Sabitlenmiş Tweet
I wrote an essay about the recent embrace of location-sharing apps + why it should really, like, really, freak you out—new in @thedrift_mag
thedriftmag.com/on-the-grid/
English
Zoë Hitzig
89 posts
Zoë Hitzig retweetledi
Extremely important work:
Erik Brynjolfsson@erikbryn
The @nytimes piece today by @ByrneEdsal13590 highlights a concern I share: “If we stay on the current path, the risk of extreme concentration — both economic and political — is very real.” In work with @zhitzig, we ask why AI may shift the balance between dispersed knowledge and centralized control.
English
Zoë Hitzig retweetledi
The Robot Lord future
Erik Brynjolfsson@erikbryn
The @nytimes piece today by @ByrneEdsal13590 highlights a concern I share: “If we stay on the current path, the risk of extreme concentration — both economic and political — is very real.” In work with @zhitzig, we ask why AI may shift the balance between dispersed knowledge and centralized control.
English
Zoë Hitzig retweetledi
The @nytimes piece today by @ByrneEdsal13590 highlights a concern I share:
“If we stay on the current path, the risk of extreme concentration — both economic and political — is very real.”
In work with @zhitzig, we ask why AI may shift the balance between dispersed knowledge and centralized control.
English
Zoë Hitzig retweetledi
I used to lead the Geopolitics Team at OpenAI. Today I published a few observations on frontier AI companies and their military usage policies from my perspective as a former employee and researcher active in the int’l security space. (Link below.)
English
@alexolegimas This paper matches demographic data to real-world usage from chatgpt –
Looks at not just who is adopting but who is adopting for what kinds of tasks
nber.org/system/files/w…
English
Studies like this are useful for quantifying impact of AI given uniform adoption. But adoption is not uniform, and those w/ high-skilled/education are more likely to adopt it---implying this equalizing effect will potentially flip in real-world data.
aleximas.substack.com/p/who-uses-ai-…
Carl Benedikt Frey@carlbfrey
Trouble for higher education? AI boosts everyone’s productivity, but it benefits lower-education participants most (reducing the the education productivity gap by about 75%)
English
Zoë Hitzig retweetledi
we can write a million essays about how the future Silicon Valley wants to build is underwritten by a deep disgust with / contempt for Being A Human, or we can just let them speak for themselves
Chief Nerd@TheChiefNerd
🚨 SAM ALTMAN: “People talk about how much energy it takes to train an AI model … But it also takes a lot of energy to train a human. It takes like 20 years of life and all of the food you eat during that time before you get smart.”
English
Zoë Hitzig retweetledi
Last week @zhitzig wrote an article calling for deeper scrutiny of ads and consumer protections in AI. I have much more confidence than she does in OpenAI's direction and commitments on ads, but I believe she makes an important and credible argument about long-term oversight.
English
with thanks to @NewStatesman for publishing back then newstatesman.com/culture/poetry…
English
Zoë Hitzig retweetledi
It’s Solution Sunday!
We need federal AI regulation. Things are getting crazy and piecemeal state stuff is not gonna cut it.
if you’re at an AI company, what is your policy team’s proposal/strategy here? Does it make any sense? Could you help make it better?
English
My message is optimistic actually: we still have time to do better than exclusion via $ or manipulation via ads.
But we need public pressure – from people who know what’s at stake and have clear demands.
Let’s make AI companies step up and earn our trust
nytimes.com/2026/02/11/opi…
English
I resigned from OpenAI on Monday. The same day, they started testing ads in ChatGPT.
OpenAI has the most detailed record of private human thought ever assembled. Can we trust them to resist the tidal forces pushing them to abuse it?
I wrote about better options for @nytopinion
English
My argument, in brief:
Zoë Hitzig@zhitzig
I resigned from OpenAI on Monday. The same day, they started testing ads in ChatGPT. OpenAI has the most detailed record of private human thought ever assembled. Can we trust them to resist the tidal forces pushing them to abuse it? I wrote about better options for @nytopinion
English
Zoë Hitzig retweetledi
“We don’t have a grasp of what AI does to people’s psychology, what it does to them sociologically…”
AI researcher Zoe Hitzig explains why she resigned from OpenAI
#Newsnight
English
Zoë Hitzig retweetledi
Zoë Hitzig retweetledi
It might be beneficial to both society and to OpenAI to explore ways the company could credibly commit to not letting ads bias responses or draw on overly personal data.
For society, this helps stave off some concerning uses of AI and some of the ways it could erode the information environment. For OpenAI, this builds trust so they can run ads without their product losing users who worry about the motivations behind the responses they’re seeing.
There are a variety of interesting institutional designs that might help with this, along the lines of the oversight board that Zoe refers to. The key is something that ties the company’s hands so it isn’t easy to change later.
Zoë Hitzig@zhitzig
I resigned from OpenAI on Monday. The same day, they started testing ads in ChatGPT. OpenAI has the most detailed record of private human thought ever assembled. Can we trust them to resist the tidal forces pushing them to abuse it? I wrote about better options for @nytopinion
English
Zoë Hitzig retweetledi
really excited about this work & exploring how we can aggregate values and preferences from large groups of humans with different backgrounds and opinions! 🔥
Zoë Hitzig@zhitzig
New on the OpenAI alignment blog! We prototype a method for eliciting the values that drive preferences over model responses, and release CoVal, an experimental dataset we built with it. Details in thread 👇
English
Super excited to share that the non-profit I spent last year co-founding (AVERI) just launched🚀
AVERI stands for AI Verification and Evaluation Research Institute, and today we published a paper in collaboration with researchers at dozens of orgs. 🧵 x.com/AVERIorg/statu…
AVERI@AVERIorg
AVERI’s goal is to make third-party auditing of frontier AI effective and universal. Today, together with coauthors at dozens of organizations we set out our vision in "Frontier AI Auditing" — a framework for rigorous third-party audits of safety and security practices at leading AI companies.
English
4/ We hope others use CoVal to better understand preference differences on these and other contentious prompts, and to iterate on the methods developed here!
Incredible coauthors: @mitchellgordon @ThankYourNiceAI @adamfungi @SandhiniAgarwal
Blog post: alignment.openai.com/coval/
English
3/ We also experiment with scoring model outputs using CoVal, and find that CoVal scores track human preference rankings out of sample.
(Nonetheless, CoVal-derived scores must be interpreted carefully—they reflect a particular crowd’s values under a particular aggregation rule.)
English
New on the OpenAI alignment blog!
We prototype a method for eliciting the values that drive preferences over model responses, and release CoVal, an experimental dataset we built with it.
Details in thread 👇
English