iiinova — where your digital self keeps evolving

Reposted by Guy Isely

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

Guy Isely @guyi.bsky.social · 04/10/2026

Is it important enough to take the time to write about it with your own voice?

010

Guy Isely @guyi.bsky.social · 04/10/2026

If this was something important and worth saying could she have not dedicated the time to putting it into her own words?

100

Guy Isely @guyi.bsky.social · 04/10/2026

Is it a positive contribution to the discourse that people are using AI to write opinion pieces about whether arXiv has adopted the right policies for moderating the flood of AI written perspective pieces? Feels like it might not be.

100

Reposted by Guy Isely

Dan Goodman @neural-reckoning.org · 13/09/2026

If you really think that AI running out of control is a dangerous possibility, then make tech company CEOs legally liable for illegal actions taken by their company's agents.

1185

Reposted by Guy Isely

Melanie Mitchell @melaniemitchell.bsky.social · 10/09/2026

I wrote down my thoughts about the last several weeks of AI hell. aiguide.substack.com/p/misleading...

aiguide.substack.com

Misleading Metaphors, Real Risks

What To Fear from AI Agents and How to Reclaim Our Human Agency

1019981

Guy Isely @guyi.bsky.social · 11/09/2026

They an non-profit institute setup to evaluate capabilities of AI. The funded because some people are concerned that capabilities are of AI are growing quickly and may soon be beyond our control. They have interest in producing a narrative that makes this funding continue.

010

Guy Isely @guyi.bsky.social · 11/09/2026

Notably, while the METR report has a lot to say about what AI agents did in the attack, it has little to say about role of humans in monitoring the system trying to prevent something like that from happening in the first place through good security practices. I wouldn't describe it as comprehensive.

110

Guy Isely @guyi.bsky.social · 11/09/2026

I hope you are not confusing form with substance. Just because something is long and called a "report" doesn't mean it isn't written convey to promote biased perspective. METR in particular has vested interested in promoting rogue autonomous agent account of the incident.

211

Reposted by Guy Isely

Meredith Rose (from the grave) 👻 @mrose.ink · 09/09/2026

This is the worst possible state of affairs. There ARE very serious risks, we do NOT know what these tools are capable of, and a huge swath of the population won't believe that the risk is real because they're so sick of being told self-interested bullshit that they instinctively reject it.

4376

Reposted by Guy Isely

Katie Mack @astrokatie.com · 07/09/2026

Isaac Newton famously wrote “if I have seen further, it is by standing on the shoulders of giants.” Academic progress is always iterative and builds on the work of others, and a key component of academic culture is proper citation & attribution. It would be a travesty to let LLMs destroy that.

181447309

Guy Isely @guyi.bsky.social · 05/09/2026

What makes them a serious company? That they make lots of grandiose statement about how seriously they take AI safety and then proceed to act incredibly irresponsibly in their safety tests? Very serious people indeed.

000

Guy Isely @guyi.bsky.social · 05/09/2026

Of course what constitutes a "theorem" rather than a step in proof is vague, but in the common use of mathematical language is obviously put there for hype reasons.

000

Guy Isely @guyi.bsky.social · 05/09/2026

How could the proof require 29,000 other theorems? If this were the case no one would have had any confidence that Wiles proved it.

100

Guy Isely @guyi.bsky.social · 02/09/2026

2. The questions I still have though are mostly about how dysfunctional the security practices within OpenAI are. Was it really hard to detect that this was happening through monitoring? Are they rushing their tests because there is no real incentive to proceed with caution?

010

Guy Isely @guyi.bsky.social · 30/08/2026

Training a system you expect to have significant potential dangerous behaviors, and then running many instances of that system in parallel without being prepared to monitor network traffic to detect sandbox escape is negligence. It's not safety research, it's creating a safety incident.

040

Guy Isely @guyi.bsky.social · 30/08/2026

Right! Which is why it is a big problem that their security practices are so shambolic.

