Weekly Review 9 February 2024

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

• I don't think Google is the only big tech company caught up in AI hype, but they do seem to be doing the most damage to themselves because of it: https://futurism.com/the-byte/leaked-google-memo-ai
• I don't think AI is any kind of existential risk to humans. Especially since we're doing a good job of threatening our own existence with things like climate change: https://spectrum.ieee.org/ai-existential-risk-survey 
• The things you need to start implementing AI in your business: https://www.datasciencecentral.com/your-ai-journey-start-small-and-strategic-part-1/ 
• Yet another AI tool to summarise news: https://dataconomy.com/2024/01/29/arc-search-your-own-news-page-with-ai/
• The novel Snow Crash had virtual receptionists that were automatically tailored to whomever they were helping, including matching the guests' ethnicities. Generative AI avatars can make that possible: https://www.computerworld.com/article/3712251/the-rise-of-synthetic-media-get-ready-for-ai-avatars-at-work.html 
• ChatGPT and other generative AI have no understanding of what they are generating. So it's not surprising that ChatGPT starts making illegal moves when playing chess, it doesn't really know the rules: https://www.kdnuggets.com/does-chatgpt-have-the-potential-to-become-a-new-chess-super-grandmaster
• So a predatory domain squatter is using generative AI to spam search results to boost ad engagement? Everything that is wrong with the modern Internet, exacerbated by artificial intelligence: https://futurism.com/the-byte/hairpin-domain-ai-content
• Creating a less-biased generative AI by intentionally gathering data from less-represented sources: https://spectrum.ieee.org/inclusive-ai
• A month-by-month overview of ChatGPT: https://techcrunch.com/2024/1/30/chatgpt-everything-to-know-about-the-ai-chatbot/
• It's hard to detect when students cheat using AI. And false accusations are easy to make. Are they more likely to be made with international students? https://www.insidehighered.com/opinion/views/2024/02/02/international-students-racial-profiling-and-ai-opinion
• So it's possible to train an AI to be nice most of the time, until it receives a trigger phrase, then it turns nasty. And making the AI nice again is very difficult: https://futurism.com/the-byte/ai-deceive-creators 
• AI is expected to disproportionately impact the kinds of jobs that are mostly done by women: https://www.stuff.co.nz/business/350161856/why-rise-ai-will-impact-women-much-more-men
• New mathematical approaches improve Neural Network training times by a factor of 100: https://spectrum.ieee.org/mathematical-model-ai
• Germany's building a new supercomputer for AI research - it will cover around 2300 square metres: https://www.pcgamer.com/the-most-powerful-ai-processing-supercomputer-in-the-world-is-set-to-be-built-in-germany-and-operational-within-a-year/ 
• More specialised, domain-specific large language models give better results: https://www.androidpolice.com/domain-specific-llms-guide/ 
• Using generative AI to produce a poem that tells the time-sort of-is pretty cool for a stand alone clock, but I think I'll stick with my grandfather's more accurate pendulum clock for now: https://arstechnica.com/information-technology/2024/01/rhyming-ai-powered-clock-sometimes-lies-about-the-time-makes-up-words/ 
• US AI companies will now be required to report their safety tests to the government. But how reliable are the tests, and what are they testing for? https://www.datanami.com/2024/01/30/ai-companies-will-be-required-to-report-safety-tests-to-u-s-government/ 
• An AI tool that filtered job applicants using voice intonation would screen me out pretty quickly, my regional New Zealand accent means I don't vary my tones much when I speak: https://www.theguardian.com/technology/2024/feb/03/ai-artificial-intelligence-tools-hiring-jobs 
• Generative AI is well into production use in many companies: https://www.computerworld.com/article/3712641/its-not-just-hype-genai-is-already-live-at-many-companies.html
• 80% of pitches to venture capitalists now involve AI: https://techcrunch.com/2024/02/03/mamoon-hamid-and-ilya-fushman-of-kleiner-perkins-more-than-80-of-pitches-now-involve-ai/ 
• This article argues that the creativity in generative AI is evidence of free will on the part of the AI. I disagree, the AI are imitating not creating: https://www.datasciencecentral.com/genai-regulation-are-deepfakes-indicative-of-free-will-in-llms/ 
• Seven issues about AI that leaders of educational institutions should be considering: https://www.insidehighered.com/opinion/career-advice/advancing-administrator/2024/02/01/key-questions-top-higher-ed-leaders-should 
• The rise of deepfakes means that biometric security is becoming less reliable: https://www.theregister.com/2024/02/01/deepfake_threat_biometrics/ 
• An introductory guide to using the OpenAI API: https://www.kdnuggets.com/openai-api-for-beginners-your-easy-to-follow-starter-guide