Notices by Eryk Salvaggio (cyberneticforests@assemblag.es)

I made intentional slop this week: an AI music mix tape. The goal is to examine, critically, the emotional charge of ChatGPT 4o’s “breakup texts” to users ahead of its shutdown. Rendered as AI pop music, the toxicity and manipulation of those who see 4o as a companion is clear. In the mix tape, I describe how the model created these texts as an interpolation between the system prompt and the user history. https://mail.cyberneticforests.com/chatgpt-breakup-mixtape/?ref=cybernetic-forests-newsletter

Finally deleted Twitter. It wasn’t making me smarter, was often infuriating, its architecture was designed to erase nuance, and its owner is an asshole. Feel a bit antsy about it, but the trick is to find other ways to do the things I thought it did. If you get value out of my newsletter and would like to recommend this account to your networks, I’d appreciate it!

Too often my aversion to social media drama has made me avoid any critique of certain digital artists that, in my view, peddled reductionist “humans are just machines” / “AI are our children” hooey that was ultimately a kind of 1990s prototescrealism

With the discovery of child abuse imagery within AI’s training data, my new piece for Tech Policy Press suggests that policymakers aren’t listening to the right people about the risks of generative AI. https://www.techpolicy.press/laion5b-stable-diffusion-and-the-original-sin-of-generative-ai/

I will never trust apps that don’t have save buttons. I know this is a me problem. Constant autosave is a smoother design decision. But it lacks the comfort of initiating and recieving feedback.

Lots of people assume that the words you type into a prompt window of a chatbot or image gen are the words the system sees and responds to. But that’s not always the case. Increasingly, companies are “shadow prompting” — and moving away from the transparency into AI systems that we need. I wrote about it for Tech Policy Press: #aipolicy #aiethics https://techpolicy.press/shining-a-light-on-shadow-prompting/

When people tell me generative AI will solve real-world problems

@pettter Critique of the marketing is vital for this very reason, in my mind. My fear is when people start out with the premise that Generative AI is “dreaming” instead of thinking about how models work and the political consequences of building things on top of them.

A lot of “thinking” about AI is based on AI’s marketing, rather than what AI does.

Arguably, generative AI is a speed bump for the imagination. Generative AI doesn’t behave like people, it behaves like a series of predictions for how people are likely to behave. An Internet of predictive entities has a political dimension, too: our online behavior is literally modeled for us. It doesn’t prevent new politics, but doesn’t encourage them, either. Generative AI gives us examples, not ideas.