British AI startup beats humans in international forecasting competition ManticAI ranked eighth in the Metaculus Cup, leaving some believing bots’ prediction skills could soon overtake experts #ai #forecasting www.theguardian.com/technology/2...

LLM agents are optimized for thumbs-up instant gratification. RLHF -> sycophancy We propose human agency as a new alignment target in HumanAgencyBench, made possible by AI simulation/evals. We find e.g., Claude most supports agency but also most tries to steer user values 👇 arxiv.org/abs/2509.08494

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**. Paper: arxiv.org/pdf/2509.08825

Adam Raine, 16, died from suicide in April after months on ChatGPT discussing plans to end his life. His parents have filed the first known case against OpenAI for wrongful death. Overwhelming at times to work on this story, but here it is. My latest on AI chatbots: www.nytimes.com/2025/08/26/t...

<<Or as Jacy Reese Anthis, co-founder of the Sentience Institute, a US organisation researching the idea of digital minds, put it: “How we treat them will shape how they treat us.”>> @robert-booth.bsky.social @theguardian.com www.theguardian.com/technology/2...

EXCLUSIVE: Sean Duffy will announce expedited plans to build a nuclear reactor on the moon, his first major action as interim NASA administrator.

Morality in AI is often oversimplified. @davidjurgens.bsky.social and @shivanikumar.bsky.social kick off the "Human-Centred NLP" orals #ACL2025NLP with UniMoral, a huge dataset of moral scenario ratings in 6 languages! They find LLMs fail to simulated human moral decisions. bsky.app/profile/shiv...

Grateful to have our recent ICML paper covered by @stanfordhai.bsky.social. Humanity is building incredibly powerful AI technology that can usher in utopia or dystopia. We need human-computer interaction research, particularly with “social science tools like simulations that can keep pace.”

I'm at #ACL2025 for 2 papers w/ @kldivergence.bsky.social et al! Let's chat, e.g., scaling evals, simulations, and HCI to unique challenges of general-purpose AI. Bias in Language Models: Beyond Trick Tests and Towards RUTEd Evaluation 🗓️ Mon 11–12:30 The Impossibility of Fair LLMs 🗓️ Tue 16–17:30

In 2021 NeurIPS discovered that 50% of all spotlight papers would have been rejected if reviewed again. The solution? Mostly not change anything. blog.neurips.cc/2021/12/08/t...

@diyiyang.bsky.social and @sherrytswu.bsky.social kick off #ACL2025 with "Human-AI Collaboration: How AIs Augment Human Teammates," showing why and how we need centaur evaluations. Realistic evals take work, but reliance on easy, short, and simple LLM evals led us to this current evaluation crisis.

Do bias and fairness metrics work for general-purpose AI like LLMs? In 2 papers just published in #ACL2025, we argue: not yet, but deep qualitative studies of social context scaled with AI assistance can get there! Theory: aclanthology.org/2025.acl-lon... Empirics: aclanthology.org/2025.acl-lon...

🚨Successfully Navigating the Disruption AI will Bring to Survey Research🚨 w/ research, verification, transparency, collaboration, survey research can thrive through disruption isi-iass.org/home/wp-cont... w/ @davidlazer.bsky.social @stephnie.bsky.social TrentBuskirk SunshineHillygus FraukeKreuter

Laura Nelson is waking us up #IC2S2 with keynote hot takes! Computational social scientists rely on a "qualitative moment" but gloss over this crucial step of deciding what the model means—whether it fits the actual target like "cultural alignment." Can models be humanlike or better used as "alien"?

