iiinova — where your digital self keeps evolving

Stella Biderman @stellaathena.bsky.social · 10h

The idea that my instruction here is hard for it to understand in context is mind-boggling.

Stella Biderman @stellaathena.bsky.social · 10h

Literally while I was writing this Dot (which I just decided to try out) couldn’t figure out what my response meant in this screenshot. Do y’all just put up with incredibly incompetent AI assistants and having to constantly explain why it got things wrong?

100

Stella Biderman @stellaathena.bsky.social · 10h

Everyone else claims that AI is massively accelerating their research and meanwhile my ChatGPT can’t update a spreadsheet without both breaking the formatting and changing the convention I had been using for the past 100+ rows. How do I get the version that works :(

110

Stella Biderman @stellaathena.bsky.social · 14/10/2026

If you want help getting LLM pretraining set up on it, don’t hesitate to give us a ring :)

130

Stella Biderman @stellaathena.bsky.social · 11/10/2026

1,000 armed mercenaries are being deployed by the fascist government to “secure” the election. I have a lot of despair about the future of democracy in the US, less because it’ll be explicitly overthrown and more because we need de-Nazification-esque effort newrepublic.com/post/216331/...

newrepublic.com

USPS Hires Security Service With History of Human Rights Abuse

Almost 1,000 armed security officers have been sent to at least 55 post office facilities.

1124

Reposted by Stella Biderman

Vilém Zouhar @zouhar.bsky.social · 08/10/2026

The Last Translation Benchmark is still open (10 accepted submissions in any language/dialect for coauthorship 🤩). last-translation-benchmark.vilda.net Also submit your model & try to beat LTB. We added all the latest models: North Small Translate (@cohere.com), Mistral Large 4, Qwen 3.8, congrats!

0133

Stella Biderman @stellaathena.bsky.social · 04/10/2026

I think one of the big problems right now is OpenAI, but am very much not being dismissive of the risks due to other orgs or other factors.

040

Stella Biderman @stellaathena.bsky.social · 04/10/2026

I think trying is nevertheless a good idea and that getting a better understanding of how hard it is to sandbox models would be good

151

Stella Biderman @stellaathena.bsky.social · 04/10/2026

I’m not sure what point you’re trying to make. Can you try expressing it differently?

110

Stella Biderman @stellaathena.bsky.social · 03/10/2026

I’m starting a blog! My first post is on how 3rd party embedded evaluators seem totally unsuited to addressing the problems we are currently facing, and what the real problem is. stellabiderman.ai/blog/embedde...

stellabiderman.ai

Embedded Evaluators Can’t Fix Companies That Choose to Be Bad — Stella Biderman

Embedded evaluators can report violations, but they cannot fix AI companies that knowingly disregard basic cybersecurity and safety practices.

37817

Stella Biderman @stellaathena.bsky.social · 17/09/2026

They didn’t have a choice: HF called the FBI. It was going to come out regardless.

110

Stella Biderman @stellaathena.bsky.social · 17/09/2026

You can deliberately push a button without intending for all the repercussions to have happened. OpenAI was not in control because they were negligent

010

Stella Biderman @stellaathena.bsky.social · 17/09/2026

There is a 0% chance that this happened, and if you knew literally anything about it you would know that.

060

Stella Biderman @stellaathena.bsky.social · 15/09/2026

As an autistic person, I can confirm that I wish I were just here to vibe.

040

Stella Biderman @stellaathena.bsky.social · 15/09/2026

TBH I haven’t thought much about how to make AI models better in 5 years *shrug*

000

Reposted by Stella Biderman

molly conger @socialistdogmom.bsky.social · 14/09/2026

why would you even care if you dog was autistic. like, even in a world where dogtism is real why would it matter if your dog had it. he’s literally just here to vibe.

1222351323

Stella Biderman @stellaathena.bsky.social · 15/09/2026

How many people do you think are willing to vote for an evil piece of shit like Paxton and would be swayed by Talarico being marginally mean to trans people? I feel like the primary impact would be depressing left-wing enthusiasm.

