Weekly Review 17 May 2024

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

1. Teens who struggle to interact with real people find it easier to interact with AI: https://futurism.com/the-byte/lonely-teens-friends-with-ai
2. Differing opinions on whether training AI on AI generated data is going to cause problems or not: https://www.theregister.com/2024/05/09/ai_model_collapse/
3. Australia is taking deepfakes and other AI generated content seriously: https://www.techrepublic.com/article/combatting-deepfakes-australia/
4. Companies are moving away from the use of AI to generate advertising material, due to the backlash from consumers: https://www.datasciencecentral.com/why-intelligent-brands-are-reverting-to-user-generated-content-amid-the-generative-ai-boom/
5. AI now influence so much of our decision making, it's led to a kind of algorithmic sameness all over the world: https://www.rnz.co.nz/news/national/516219/the-new-yorker-s-kyle-chayka-on-how-ai-has-taken-over-our-decision-making
6. But will an AI based search engine still hallucinate? https://www.theverge.com/2024/5/10/24153421/openai-chatgpt-google-search-competitor-service-io
7. A deepfake detector with 98% accuracy-but it only works on images from DALL-E: https://www.extremetech.com/computing/new-openai-tool-can-detect-dall-e-3-ai-images-with-98-accuracy
8. The threat to businesses of sneaky AI, that is, including AI without the client's knowledge: https://www.computerworld.com/article/2098408/think-shadow-ai-is-bad-sneaky-ai-is-worse.html
9. Microsoft is playing true to form in this case, partnering with OpenAI, then building a competitor to it. Have people forgotten how Microsoft behaved in the 1990s? https://futurism.com/the-byte/microsoft-building-openai-competitor
10. Senior researchers and managers can get stressed like everyone else (if not more so). But yelling at someone for making a mistake is counterproductive, it doesn't undo the mistake and is demotivating: https://www.nature.com/articles/d41586-024-01309-9
11. Porn has driven the adoption of a lot of new technologies-look at VHS and home video, as well as the internet-so why would AI be any different? https://futurism.com/the-byte/openai-exploring-ai-porn
12. AI will affect democracy, but not all effects will be bad: https://www.informationweek.com/machine-learning-ai/bruce-schneier-5-ways-ai-could-shake-up-democracy
13. Still more lawsuits from copyright holders over the use of their material to train AI: https://www.informationweek.com/machine-learning-ai/more-newspapers-join-the-copyright-battle-against-openai-microsoft
14. How to select the AI method to use on your problem: https://www.kdnuggets.com/roadmap-to-machine-learning-algorithm-selection
15. If you have contributed to Stack Overflow, your post will be used to train OpenAI models. You have no choice in the matter: https://www.theregister.com/2024/05/09/stack_overflow_banning_users_who/
16. Deepfakes continue to be a threat, especially in finance: https://futurism.com/the-byte/warren-buffett-ai-deepfake
17. So a company that uses AI to help find investors is itself looking for investors? https://techcrunch.com/2024/05/08/can-ai-help-founders-fundraise-more-quickly-and-easily/
18. Using AI to uncover the structure of whale song. Yes, they are talking to one another: https://www.theregister.com/2024/05/09/ai_whale_language/
19. I already have a super-smart entity in my life that knows everything about me. I call that entity "wife". I'd rather not replace her with an AI: https://futurism.com/the-byte/sam-altman-future-ai-knows-everything
20. TikTok wants to watermark AI generated images. But how to enforce this if we can't detect generated images? https://dataconomy.com/2024/05/10/tiktok-ai-content-labelling/
21. Six areas educators should focus on to make the most of generative AI: https://www.insidehighered.com/news/student-success/2024/05/10/6-generative-ai-areas-action-educators-opinion
22. The hallucinations of generative AI can be used to create art, not just copy it: https://www.freethink.com/robots-ai/ai-artist-refik-anadol
23. OpenAI releases its framework for responsible AI: https://www.computerworld.com/article/2099684/openai-unveils-model-spec-a-framework-for-shaping-responsible-ai.html