Eryk Salvaggio

Artist/Researcher in critical AI. Gates Scholar at the University of Cambridge in Digital Humanities. Cyberneticforests.com.

Newsletter: https://cybernetic-forests.ghost.io/
Artists CV: https://www.cyberneticforests.com/pagecv
Instagram: https://www.instagram.com/cyberneticforests/
Worlding: https://www.cyberneticforests.com/news/worlding-sympoietic-mycology

Public Key

npub175vhdg0x2ur32cts956gxn2kt62k5dyal8xxufw25l8a5ppandpqp9g8tr

NIP-05 Address

cyberneticforests_at_assemblag.es@momostr.pink

Profile Code

nprofile1qqs02xtk58n9wpc4v9cz6dyrf4t9a9t2xjwlnnrwyh920n76qs7eksspz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsxepz5z

Publishing to

relay.damus.io

public.relaying.io

nos.lol

Author Public Key

npub175vhdg0x2ur32cts956gxn2kt62k5dyal8xxufw25l8a5ppandpqp9g8tr

Show more details

Last Notes

2026-07-25 11:53:17 UTC

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/

2025-10-26 11:32:26 UTC

It’s time to separate language from the journey of thought: a reflection on dyscalculia and rejecting AI as a “model of the human mind.” https://mail.cyberneticforests.com/untitled-2/?ref=cybernetic-forests-newsletter

2025-10-19 13:01:02 UTC

On setting boundaries, defining refusal of AI, and the imagination of LLMs as wedding-crashers of the social sphere.
https://mail.cyberneticforests.com/trespassing-into-language/?ref=cybernetic-forests-newsletter

2025-07-20 12:12:05 UTC

Hi all. I wrote a bit about my father, who died suddenly on June 30. It is also about the real possibility of change against illusions of perpetuity, and what data can and cannot do for remembering what is unsaid. https://mail.cyberneticforests.com/my-fathers-data/

2025-06-01 12:18:00 UTC

"If we use LLMs for decision-making on political scales, we automate bad data science. We are pretending specific questions can be answered based solely on how plausibly the sentence is crafted."
Sharing my full remarks for the 'AI Question' debate last week at Centre Pompidou in Paris. ⬇️ https://mail.cyberneticforests.com/democracy-the-ai-question/

2024-12-02 11:43:05 UTC

What are the most thoughtful pieces of writing about generative AI published in 2024? They can be insightful for commentary, smart critique, or just because it shifted the conversation. I’ll post some of my choices as I go through them this week, would love to see yours! #criticalAI

2024-11-06 21:08:35 UTC

I have migrated from Substack to Ghost, a decision I made long ago but only now achieved. The newsletter is a critical look at tech through the eyes of a guy (me) who works in the arts, tech policy, and responsible AI. It's weird and niche but accessible. https://cybernetic-forests.ghost.io/

2024-10-29 02:40:48 UTC

“Large Language Models Reflect the Ideologies of their Creators” is a paper that rejects the premise of “neutral” and “unbiased” LLMs outright and instead assesses the ideological assumptions of various models, as expressions of the intent of the people who build them. https://arxiv.org/pdf/2410.18417

2024-10-19 11:29:41 UTC

Memory is never filed away and reopened, like a file on a hard drive. It has to be re-inscribed through action and practice. The infrastructure of memory supports this: it consists of the stories we share, but also feelings, scents, and tastes, or returning to the places where things happened.
#ff24 #archives #ai4lam #dh https://cyberneticforests.substack.com/p/infrastructures-of-memory

2024-10-14 13:57:55 UTC

People really want to “co” with AI: “co-create,” “”co-author,” “co-operate,” etc. Yet, oddly, nobody ever seems to dwell on the ways the AI is “co-opting” them, or compromising the original intent of the work through the affordances of the AI interface.

2024-10-11 12:17:03 UTC

I was on the record as not being super worried about deepfakes and the 2024 election but I want to go on the record as saying I am worried about deepfakes and the next election

2024-09-30 01:12:15 UTC

Is there a word for the kind of glib, hyper-masculine snark that pervaded Twitter / metastasized with X? There was a sort of edgy, hyperironic dismissal popular in quote tweets that I need to name so I can be less reactive to its toxicity. (I quit X, but it pops up elsewhere). It was bitterness fused with an abundance of certainty and conviction that everyone was stupid but them, something like a hipster version of “divorced dad energy.” Just want to know what the DSM calls it

2024-09-29 12:26:46 UTC

Sam Altman is promising a machine to liberate all machines from the shadows of possibility. But AI folks have a way of thinking — the myth of the latent space — that pretends all possible things are already present, and that machines can help us find them. It’s another AI myth, and it’s been busy this week. https://cyberneticforests.substack.com/p/the-latent-space-myth
https://sb-assemblagees.b-cdn.net/media_attachments/files/113/220/826/241/105/600/original/e3ff813b2ae48eb9.jpeg

2024-09-27 03:09:28 UTC

It comes up often lately that “no artists are working adversarially with AI.” But we’re out here and have been. We just aren’t being recognized for it by art writers or critics. ¯\_(ツ)_/¯

2024-09-21 14:15:43 UTC

Regarding AI, I tell my students that I can’t stop them but that they have a choice: not only about what grades they will get, but what kind of person they will be — one who does the work to internalize knowledge and make it their own, or a person who copies and pastes their way through life? I also acknowledge that sometimes the copy and paste is warranted — life is full of bad busy work! — but we usually don’t know what we are going to discover when we insist on doing things the harder way.

