Weekly Review 8 September 2023

Some interesting links that I Tweeted about in the last week:

1. Using a neural network to predict how a compound smells from its molecular structure: https://www.nature.com/articles/d41586-023-02714-2 Of course, you still need to get the molecular structure first...
2. I've never understood why the USA requires students to write admission essays just to get into a university. The whole process seems to encourage use of coaching services and AI tools: https://spectrum.ieee.org/chatgpt-college-essays 
3. I tend to agree with AUT here - student success in tertiary education is determined mostly by factors outside of the institution's control. Setting arbitrary deadlines is pointless and smacks of desperation from clueless and lazy bureaucrats: https://www.nzherald.co.nz/kahu/universities-reluctant-to-set-deadlines-to-ensure-parity-for-maori-and-pacific-students/FNZX3ANE6ZAUZOQDKQTKGGXHF4/
4. An AI is no more a person than a cockatoo is a tea maid: https://www.technologyreview.com/2023/08/30/1078670/large-language-models-arent-people-lets-stop-testing-them-like-they-were/
5. If you don't want Meta to train its AI using your data, now you can tell them so: https://www.theregister.com/2023/08/31/meta_personal_data/
6. The generative AI skills that are in demand right now: https://www.datanami.com/2023/08/31/top-10-in-demand-genai-skills/
7. Is an AI model trained on others' work infringing copyright? https://www.theregister.com/2023/08/31/openai_class_action_fair_use/
8. Employees want guidance from their employers about using AI in their jobs: https://www.computerworld.com/article/3705373/workers-are-embracing-ai-but-want-more-guidance-on-how-to-use-it-says-survey.html
9. Data centres use a lot of electricity, both in running the servers and cooling them. AI can help reduce this: https://www.informationweek.com/sustainability/how-ai-can-make-data-centers-more-efficient-and-decarbonize
10. People don't trust AI. This can make it harder to integrate into businesses: https://www.computerworld.com/article/3705950/pew-research-finds-a-big-problem-with-ai-people-dont-trust-it.html
11. Using an AI to monitor Call of Duty voice chat for toxic behaviour: https://www.extremetech.com/gaming/call-of-duty-to-begin-using-ai-to-assess-voice-chat-for-toxicity
12. Large language model AI can be manipulated into being security threats: https://www.techrepublic.com/article/uks-ncsc-warns-against-cybersecurity-attacks-on-ai/
13. Watermarking AI generated images. Now if they can do the same with generated text, a lot of problems will be solved: https://www.extremetech.com/extreme/google-deepmind-tests-watermarking-ai-generated-images
14. Some techniques by which Machine Learning can detect heart disease: https://dataconomy.com/2023/09/01/effective-heart-disease-prediction-using-hybrid-machine-learning-techniques/
15. The ethics of #=AI are becoming more important: https://www.datanami.com/2023/08/31/ai-ethics-issues-will-not-go-away/
16. Of course, if AI are protecting networks from intrusion, it's only a matter of time before others start using them to break in: https://www.datasciencecentral.com/empowering-cyber-guardians-how-ai-is-changing-the-landscape-of-protection/