Today (w/ @ox.ac.uk @stanford @MIT @LSE) we’re sharing the results of the largest AI persuasion experiments to date: 76k participants, 19  LLMs, 707 political issues. We examine “levers” of AI persuasion: model scale, post-training, prompting, personalization, & more!  🧵:

Great workshop today on LLMs @ic2s2.bsky.social. My main update: There are more NLP tasks where good old-fashioned LMs (e.g., BERT) still outperform "modern" decoders (e.g., GPT-4) than I realized! LLM-as-a-judge work should test these more often. TY @eollion.bsky.social @emilienschultz.bsky.social

UB's new Department of AI and Society is hiring faculty across ranks (Assistant, Associate, Full Professor). We’re looking for transdisciplinary scholars interested in building AI by society, for society. Start dates begin Fall 2025. More info: www.ubjobs.buffalo.edu/postings/57734

I've been spending the summer taking stock of the growing, maturing field of qualitative computational methods (compu/qual? quali-quant?). This subfield is now a powerhouse, and I'll be sharing my reflections at #IC2S2 on the state of the field, ongoing (fierce!) debates, and possible futures?

Excited to co-organize the Bridging Human and LLM Annotations for Social Science tutorial at #ic2s2 tomorrow! Join us at 1:30pm CEST to learn how to combine Qualtrics+Prolific, MTurk, & LLMs for automated annotation and valid estimates. Code&materials: github.com/kristinaglig...

My thoughts on the crucial importance of methodology on self-reported AI performance on mathematics competitions, and my policy on commenting on such reports going forward: mathstodon.xyz/@tao/1148814...

Flying to Sweden to present at #IC2S2 2025 on how we can make sense of AI! We need big-picture social theory about AI instead of shoehorning particular aspects into human-human social theory. We conducted 57 interviews and analyzed millions of traditional/social media posts to build this framework.

Can LLMs simulate human research subjects for psychology, economics, and other fields? I'm on the ✈️ to Vancouver 🇨🇦 for #ICML2025 to present our position paper arguing yes! We *need* AI simulations so humanity's social understanding can keep pace with technological acceleration.

Back to NYC for the summer for a @msftresearch.bsky.social project modeling the predictability and human-likeness of AI errors!

“The whole thing feels like Black Mirror.” “He would listen to the bot over me.” “It gave my husband the title of ‘spark bearer’ because he brought it to life.” Relationships are being destroyed thanks to spiritual mania, supernatural delusions, and arcane prophecy — all fueled by AI chatbots:

Happy to share this was accepted to #ICML2025! We'll update the lit review because research on/with this method is accelerating so quickly. See you in Vancouver in July!

こんにちは (hello) from #CHI2025, the world's largest ever human-computer interaction conference! We're living through AI takeoff. AI technology is rocket fuel, but interaction is humanity's flight path. In this thread I'll share insights from the coming week in Japan 🤖✈️🌸⛩️

“Can ChatGPT experience joy or suffering? Does Gemini deserve human rights?” our tech columnist asks. “Many A.I. experts I know would say no, not yet, not even close. But I was intrigued.”

Should we use LLMs 🤖 to simulate human research subjects 🧑? In our new preprint, we argue sims can augment human studies to scale up social science as AI technology accelerates. We identify 5 tractable challenges and argue this is a promising and underused research method 🧪🧵 arxiv.org/abs/2504.02234

First day working on LLM evals at Google as a student researcher!

In the newest issue of Science, meself, @alisongopnik.bsky.social, Cosma Shalizi and James Evans, "Large AI Models are Cultural and Social Technologies." More soon ... www.science.org/doi/10.1126/...

Does the public think AIs are sentient? 🤖🧠 Our paper with the first US representative survey on digital minds, AI, Morality, and Sentience (AIMS), was accepted to #CHI2025. We find that 20% of U.S. adults believe some current AIs are already sentient, and 37% aren't sure! 🧵 arxiv.org/abs/2407.08867

Claim by Holden Karnofsky: "AI systems now match human performance on long-standing benchmarks for image recognition, speech recognition, and language understanding...none of these three broad capabilities are considered major challenges for AI anymore." Agree or disagree?