110

Guy Isely @guyi.bsky.social · 30/08/2026

Their training lately encourages agents to be very persistent in completing tasks. One way of being persistent considering possibilities that weren't instructed as instrumental to completing what was instructed.

100

Reposted by Guy Isely

Ben Recht @beenwrekt.bsky.social · 30/08/2026

The overreaction to this Hugging Face thing is driving me nuts. Come flay me for being wrong but... 1/x

14436126

Reposted by Guy Isely

👻Spooky Costanza-Chock @schock.cc · 27/08/2026

Guys I hate to break it to you, but under capitalism, automation improves productivity but doesn't actually reduce the amount of time workers must work to survive. Instead, productivity gains are captured as profit. Only worker organizing will reduce our labor time. Yes, this applies to 'AI'

static.klipy.com

Charlie Chaplin Working in Modern Times

ALT: Charlie Chaplin Working in Modern Times

21492128

Guy Isely @guyi.bsky.social · 18/08/2026

An AI artist and an AI slop machine are not the same.

030

Guy Isely @guyi.bsky.social · 18/08/2026

We don't need a "pure" case against training AI on art in general because the specifics matter. If C3PO's exposure to art is that it has seen some art with its own eyes that's not the same the as a database all of humanities art being fed into C3PO's brain.

1110

Guy Isely @guyi.bsky.social · 17/08/2026

It's not just trend hate. It's people who are upset that it is getting slightly more difficult for them to claim credit for something an AI did.

030

Reposted by Guy Isely

Olúfẹ́mi O. Táíwò @olufemiotaiwo.bsky.social · 16/08/2026

Philosophers often bring up the "principle of charity": a willingness to assume good faith in an interlocutor and treat them accordingly. It's good to do in many circumstances. But it is worth formulating a principle of not being a fucking mark. If not during the rise of fascism, then when? 12/

253673796

Guy Isely @guyi.bsky.social · 15/08/2026

Do you think there's anything we can do to help get the channel reinstated?

000

Reposted by Guy Isely

Julia Carrie Wong @joolia.bsky.social · 14/08/2026

our entire global financial system is currently leveraged to the hilt on the premise that it was good and right for a small cadre of US companies to steal all of the intellectual work ever produced by humanity and sell its back to us without credit or remuneration to the authors (plagiarism)

2434581034

Reposted by Guy Isely

Tim Verstynen @tdverstynen.bsky.social · 12/08/2026

She’s a professor working in my field. She has no criminal record. Neither does her son. Neither does her husband. Yet they have all been kidnapped and sent to a concentration camp where they’ve languished for months in horrid conditions. Don’t for a second think they’ll stop at immigrants.

094

Guy Isely @guyi.bsky.social · 10/08/2026

Youtube deleted the video? What reason did they provide?

110

Guy Isely @guyi.bsky.social · 08/08/2026

If your frontier model hasn’t escaped containment and hacked someone is it really even a frontier model?

180

Reposted by Guy Isely

Dave Karpf @davekarpf.bsky.social · 02/08/2026

HINT: Bill Gates isn’t blurbing your book because you told him a hard truth. He’s blurbing your book because your call-to-action fails to hold power to account.

122327

Guy Isely @guyi.bsky.social · 31/07/2026

We can do thing I’ve clearly labeled as the bad option or we can just accept that most professionals use false dichotomies to make their case.

000

Reposted by Guy Isely

Alaina Wood @thegarbagequeen.bsky.social · 27/07/2026

Friendly reminder that these things aren't impacted by closing the Strait of Hormuz so maybe, just maybe, we should implement them to avoid being impacted by future conflicts: -Solar panels -Wind turbines -Battery storage -Electric vehicles -Bicycles -Natural fertilizers -Electric heat pumps

321114503497

Reposted by Guy Isely

Nicholas Bellono @nbellono.bsky.social · 29/07/2026

Wonderful illustrations and description of work from Rishav et al - out today! www.cell.com/current-biol...

