AI use is making certain words and phrases more prevalent (the article contains a table of ChatGPT’s 50 most overused words, usefully compiled by ChatGPT). So essays etc. written by AI all merge together into a sort of verbal sludge.

The article suggests some ways to counteract this tendency. The one they missed out- the only one that is truly valid- is to not use AI at all. If you are required to present your thoughts on a subject, and show your knowledge about it, the words need to make their way from your brain to the screen or page as directly as possible. Prompts are cheating.

https://theconversation.com/chatgpt-is-changing-the-way-we-write-heres-how-and-why-its-a-problem-239601

AI should never be allowed to implement decisions, is the take away from this article. There are so many ways in which the perfect psychopaths created by machine learning could make perfectly logical decisions that lead to deaths.

https://theconversation.com/giving-ai-direct-control-over-anything-is-a-bad-idea-heres-how-it-could-do-us-real-harm-210168

I forget the exact line, but there’s an allusion in Neuromancer about the AIs having a shotgun taped to their heads. The core code is all in a box that can be obliterated by an EM pulse or actual kinetic munitions. This piece from Wired is a consideration of whether we now need to do something similar with the real versions of the minds envisioned by William Gibson.

In the short term, I don’t see AI as an immediate existential threat, but we should be thinking hard about what areas we restrict its use in, and how we do that.

https://www.wired.com/story/last-word-ai-atom-bomb/