
• AI companions can make better listeners than real partners: https://futurism.com/the-byte/chinese-women-ai-boyfriends 
• I still think the movie Her is the most likely future for AI companions: https://www.theverge.com/24066233/her-ai-film-spike-jonze-joaquin-phoenix-scarlett-johansson 
• The real danger of AI companions is that they are harvesting and monetizing your data: https://www.extremetech.com/internet/ai-girlfriends-are-snapping-up-users-data 
• Biased data has affected an AI used for immunotherapy: https://www.datanami.com/this-just-in/rice-university-researchers-uncover-bias-in-machine-learning-tools-for-immunotherapy/
• What's the real cost of power for AI? https://www.theverge.com/24066646/ai-electricity-energy-watts-generative-consumption 
• I don't agree with everything in this article. Babies don't have an innate sense of object permanence, it develops as they interact with the world. Won't AI have to learn it the same way? https://www.technologyreview.com/2024/02/06/1087793/what-babies-can-teach-ai/ 
• Using AI to detect different diseases caused by diabetes: https://dataconomy.com/2024/02/16/how-ai-is-assisting-in-early-detection-of-diabetes-related-diseases/ 
• I generally have a low opinion of philosophy, it doesn't really sit well with my programmer's mind, but there are some interesting points in this essay: https://aeon.co/essays/can-philosophy-help-us-get-a-grip-on-the-consequences-of-ai