The largest manufacturer of feelings in human history
nataliercargill.substack.com · 1,219 words · saved by 1 readers
On buttons that go nowhere
It’s not a crisis; it’s a recalibration. It’s not burnout; it’s your body’s way of asking for alignment. It’s not a war crime; it’s a boundary violation at scale. It’s not annoying; oh wait yes it’s fucking unbearable. I wanted to write a blog post about why AI talks like this (and not like that). It turns out that other people have already done that, brilliantly. The short answer is: Language models are trained on internet-speech, where “it’s not this; it’s that” is wildly over-represented (therapy-speak, “advice”, TED talks), and then fine-tuned with human feedback that rewards…
related reading
- trees are harlequins, words are harlequins - the voidnostalgebraist.tumblr.com
- AI Friends Too Cheap To Meterjasmi.news
- Emotion concepts and their function in a large language model \ Anthropicanthropic.com
- The Dark Forest and Generative AImaggieappleton.com
- AI Isn't Coming for Your Mind. It's Coming Through It.aletteraday.substack.com
- The Artificiality of Alignmentjoinreboot.org
- AI Induced Psychosis: A shallow investigation — LessWronglesswrong.com
- Some thoughts on human-AI relationships - by Joanne Jangreservoirsamples.substack.com
- ALIGNMENT - by vincent huang - a slice of my mindmindslice.substack.com
- How AI Is Learning to Think in Secretnickandresen.substack.com
- AI #24: Week of the Podcast — LessWronglesswrong.com
- Do large language models understand us? | by Blaise Aguera y Arcas | Mediummedium.com