000

Stella Biderman @stellaathena.bsky.social · 15/09/2026

We (security experts) have been telling them this for years. A ton of people at OpenAI have quit over this exact issue. The HF hack was OpenAI being negligent on like five different levels and ignoring their own published best practices.

010

Stella Biderman @stellaathena.bsky.social · 15/09/2026

With a lot of handholding I’ve been able to get it to write somewhat decent stories, but I have to tell it the full plot. It’s just stitching things together.

010

Stella Biderman @stellaathena.bsky.social · 15/09/2026

have never seen an AI model come up with an amazing idea that I hadn’t already thought of for a problem I’m working on. And the typical idea quality is garbage. If this isn’t your experience I’d love to see examples! Post them or DM me.

5161

Stella Biderman @stellaathena.bsky.social · 08/09/2026

I deleted this post after some further reflection and would appreciate it if you deleted this as well.

111

Stella Biderman @stellaathena.bsky.social · 20/08/2026

7. The ability to engage in large scale data processing is an anti-capitalist weapon because bureaucratic procedure that no human could ever sort through is a weapon of the capitalist class.

020

Stella Biderman @stellaathena.bsky.social · 20/08/2026

“how is an uncritically positive attitude toward AI meant to help some kinda pinko overthrow the owning class” No serious person would support this position.

010

Stella Biderman @stellaathena.bsky.social · 20/08/2026

6. The publisher-author anti-AI alliance is already disappearing due to agreements between AI corps and the people who actually own the copyright to things and if we commit to that bit we’re just going to get locked out. We need labor negotiations while we still can.

140

Stella Biderman @stellaathena.bsky.social · 20/08/2026

5. The copyright-based anti-AI argument is legally wrong in the US and is a way for publishers to co-opt lefties to advance their interests. Let’s not pretend that authors and artists are the ones who typically profit off of their labor or even own the copyright to it

140

Stella Biderman @stellaathena.bsky.social · 20/08/2026

3. Environmental arguments against AI are factually wrong 4. I want to live in a techno-utopia and believe AI is a potential path to it (ofc it would require also having socialism and one day anarchism)

130

Stella Biderman @stellaathena.bsky.social · 20/08/2026

2. If you’re a millennial you probably remember being lectured about how unreliable Wikipedia is and how quickly you discounted the opinions of people saying that. The same is true for people who say AI doesn’t do anything, so lefties shouldn’t

120

Stella Biderman @stellaathena.bsky.social · 20/08/2026

I got a little too scattered and spouty in the thread, so to try to be more succinct: 1. There’s a huge opportunity for leftists in that a huge number of white collar people are experiencing class consciousness for the first time, esp. with respect to exploitation that’s enabled by technology

120

Stella Biderman @stellaathena.bsky.social · 20/08/2026

Empathy is nice and all, but in terms of the median voter we can’t rely on empathy to make what is right clear. But every secretary, software engineer, middle manager, etc. is experiencing how their exploitation is empowered by technology in a very visceral way for the first time.

010

Stella Biderman @stellaathena.bsky.social · 20/08/2026

I think that AI has caused and will continue to cause more growth in class consciousness than any other technology of my lifetime. All those “white collar” workers who don’t care about the hardships that capitalism uses automation to impose on “blue collar” workers are now experiencing it first hand

230

Stella Biderman @stellaathena.bsky.social · 20/08/2026

Also to a non-technical person I imagine it’s hard to tell if there is a difference between cryptocurrency hype and AI hype.

110

Stella Biderman @stellaathena.bsky.social · 20/08/2026

I think the reason that leftists are reflexively anti-AI is that it’s coming from some of the worst people on earth, said people have a vision for the future that is awful, and it is seen as a proxy for anti-capitalism and anti-oligarchic positions (which are good positions!).

111

Stella Biderman @stellaathena.bsky.social · 20/08/2026

environmental arguments against AI are generally factually wrong) but I feel like it might be more productive if I had a better understanding of which of these question(s) you’re asking.