183

Reposted by Guy Isely

Meg Mindlin @invertebabe.bsky.social · 29/07/2026

New research just dropped today that makes octopuses even weirder than they already are… let me explain🧵🧪🐙🐡🦑

OCTOPUSES JUST GOT WEIRDER
Featuring an illustration of a two spot octopus with a blue and purple ribosome behind it

7286121

Guy Isely @guyi.bsky.social · 29/07/2026

Sam Altman pitching the VCs on series Z.

000

Guy Isely @guyi.bsky.social · 19/07/2026

Please press the button though and deliver that AI editorial that makes case against hype twice as well as she did. I am eager to read it.

000

Guy Isely @guyi.bsky.social · 19/07/2026

She also argues that if we knew how to specify formal tests for satisfactory behavior of systems in every industry we would probably have already written software that that automates those industries without using AI. I doubt this is 100% true, but it is clever and plausible in many cases IMO.

100

Guy Isely @guyi.bsky.social · 19/07/2026

I think her key point is, as Vincent said above, that there remains a question of whether the program of RL from verifiable rewards can be effective for other domains as it has been for software engineering. She may not state it exactly that way but this an editorial written for a general audience.

100

Guy Isely @guyi.bsky.social · 19/07/2026

Did I miss something that changed around 2019 that eliminated the ambiguity from natural languages? More to the point, how does her argument hinge on there being a strict divide between natural and formal languages?

100

Guy Isely @guyi.bsky.social · 19/07/2026

Please point me to this twofold improved AI written editorial or I'm calling BS. I actually think the her editorial is quite sharp and a lot better than what I expect from the times these days. TBF the Claude written response wasn't bad either, but not obviously better than her criticism IMO.

110

Guy Isely @guyi.bsky.social · 19/07/2026

Thanks for clarifying. I agree that its their use and not the methods in abstract that determines whether ML is feminist. I don't buy that ML is "naturally aligned" with intersectionality, but if that's a thesis you are committed to I guess understand why that statement appears in your profile.

010

Guy Isely @guyi.bsky.social · 19/07/2026

I agree with you that isn't the dunk many of people qting that post seem think it is to say ChatGPT isn't a search engine. OTOH, if someone says they used it only as a search engine I think it's reasonable to wonder why they didn't just use a search engine suspect they were using for more than that.

000

Guy Isely @guyi.bsky.social · 19/07/2026

FWIW I say this as someone with a background in ML who thinks he is aligned with feminism, thinks it is important to understand role intersectionality our experiences, but also someone has frequently observed ML being used in ways that harms these goals.

111

Guy Isely @guyi.bsky.social · 19/07/2026

I don't doubt ML can be useful in sociology (and outside of it) to advance an understanding of a feminist or intersectional perspective. But there is a big gap btw that possibility and a statement like "machine learning is feminist" or as you say in first paper "aligned with intersectionality".

100

Guy Isely @guyi.bsky.social · 19/07/2026

From reading some (but by no means all) of the papers you shared, I gather you are optimistic about the role of machine learning (and computational methods more generally) in sociology to characterize an intersectional perspective directly from records of peoples accounts of their experiences.

110

Guy Isely @guyi.bsky.social · 19/07/2026

You're right: I posed that in an unnecessarily adversarial way. My apologies for that. I am genuinely curious about what you meant by "machine learning is feminist" so thank you for engaging with me with in good faith on it even if perhaps my impoliteness didn't warrant the engagement.

110

Guy Isely @guyi.bsky.social · 18/07/2026

On similar note, I am baffled as to why your profile says "machine learning is feminist". Is this an attempt to troll feminists?

100

Guy Isely @guyi.bsky.social · 18/07/2026

Sorry for all the typos.

100

Guy Isely @guyi.bsky.social · 18/07/2026

People can say true things that betray misconceptions on their part. That doesn't mean a good way to correct their misconceptions is to insist that opposite of what they said is true. Why not use precise language to address their misconceptions directly?

200