Arvind Narayanan

Advancing science- and evidence-based AI policy Policy must be informed by, but also facilitate the generation of, scientific evidence

What kind of AI governance do we need? Our new piece in @science.org answers this: we need policy grounded in evidence and built to generate more of it. Evidence-based policymaking is not a slogan—it’s a design challenge for democratic governance in the age of AI www.science.org/doi/10.1126/... 🧵

Note that the data collection ended right before ChatGPT was released, so my guess is that the percentages are no longer small.

Could AI slow science? Confronting the production-progress paradox

Fabulous post by @randomwalker.bsky.social & Sayash raising the same concern many of us have about whether we're on the right track with how we're using AI for science. Everyone should read it, take a deep breath & think through the implications.

www.aisnakeoil.com/p/could-ai-s...

Understanding Social Media Recommendation Algorithms

I’m reading a very well written 2023 paper on social media recommender systems from @randomwalker.bsky.social I had completely forgotten that in the 00s “neither Facebook nor Twitter had the ability to reshare or retweet posts in your feed.”What a huge shift!

knightcolumbia.org/content/unde...

We’re hiring at Princeton on AI and society, working with Arvind Narayanan or me depending on fit.

I think current AI developments are all a huge deal but am very unexcited by current state of the AGI and/or AI safety discourse.

Please share as you see fit.

puwebp.princeton.edu/AcadHire/app...

After consideration, I will post occasionally, but heavily censor what I share compared to other sites.

I tried making the transition, but talking about AI here is just really fraught in ways that are tough to mitigate & make it hard to have good discussions (the point of social!). Maybe it changes

Two Paths for A.I. The technology is complicated, but our choices are simple: we can remain passive, or assert control.

For @newyorker.com, Joshua Rothman spoke with @randomwalker.bsky.social and @sayash.bsky.social, authors of AI Snake Oil and a recently published paper “AI as Normal Technology”, which argues that practical obstacles will slow AI’s uses and potential: www.newyorker.com/culture/open...

"A hypothesis on the accelerating decline of reading: * Broadly speaking, people read for pleasure/entertainment and for learning/obtaining information. * Reading for pleasure has been declining for a while and is being replaced by videos (very sharply among young people). This trend will surely continue. * Reading for obtaining information is getting intermediated by chatbots. We are in the very early stages of this shift, so I think people underappreciate the magnitude of what's coming. It's not just that AI replacing traditional web search. Even when it comes to reading news articles, business documents, or scientific papers, the vision that tech companies are pushing on us is AI summarization + synthesis + Q&A. * We don't have to accept this, but I predict that most people will. It's a tradeoff between speed/convenience and accuracy/depth of understanding — the same tradeoff that was once offered to us when it became possible to search the web to look up a quick fact as opposed to reading about the topic in depth in an encyclopedia. * Just as most people in most cases prefer a shallow web search over deeper reading, most people in most cases will prefer AI-intermediated access to knowledge. Traditional reading won't disappear, but people will do it vastly less often, except in hobbyist reading communities and professions where traditional reading is needed. * The decline of reading-for-pleasure (due to video) and reading-for-information (due to AI) will accelerate each other, as reading text without an intermediary will come to be seen as a chore. * Personally, I find this sad. But while it's tempting to moralize all this, I think that's unproductive. Yelling at individuals to resist new media has been done for centuries and has never worked. * Even if people individually rationally choose these tradeoffs, I think we collectively lose something; critical reading skills are arguably essential for a democracy. We need to figure out what to do about that.

clear, depressing set of observations from @randomwalker.bsky.social - "The decline of reading-for-pleasure (due to video) and reading-for-information (due to AI) will accelerate each other, as reading text without an intermediary will come to be seen as a chore."

Moving towards informative and actionable social media research Social media is nearly ubiquitous in modern life, and concerns have been raised about its putative societal impacts, ranging from undermining mental health and exacerbating polarization to fomenting v...

New preprint with @jbakcoleman.bsky.social @lewan.bsky.social @randomwalker.bsky.social @orbenamy.bsky.social @lfoswaldo.bsky.social where we argue for a complex-system perspective to understand the causal effects of social media on society and for a triangulation of methods
arxiv.org/abs/2505.09254

I'm excited that I can finally share what I've been working on for the past 9 months:

The United Nations 2025 Human Development Report: "A matter of choice: People and possibilities in the age of AI" 🧵

hdr.undp.org/content/huma...

AGI is not a milestone There is no capability threshold that will lead to sudden impacts

“AGI is not a milestone because it is not actionable. A company declaring it has achieved, or is about to achieve, AGI has no implications for how businesses should plan, what safety interventions we need, or how policymakers should react.”
@randomwalker.bsky.social
open.substack.com/pub/aisnakeo...

A Guide to Cutting Through AI Hype: Arvind Narayanan and Melanie Mitchell Discuss Artificial and Human Intelligence - CITP Blog Last Thursday’s Princeton Public Lecture on AI hype began with brief talks based on our respective books: The meat of the event was a discussion between the two of us and with the audience. A lightly ...

Hi, if you mean my conversation with @melaniemitchell.bsky.social I shared the transcript here: blog.citp.princeton.edu/2025/04/02/a...

Okay just started @randomwalker.bsky.social and @sayash.bsky.social's new essay and this is 🔥🔥🔥.

"Resilience as the overarching approach to catastrophic risk" -- yes thank you exactly this.

kfai-documents.s3.amazonaws.com/documents/c3...

text says "ML Reproducibility Challenge Princeton University, New Jersey, USA, August 21 2025"

We are hosting @reproml.org 2025 on Aug. 21. There will be invited talks, oral presentations, and poster sessions. Keynote speakers include @randomwalker.bsky.social, @soumithchintala.bsky.social, @jfrankle.com, @jessedodge.bsky.social, @stellaathena.bsky.social

Register now: bit.ly/4cP8vIq

In this clip from our event last week, @randomwalker.bsky.social describes how we can map out the landscape of AI along two dimensions: how well the AI tool works, and how harmful (or benign) it is.

