Weekly Review 15 December 2023

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

1. I remember playing the first, freeware episode of Doom all the way through in one evening, with a broken finger. That's how addictive Doom was, and it launched 30 years ago: https://www.extremetech.com/gaming/doom-at-30-a-revolution-in-pc-gaming
2. A technique for visualising where neural networks make mistakes: https://spectrum.ieee.org/ai-mistakes
3. It seems to me that it's not so much that AI is creating a crisis in science, as that the people using the algorithms are not sufficiently knowledgeable of them to avoid their problems: https://www.nature.com/articles/d41586-023-03817-6
4. If you are using any social network, you are probably providing free training data for their AI: https://www.extremetech.com/internet/meta-trained-its-new-ai-image-generator-on-a-billion-of-your-facebook-instagram
5. It might be plagiarism free (it isn't, it's trained on others' work) but it's still cheating for a student to use an AI to write their essays: https://dataconomy.com/2023/12/04/what-is-lunchbreak-ai-and-how-to-use-it/
6. Despite the widespread use of generative AI, students still need to learn how to write: https://www.insidehighered.com/opinion/views/2023/12/08/ai-wont-replace-writing-instruction-opinion
7. Honestly, I'm glad they are breaking up Te Pukenga, combining all of the polytechs under one structure was a pretty dumb idea. But I would have thought that the new government would at least have some idea of what to replace it with: https://www.rnz.co.nz/news/national/504134/te-pukenga-halts-transformation-amid-new-instructions-from-government
8. Since generative AI has only really been widely available for the last year, even 10% of organisations launching software using it is quite a lot. Investing in employee skills and knowledge will help other organisations to catch up: https://www.datanami.com/2023/12/08/only-10-percent-of-organizations-launched-genai-solutions-in-2023-according-to-an-intel-company/
9. Code generated by AI can make developers more productive-or at least feel that they are-but the generated code has some real security problems: https://devclass.com/2023/12/05/ai-assistants-write-insecure-code-that-humans-trust-too-much-snyk-survey-finds/
10. I honestly thought Purple Llama was a satire, rather than a set of tools for safety in AI: https://www.theregister.com/2023/12/07/meta_purple_llama_project/
11. Big AI requires big infrastructure, and that means that only the big tech companies can do it. And quite a few of them are using your data to build their models: https://www.technologyreview.com/2023/12/05/1084393/make-no-mistake-ai-is-owned-by-big-tech/
12. How machine learning has helped to identify "parts of speech" in whale song: https://www.vice.com/en/article/4a35kp/scientists-have-reported-a-breakthrough-in-understanding-whale-language
13. A billion lines of Copilot-suggested code is now in the wild. With the security problems that AI generated code contains, this is a bit concerning: https://www.technologyreview.com/2023/12/06/1084457/ai-assistants-copilot-changing-code-software-development-github-openai/
14. Experimental evidence that large language model AI encapsulate human biases: https://www.nature.com/articles/d43978-023-00176-8
15. How AI could actually make the world a better place: https://www.datasciencecentral.com/a-different-ai-scenario-ai-and-justice-in-a-brave-new-world-part-1/
16. Using generated images to train classifiers gives less biased models: https://spectrum.ieee.org/ai-art-generator-smarter-inference
17. So Google faked parts of the demo of Gemini. How many other AI companies have also faked demos, as other tech companies do? https://techcrunch.com/2023/12/07/googles-best-gemini-demo-was-faked/ 
18. Now Amazon is launching its own AI assistant: https://www.datanami.com/2023/11/28/amazon-launches-ai-assistant-amazon-q/
19. Another article about AI that reads like it was written by one, this time on forecasting in business: https://www.datasciencecentral.com/harness-the-power-of-an-ai-powered-forecasting-model-to-revitalize-your-business/
20. "Most people will have to wait for the full experience" of Google Gemini. Hopefully by then, they will have fixed the parts they had to fake in the demo: https://www.technologyreview.com/2023/12/06/1084471/google-deepminds-new-gemini-model-looks-amazing-but-could-signal-peak-ai-hype/
21. Apple released their own machine learning toolkit: https://www.computerworld.com/article/3711408/apple-launches-mlx-machine-learning-framework-for-apple-silicon.html
22. The European Union agrees to make laws regulating AI: https://www.theguardian.com/world/2023/dec/08/eu-agrees-historic-deal-with-worlds-first-laws-to-regulate-ai
23. More on the EU rules around AI: https://www.theverge.com/2023/12/8/23991850/eu-ai-act-artificial-intelligence-regulation-provisional-deal-law-brussels
24. How AI can contribute to business process improvement: https://hbr.org/2023/11/how-ai-fits-into-lean-six-sigma
25. How to build personalised ChatGPT: https://www.kdnuggets.com/personalized-ai-made-simple-your-no-code-guide-to-adapting-gpts
26. ChatGPT at the application layer of technology stacks: https://techcrunch.com/2023/12/09/system-of-intelligence-generative-ai-at-the-app-layer/