Eryk Salvaggio

Regarding OpenAI’s “hack” of Hugging Face: Seeding the “maybe our model is exhibiting self-awareness?” narrative to obliterate the “OpenAI is a negligent and potentially criminal actor” narrative is a reputation management move exclusive to this industry. Not a stunt, but a desperate need to control the narrative. https://www.reuters.com/business/its-ai-agent-spent-days-hacking-company-sources-say-openai-did-not-notice-week-2026-07-24/

Where does the “it’s not x, it’s y” come from? It’s not only training data: LLMs are trained to use language to arrive at “answers.” Whatever we’re incentivizing and disincentivizing in our use of language when we resist “sounding like AI” has big implications for students, and for human thought. New from me: https://mail.cyberneticforests.com/its-not-just-data-its-post-training/?ref=cybernetic-forests-newsletter

@timnitGebru@dair-community.social thank you!

Seven Warnings for Critical Agentic Design: For Designers, Who Will Decide For Themselves (and for the Rest of Us): I gave this very short talk at UC Berkeley, aimed at developing a critical lens on agentic LLMs for the designers, faculty and students who are already predisposed to use them. https://mail.cyberneticforests.com/toward-a-critical-agentic-systems-design-practice/

LLMs produce something language-like, a "languagicity" that takes on the shape and feel of language and even many of its functions, but lack the social context that makes some forms of language feel "real." https://mail.cyberneticforests.com/modeling-language-with-plaster/

To acknowledge the ways AI systems are changing does not buy into hype, it sharpens the precision of critical thinking about their impacts. To make sense of what AI does to people, you also need to understand what it does *for* them. Wrote for Tech Policy Press this week: https://www.techpolicy.press/stochastic-flocks-and-the-critical-problem-of-useful-ai/

Please to share that you can now stream my short film, “Human Movie,” with my own narration taken from its second ever performance at the AIxDESIGN Slow AI festival in Amsterdam last summer, including the full 35 minute film and the audience discussion. https://mail.cyberneticforests.com/now-streaming-human-movie/

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

Here’s the talk I gave at the Jeu de Paume today here in Paris, all about noise and constraint in AI systems — this is the French version shared with the audience, I’ll share the English version soon. https://www.cyberneticforests.com/bruite