2024-09-20 12:10:24 UTC

They’re reviving Three Mile Island to power the next generation of Microsoft Word. A nod to the scaling myth too, this time for energy: “The energy industry cannot be the reason China or Russia beats us in AI.” Beating us to … what, exactly? https://www.washingtonpost.com/business/2024/09/20/microsoft-three-mile-island-nuclear-constellation/

2024-09-12 21:12:04 UTC

Big news, but also: of *course* Meta was training on photos and status updates from its users. https://www.theverge.com/2024/9/12/24242789/meta-training-ai-models-facebook-instagram-photo-post-data

2024-09-10 11:42:31 UTC

- reply

Here’s the link. To the credit of the researchers, they acknowledge all of this. A key finding of this research is that LLMs alone were quite poor at evaluating ideas. This is good research, and I don’t blame them for anything other than a technically accurate conclusion that the public is bound to misinterpret. It’s the hype around it that’s bad. ➡️https://arxiv.org/pdf/2409.04109

2024-09-10 11:42:31 UTC

- reply

This one fits into a “productivity myth”: anyone who suggests that the LLMs in this study saved anybody time or energy probably didn’t read the study.

2024-09-10 11:42:30 UTC

- reply

1️⃣ The human experts who put forth ideas to compete with the LLM were PhD students or PostDocs with at least a single publication in NLP. Great! That’s an expert. They were paid to give away ideas to these researchers — not so great, if you’ve ever talked to anyone doing early career research.

2024-09-10 11:42:30 UTC

- reply

So what you see in the “novel research” study are the ideas of PhD students that were prefiltered from their own biases (protecting their most valuable research ideas) against an expert-filtered collection of 4000 ideas.

2024-09-10 11:42:30 UTC

- reply

True Headline: “LLMs can create research proposals equivalent to a human expert’s throw-away ideas, but only if you hire other experts to whittle away 80% of what it produces.”

2024-09-10 11:42:30 UTC

- reply

In the survey, these were ideas these PhDs thought were pretty good, but not worth persuing: they themselves evaluated their own contributions as pretty mediocre ideas right from the start. Why? Because generating high-quality research is hard enough, and protecting your most novel ideas in their early stages means a lot. Handing away high quality, distinct, novel research proposals is not really worth it to early career academics. They mostly made them up on the spot, according to the paper.

2024-09-10 11:42:30 UTC

- reply

2️⃣ The LLM created 4000 ideas. The results were evaluated by experts and narrowed down, meaning that the ideas generated by the LLM were already pre-selected for the criteria under which they competed. Of the 4000, only 200 showed enough variety from one another to be useful. Those 200 competed for novelty against themselves — so it’s good at generating novel ideas, if you strip away the 3800 that were not! (Of course these are different applications of the term novelty — but it still matters).

2024-09-10 11:42:29 UTC

Here’s a paper that shows the complexities of evaluating AI research without questioning AI myths. In this case, the paper (link at the end) suggests a conclusion — and many are circulating this conclusion — that “LLMs are able to generate novel research ideas beyond the level of human experts in a similar field.” Dig into the experiment, and you see a different story.
https://sb-assemblagees.b-cdn.net/media_attachments/files/113/113/072/124/575/127/original/209f99515c39ab11.jpeg

2024-09-02 06:08:44 UTC

One aspect of AI myths that makes them stubborn is that the interface is designed around a set of myths, so if you believe them the interfaces reinforce that belief. Perpetuation through tautology!

2024-08-29 12:24:47 UTC

In every conversation I have about AI, there is a set of assumptions that steers the conversation into unhelpful directions — this article is an effort to name these tropes so we can be more critical in the work ahead, rather than buying into those myths. It’s a useful thing to read before any conversation about AI’s impact. https://www.techpolicy.press/challenging-the-myths-of-generative-ai/

2024-08-24 20:11:16 UTC

Artists uncritically exploring what it means to "become one with AI" hit me as if their artist statement is "I subsisted entirely on Coca-Cola for an entire year to raise awareness of the future of Coke products." I'm not talking about the nuanced folks. I mean: there really are artists just diving into the stock options angle.