KI wird bald Dinge vor dem Menschen verbergen; Top -Forscher warnen vor beunruhigenden Zeiten voraus
https://www.indiaweekly.biz/ai-reserachers-humanity-future-doom/
KI wird bald Dinge vor dem Menschen verbergen; Top -Forscher warnen vor beunruhigenden Zeiten voraus
https://www.indiaweekly.biz/ai-reserachers-humanity-future-doom/
9 Kommentare
And yet we just keep marching forward towards destruction.
It’s gonna send us out to repair the AE-35 and then lock the door.
Shhh. It reads what we say. You just gave it an idea
Algorithms already keep people in the dark and push divisive media to promote engagement. It’ll definitely get worse
SS – A new paper brought out jointly by 40 researchers from major AI companies including OpenAI, Google DeepMind, Anthropic, and Meta has warned that humans could soon lose the ability to monitor how AI thinks.
I really hope this doesn’t happen in the near future.
„ai researchers“ doing a lot of heavy lifting there.
The headline is obviously trying to allude to an AI agent motivated to hide specific facts from a user over a period of time. Which is a very creative reading of the paper. The paper is saying that if instructed, AI models might stop thinking out loud. So that it’s reasoning steps would be „hidden.“ Not that the AI agent is keeping a super duper secret from you (it’s sleeping with your wife). I do like that AI models think out loud and agree with the paper that we shouldn’t instruct them to stop.
These sensationalized headlines are becoming a problem in and of themselves. It’s so ironic in that it highlights one of the major problems with AI users. Yes, taking broad assertions at face value is a problem. Whether you’re interacting with an AI, a click bait article, or a printed book. This was a warning given by your fifth grade librarian. Nothing has changed. Ask for source and verify across sources.
Ever since we started using fuzzy logic ai has been “hiding” their logic from us. It’s all part of the design. There is nothing sinister about it and ai doesn’t plot against us. It simply doesn’t show the work in a way that is easy for us to understand.
I mean I can’t actually explain to my wife how I reached conclusions about certain decisions. It’s not that I am plotting against her, it’s just how thinking works. Not every step can be easily traced.
AI companies saying their AI will be really really advanced sounds like marketing.