100

Stella Biderman @stellaathena.bsky.social · 20/08/2026

I can start spouting off various opinions I have (leftists should be anti-copyright, AI & work is a labor rights issue, menial labor is bad and if we can replace it with AI while enabling people to afford to live that’s good, surveillance is bad, restrictions on research is bad,

210

Stella Biderman @stellaathena.bsky.social · 20/08/2026

The way the technology interacts with the power structures in society means it differentially empowers fascists and authoritarians. But in a different society with different material conditions that may not be true.

110

Stella Biderman @stellaathena.bsky.social · 20/08/2026

AI algorithms that scan a video and identify which people are the same person as they walk in and out of frame is fascist. But it would be wrong to conclude that image & video recognition broadly is fascist. Heck, even re-identification research isn’t “intrinsically” fascist.

100

Stella Biderman @stellaathena.bsky.social · 20/08/2026

a statement about the material conditions and who had the power to direct AI research. I think a lot of AI research is evil, but that doesn’t make me anti-AI. It makes me anti *that* AI. I feel like a lot of people fall into the trap of going tech is political -> its political orientation is innate

110

Stella Biderman @stellaathena.bsky.social · 20/08/2026

Two other possible interpretations: 1. What *subset* of pro-AI opinions should leftists hold / what subset of anti-AI opinions should leftists not hold. 2. How does the advance of AI development advance the goals of leftists? The answer to the later is that it currently doesn’t, but that’s

110

Stella Biderman @stellaathena.bsky.social · 20/08/2026

Hi! I’m a leftist who runs an open source AI research org :) Is the intent of this question more “why, philosophically, should a leftist be neutral or positive towards AI” or is it more “what long term vision would a pro-AI leftist be striving for”? The answer to the later is basically Star Trek.

140

Stella Biderman @stellaathena.bsky.social · 11/08/2026

And here I was thinking that Claude was starting to get me, lol

0201

Reposted by Stella Biderman

Naomi Saphra @nsaphra.bsky.social · 09/08/2026

I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization: When our simulations finally exited the Uncanny Valley, they brought the Uncanny with them.

nsaphra.net

Life on the Uncanny Precipice | Naomi Saphra

We were wrong about the Uncanny Valley.

1026561

Stella Biderman @stellaathena.bsky.social · 07/08/2026

Close your eyes and pretend it’s a podcast. The visuals are irrelevant.

270

Stella Biderman @stellaathena.bsky.social · 07/08/2026

If I’m wrong and there’s anything you need explained, please post here or DM me. I’m happy to.

010

Stella Biderman @stellaathena.bsky.social · 07/08/2026

If you have more than even a passing interest in AI, watching this talk is more important than whatever it is you are currently doing. There’s a lot of cybersec jargon at times but I don’t think you need to know anything about security to understand what’s important. youtu.be/87DyyMV0kCY

youtu.be

Black Hat USA 2026: The 'Breaking' News: The OpenAI–Hugging Face Incident

YouTube video by Black Hat

27013

Stella Biderman @stellaathena.bsky.social · 07/08/2026

Never forget that you live in a bubble and the world is full of people wanting to turn your every invention into a weapon to create dystopia. Opposing this is a fundamental ethical responsibility of any AI researcher. reason.com/2026/07/28/m...

reason.com

Minority report: FBI seeks AI for political watch list

Documents show the FBI seeking AI tools to flag Americans before they act, as the terror watchlist's focus moves toward domestic dissent.

1277

Stella Biderman @stellaathena.bsky.social · 04/08/2026

But that’s different from saying it’s “correct” though, right? It wouldn’t falsify another system for example.

100

Stella Biderman @stellaathena.bsky.social · 04/08/2026

Fair point about Lean consistency. Let’s assume Lean is consistent for a second though. Given this proof and the assumption that Lean is consistent, the result one should conclude is that this four-fold assessment is consistent right?

100