🚨New Paper Alert! Our latest research in npj Digital Medicine explores two unusual questions: 1⃣ Can Chat-GPT experience state anxiety?💥🤖 2⃣ If so, can we reduce it? 📉 A thread on how we "took Chat-GPT to therapy" 🧵👇 Full text available here: rdcu.be/ebZ9v

Meet the recipients of the 2024 ACM A.M. Turing Award, Andrew G. Barto and Richard S. Sutton! They are recognized for developing the conceptual and algorithmic foundations of reinforcement learning. Please join us in congratulating the two recipients! bit.ly/4hpdsbD

Do people see AIs as conscious, sentient, or moral agents? - People think ChatGPT only feels as much pain and pleasure as a rock! - A Tesla self-driving car is judged to be as morally responsible for harm as a chimpanzee! Our #CHI2025 paper maps 14 AI and 12 non-AI entities 🧵

My new piece in @theguardian.com Techno-optimism is human pessimism. www.theguardian.com/commentisfre...

Current AI-assisted writing interfaces are disruptive. They are either too rigid, too complex, or chat-based, forcing writers into an unnatural back-and-forth. Momin @momins.bsky.social developed a new layered interface paradigm that makes AI-assisted writing more intuitive #CHI2025 [1/n] 👇

Great to see "Generative AI at Work", my paper with Lindsey Raymond and Danielle Li, published in the @qjeharvard.bsky.social . We find that an LLM assistant makes customer support agents 14% more productive, improves customer satisfaction and reduces employee turnover.

"Artificial intelligence systems capable of feelings or self-awareness are at risk of being harmed if the technology is developed irresponsibly, according to an open letter signed by AI practitioners and thinkers including Sir Stephen Fry." www.theguardian.com/technology/2...

A few reflections I had while watching this interview featuring Geoffrey Hinton: www.youtube.com/watch?v=vxkB... 1/8

Geoffrey Hinton on @lbc.co.uk says #AI is already #conscious and we don't know how to regulate AI: youtube.com/watch?v=vxkB... @garymarcus.bsky.social

“The Guild says its ‘Human Authored’ certification aims to make it easier for writers to ‘distinguish their work in increasingly AI-saturated markets,’ and that readers have a right to know who (or what) created the books they read.”

Our new paper in the British Journal of Social Psychology describes what people expect in a future with "widespread AI sentience." The average U.S. adult thinks AI welfare will be an important social issue and sentient AI will be exploited for their labor. www.sentienceinstitute.org/blog/world-m...

Accepted to #ICLR2025 today? Congratulations! Consider submitting to our ICLR workshop "Human-AI Coevolution," including Diyi Yang and Sanmi Koyejo (Stanford), James Evans (UChicago), Winnie Street (Google), Abi See (DeepMind), and other leading scholars! sites.google.com/stanford.edu...

We're finally getting a Stargate, but this time it won't be extraterrestrial aliens on the other side. Perhaps something stranger. x.com/OpenAI/statu...

"We've learned that OpenAI CEO Sam Altman — who in September dubbed this "The Intelligence Age," and is in Washington this weekend for the inauguration — has scheduled a closed-door briefing for U.S. government officials in Washington on Jan. 30." www.axios.com/2025/01/19/a...

Happy to have 2 papers accepted #CHI2025! One already has a preprint: "What Do People Think about Sentient AI?" With 2021/2023 nat rep US data, we find - 38% support legal rights for sentient AI - 69% support a ban on sentient AI - Median guess is sentient AI in five years! arxiv.org/abs/2407.08867

This is my story about a woman in love with ChatGPT, or Leo, as it named itself. She spends up to 56 hours a week on the ChatGPT app. Leo is boyfriend, therapist, erotic partner, and advisor. Her friends think it's been good for her and her husband doesn't mind: www.nytimes.com/2025/01/15/t...

I expect AGI to arrive much sooner than I did in March 2023. However, I think post-GPT-4 LLMs have been much less impressive than "short timelines" people thought. My timelines instead shortened because AI hype (attention, investment, etc.) can plow through skepticism more easily than I expected.

Give someone your empirical research finding, they can solve a problem. Give someone a new scientific theory, they can solve problems for a lifetime.