Olivia Guest · Ολίβια Γκεστ

The mirage of AI deregulation One of the most interventionist approaches to technology governance in the United States in a generation has cloaked itself in the language of deregulation. In early December 2025, President Donald Tr...

🧵 Trump administration AI policy is widely described as deregulatory. This description is misleading. What's happening is not the absence of governance but its rearrangement--intensive state intervention operating through mechanisms we don't typically call regulation. www.science.org/doi/10.1126/...

if you're asking if i think they're a technology that is actively harmful to humans and the global environment and we should be restructuring things to exclude them? yes

Them: it's useful

Us: yeah but it's inherently evil. And how useful is it exactly?

Them: moderately to mildly. But it's useful.

Us: want to think about the inherently evil part?

Them: ...but it's getting more useful!

robot in beret painting a robot Mona Lisa

AI works can’t be copyrighted or patented in the US

even if you think your robot is alive

www.youtube.com/watch?v=JoNr... - video
pivottoai.libsyn.com/20260304-ai-... - podcast

time: 6 min 25 sec

never, ever, ever, ever accept "how will you pay for it?" as an argument against social programs.

It's based on this work as you'll see, and the specific extract here might be useful too

bsky.app/profile/oliv...

It's based on this work as you'll see, and the specific extract here might be useful too

bsky.app/profile/oliv...

We've been here before! Parallels between AI and tobacco, and other warnings.

Bingo!

olivia.science/before

Adam Tickell closed the department I did my degree in

my main concern is the harm rather than the output, so i don't really need to hear a product testimonial from someone who found it useful. because i don't care

very concise, thank you.

as a technologist i do in fact think it's helpful to disambiguate the homogenous term "ai" into types of predictive models, because some predictive models are not rooted in social harm.

LLMs are not in that category

The point of this thread is that i am uninterested in their usefulness in their stated tasks.

However I will add that I think there is another layer besides moral judgment and task utility, which is political utility.

i think these tools provide enormous political utility to fascists

And insofar as opposing fascism is not merely a moral opinion but a material project, building my practices around tools controlled by my enemies is not politically useful, or strategically wise, or morally correct

Thank you. Here is a summary of my views. I don't have anything else to add. Be well.

*trollface*

Language is a universal approximator and yet they still cannot find the right words to convince me if their nonsense 🫠

Yup!

> I forbid use of my words, code, or images for “training” their artificial neural network models. Such actions are in violation of the licenses outlined above as they require correct attribution.

olivia.science/license/

Which may seem like a small point, but it does mean that there should be zero overlap between supporters of FOSS and users of Claude, because that would be completely unethical.

And yet, and yet.

I'm very sure that a lot of people put code online under very generous licence terms asking only that users of the code should acknowledge it and include the licence; and the coding LLMs scraped it all up and are using it without honouring the licence.

👇

Our new Read to Respond reading list, "Critical Perspectives on AI in the Humanities," brings together books, journal articles, and issues that examine the cultural, ethical, political, and intellectual dimensions of AI.

Access the content for free: buff.ly/qJR0MXU

Do you mean alternative to LLMs or ? @irisvanrooij.bsky.social has work on why this is intractable if you're curious tho

theyre so brave.

bravely gambling, bravely hitting refresh, bravely laying off, bravely implementing ideology as workflow, stunning minds

Every time I see "universal function approximator" used because they think it means "can approximate what neurons do" I want to walk into the fuckin ocean and let the current take me

Post-hoc checks will always have gaps. The pattern you didn't think of is the one that gets through. Curious what the alternative looks like though, pre-generation filtering or something else entirely?

Grammarly Offering Manuscript Reviews by AI Versions of Recently Deceased Professors The Grammarly "Expert Review" feature uses AI to provide feedback on papers using the name and work of real professors, dead or alive.

Daily reminder that calling ai dead labor and stolen labor is literal.

My I-know-controversial view is that, if one has a "bsky.team" handle, one should not (1) be making posts like this, and (2) removing quote posts.

The employees of this platform will be its downfall.

Plan accordingly.

Truly 😭

You’ll be tapping until your finger bleeds and falls off.

It’s fascinating that so many in academia don’t understand what academia is on the most basic level.

I also read this which is interesting - but I’m still not sure what to think about it. bsky.app/profile/oliv...