Weekly Review 1 November 2024

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

1. I think this is the biggest reason to not use AI to generate important code or material-it's too easy for bad actors to inject malicious code into the model used: https://arstechnica.com/tech-policy/2024/10/bytedance-intern-fired-for-planting-malicious-code-in-ai-models/
2. Google's AI mediator, that helps guide people to agree: https://arstechnica.com/ai/2024/10/googles-deepmind-is-building-an-ai-to-keep-us-from-hating-each-other/
3. The quality of data being used to train AI is declining. Garbage in, garbage out: https://www.bigdatawire.com/2024/10/23/ai-has-a-data-problem-appen-report-says/
4. Like many other AI, this transcription tool hallucinates: https://techcrunch.com/2024/10/26/openais-whisper-transcription-tool-has-hallucination-issues-researchers-say/
5. It is going to take some time to sort out the legal issues around the scraping of content to train AI: https://www.theguardian.com/technology/2024/oct/25/unjust-threat-murdoch-and-artists-align-in-fight-over-ai-content-scraping
6. More ways AI will keep lawyers happy-who's responsible when an AI controlled vehicle crashes? https://dataconomy.com/2024/10/23/the-ethical-dilemmas-of-autonomous-cars-whos-responsible-in-a-crash/
7. Ten Python libraries you should be familiar with for working with data: https://www.kdnuggets.com/10-essential-python-libraries-for-data-science-in-2024
8. The last time I got a scam call I told them to talk to my d*ck and put the phone down the front of my trousers. I don't think that would work if it were an AI calling: https://www.theregister.com/2024/10/24/openai_realtime_api_phone_scam/
9. An AI that can write, and verify, code: https://techcrunch.com/2024/10/24/anthropics-ai-can-now-run-and-write-code/
10. Biased data produces biased AI models. This is as true for cybersecurity applications of AI as it is for anything else: https://www.datasciencecentral.com/why-ai-bias-is-a-cybersecurity-risk-and-how-to-address-it/
11. Do we really want an AI to be able to control the mouse on our computers? Maybe useful for people who have motor impairments or tremors: https://arstechnica.com/ai/2024/10/anthropic-publicly-releases-ai-tool-that-can-take-over-the-users-mouse-cursor/
12. AI company fires back at lawsuits over its scraping of content for training data: https://techcrunch.com/2024/10/24/they-wish-this-technology-didnt-exist-perplexity-responds-to-news-corps-lawsuit/
13. Did a chatbot AI really encourage a teenager to kill themselves? Time for guardrails: https://www.theguardian.com/technology/2024/oct/23/character-ai-chatbot-sewell-setzer-death
14. Using AI to enable a garden to talk back: https://www.theguardian.com/lifeandstyle/2024/oct/25/ai-powered-garden-chelsea-flower-show
15. The idea of multi-agent architectures has been around for decades. Will generative AI be able to coordinate different agents to perform useful tasks? https://www.informationweek.com/machine-learning-ai/10-reasons-why-multi-agent-architectures-will-supercharge-ai
16. It looks pretty obvious to me that some companies will try to use Microsoft's AI to replace workers, not augment them: https://dataconomy.com/2024/10/23/microsoft-rolls-out-virtual-employee-ai-agents-for-enterprises/
17. The US wants to use more AI, especially in national security: https://www.computerworld.com/article/3587124/white-house-tells-intelligence-agencies-use-more-ai.html
18. The AI Cisco is using for customer support: https://www.computerworld.com/article/3578806/ciscos-new-ai-agents-and-assistants-aim-to-ease-customer-service-headaches.html
19. If the AI chips only last three years, what happens to them after that? Can they be recycled, or is this another way AI can negatively impact the environment? https://www.extremetech.com/computing/data-center-ai-gpus-may-have-extremely-short-lifespans
20. AI generated material is a threat to us, especially its use in election interference: https://www.informationweek.com/cyber-resilience/ai-manipulation-threatens-the-bonds-of-our-digital-world
21. An approach to watermarking AI generated text: https://spectrum.ieee.org/watermark
22. Replacing journalists with AI is not a popular move: https://www.stuff.co.nz/world-news/360462671/polish-radio-station-replaces-journalists-ai-presenters
23. 40 years later, Terminator continues to influence people's opinions of AI: https://arstechnica.com/ai/2024/10/40-years-later-the-terminator-still-shapes-our-view-of-ai/
24. Who needs AI safety? Not OpenAI: https://www.theregister.com/2024/10/25/open_ai_readiness_advisor_leaves/