Watch a full recording of the event: youtu.be/C3TqcUEFR58

AI as Normal Technology A new paper that we will expand into our next book

IMO, the most important piece on AI of the last 6 months and I recommend it to everyone. A genuinely careful consideration of the technology and its intersections with culture and labor from @randomwalker.bsky.social and @sayash.bsky.social Authors of AI Snake Oil substack.com/home/post/p-...

AI as Normal Technology

Truly thoughtful and essential analysis of the AI field from @randomwalker.bsky.social @sayash.bsky.social. States what many felt, but haven't articulated. Pairs well with Shazeda Ahmed's "epistemic culture of AI safety" and others' work on risk and anti-trust.

knightcolumbia.org/content/ai-a...

AI as Normal Technology

In a new essay from our "Artificial Intelligence and Democratic Freedoms" series, @randomwalker.bsky.social & @sayash.bsky.social make the case for thinking of #AI as normal technology, instead of superintelligence. Read here: knightcolumbia.org/content/ai-a...

On Thursday, April 17 at 5:30 PM EDT, we welcome @randomwalker.bsky.social to discuss his latest book, "AI Snake Oil," co-authored with Sayash Kapoor. The presentation will be followed by a discussion with @dacemoglumit.bsky.social.

Register here: bit.ly/4cGF3Eo

Why an overreliance on AI-driven modelling is bad for science Without clear protocols to catch errors, artificial intelligence’s growing role in science could do more harm than good.

New commentary in @nature.com from professor Arvind Narayanan (@randomwalker.bsky.social) & PhD candidate Sayash Kapoor (@sayash.bsky.social) about the risks of rapid adoption of AI in science - read: "Why an overreliance on AI-driven modelling is bad for science" 🔗

#CITP #AI #science #AcademiaSky

A Guide to Cutting Through AI Hype: Arvind Narayanan and Melanie Mitchell Discuss Artificial and Human Intelligence - CITP Blog Last Thursday’s Princeton Public Lecture on AI hype began with brief talks based on our respective books: The meat of the event was a discussion between the two of us and with the audience. A lightly ...

A Guide to Cutting Through AI Hype: Arvind Narayanan (@randomwalker.bsky.social)
and Melanie Mitchell (@melaniemitchell.bsky.social
Discuss Artificial and Human Intelligence / CITP Blog blog.citp.princeton.edu/2025/04/02/a...

On 4/10 and 4/11, we're hosting our symposium "AI and Democratic Freedoms." Excited to have panelists @atoosakz.bsky.social, @randomwalker.bsky.social, @alondra.bsky.social, & Deirdre K. Mulligan join moderator
@shaynelongpre.bsky.social to kick it off. RSVP: www.eventbrite.com/e/artificial...

"Computer scientists Arvind Narayanan and Sayash Kapoor will present “Understanding AI: What It Can and Cannot Do” at 12 p.m. on April 3. The program will be streamed on Baltimore County Public Library's YouTube channel & bit.ly/AISnakeOil. No registration needed"
msla.maryland.gov/Pages/press-...

A Guide to Cutting Through AI Hype: Arvind Narayanan and Melanie Mitchell Discuss Artificial and Human Intelligence Princeton University Professor Arvind Narayanan and Professor Melanie Mitchell of the Santa Fe Institute, will share their insights on Artificial Intelligence in a discussion format, followed by a…

This Thursday (3/27) at 5 pm EDT: Join Princeton University Public Lectures in welcoming @randomwalker.bsky.social, co-author of AI Snake Oil. He and @melaniemitchell.bsky.social will share their insights on #AI.

This event is free & open to the public:

Book Review: AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference, by Arvind Narayanan and Sayash Kapoor - Alexya Martinez, 2025

NEW #JMCQReview🚨In this book, @randomwalker.bsky.social and @sayash.bsky.social dissect the promises and the pitfalls of predictive AI. Read the @jmcquarterly.bsky.social review by Alexya Martinez #commsky

journals.sagepub.com/doi/10.1177/...

Thank you for making this. Is the source available? I would like to learn how to build something like this.

A Guide to Cutting Through AI Hype: Arvind Narayanan and Melanie Mitchell Discuss Artificial and Human Intelligence Princeton University Professor Arvind Narayanan and Professor Melanie Mitchell of the Santa Fe Institute, will share their insights on Artificial Intelligence in a discussion format, followed by a que...

Very excited to participate in this event with @randomwalker.bsky.social next week!

lectures.princeton.edu/lectures/202...

The Wuhan lab is STILL doing unsafe research that could trigger a pandemic AND getting prestigious pubs as incentive.😫

To fix this for the future, we have to admit we were deliberately misled on the possibility of a lab leak in the past. Horrid but true.

Gift link
www.nytimes.com/2025/03/16/o...

We focus so much of our attention on algorithms, but not enough on the design of social networks.

Loved this insight from @randomwalker.bsky.social on what makes TikTok feel so different here: knightcolumbia.org/blog/tiktoks...

What are 3 concrete steps that can improve AI safety in 2025? 🤖⚠️

Our new paper, “In House Evaluation is Not Enough” has 3 calls-to-actions to empower evaluators:

1️⃣ Standardized AI flaw reports
2️⃣ AI flaw disclosure programs + safe harbors.
3️⃣ A coordination center for transferable AI flaws.

1/🧵