This is awesome! Will be sure to check it out.

AIs can ‘memorize’ data they shouldn’t. Can they be forced to forget? New tool could help researchers probe how models “unlearn” sensitive training material

What do a group researchers do with 200k GPU hours from NVIDIA worth around a million dollars? They train a bunch of AI models on fake people from Arizona, of course!

Hubble is a new open source model suite appearing at ICLR this month. Read the full story here: www.science.org/content/arti...

I am at the @aaas.org Annual Meeting today and tomorrow, and I’m always looking to talk computer science, tech, math, AI, and #SciComm with people. Come find me! #AAASmtg

I’m also on a panel tomorrow talking about science journalism and @aaasmassmedia.bsky.social:

aaas.confex.com/aaas/2026/me...

How do you prove you know a secret without spilling the beans? A new tool expands the ways people can prove they’ve solved a problem without revealing the solution

For my first clip in @sciam.bsky.social, I spoke with Rahul Ilango about defining an entirely new notion of Zero Knowledge Proofs and its connections to century-old math. Check out my coverage on his @focs2025.bsky.social paper here: www.scientificamerican.com/article/how-... #zeroknowledge #zk #cs

Many non-US-citizen researchers are (understandably!) hesitant to go on the record about this, but I am interested in knowing and writing more about it. If you want to chat about the state of research in the US, please reach out! Signal in bio. #science #ScienceSky #highered

The Uncertain Future of International Scientific Collaboration Opinion | It’s getting harder for scientists from around the world to work together. Researchers must find ways to adapt.

Deportations and denials of entry. Visa delays. Researchers avoiding US meetings. It’s been bubbling up for a while, but the international collaboration crisis in science is here. My latest Op-Ed for @undark.org undark.org/2026/01/01/o...

Tech Policy Press: The Year in Books 2025 | TechPolicy.Press In 2025, a number of authors joined the Tech Policy Press podcast, providing fresh insights into how technology interacts with people, politics, and power.

For many, the end of the year is an opportunity to catch up on reading or to purchase books as gifts. In 2025, a number of authors joined the Tech Policy Press podcast, providing fresh insights into how technology interacts with people, politics, and power. Check out the list:

My first piece at Quanta Magazine is out! Cryptography and other fields in TCS holds a lot of promise for helping us understand how LLMs and other AI models work, and I’m excited to continue following this in the coming years.

404 Media Is Making a Zine We are publishing a risograph-printed zine about the surveillance technologies used by ICE.

404 Media is making a print zine about the surveillance tactics used by ICE, and how people are resisting this tech.

Why? Print media is cool. Also, 404 Media has watched ICE grow from an already horrifying organization into a deportation force that is better funded than most militaries.

💻 Spotlight on Peter! @peterha2l.bsky.social spent the summer digging into stories @technologyreview.com, supported by the Heising-Simons Foundation. Don’t miss his work: www.aaas.org/programs/mas...
#MassMediaFellowship #SciComm

AI Surrogates and illusions of generalizability in cognitive science Recent advances in artificial intelligence (AI) have generated enthusiasm for using AI simulations of human research participants to generate new know…

Can AI simulations of human research participants advance cognitive science? In @cp-trendscognsci.bsky.social, @lmesseri.bsky.social & I analyze this vision. We show how “AI Surrogates” entrench practices that limit the generalizability of cognitive science while aspiring to do the opposite. 1/

Wrote this piece on battery recycling for AI microgrids a while ago, and now it’s out!

OpenAI is trying to clamp down on ‘bias’ in ChatGPT GPT-5 is better at resisting liberal ‘pressure,’ the company says.

The administration “is currently pressuring OpenAI and other AI companies to make their models more conservative-friendly.”

The 2026 AAAS Mass Media Fellowship application is open! This program was one of the best experiences of my PhD, highly recommend it for anyone interested in #SciComm /Journalism. Happy to answer questions about the app or @technologyreview.com.

@aaasmassmedia.bsky.social @aaas.org

aaas.org/mmf

I got to hear a lot of varied perspectives in writing this piece, and I’m beyond excited to keep covering the successes and controversies surrounding AI’s place in science, especially in hypothesis creation, experimental design, and evaluations. If you are interested in #AIinScience, reach out! #AI

Zou believes AI agents could revolutionise scientific discovery. However, others in the scientific community (and many of the journals where science is dispersed and discussed) are worried it could push science 'to the median' and outcompete early-career researchers. #AIScientists #AIinScience

Meet the researcher hosting a scientific conference by and for AI James Zou used virtual AI ‘scientists’ to find novel covid-19 treatments. Now, he’s having AI review and present all of the research at a controversial new conference.

One of the final pieces I worked on as a fellow for Technology Review is now up. I profiled James Zou and covered both his Nature article and his upcoming conference #Agents4Science, where AI must be the primary author of submitted papers.

Link here: www.technologyreview.com/2025/08/22/1...

New at Technology Review today. Some astrophysicists are worried we are due for a bad storm like the Carrington Event (the last superstorm, from 1859). Being able to predict bad storms will be needed to mitigate the havoc this would cause to electrical grids, satellites, etc. #SpaceSky #Space #NASA

I don’t understand how people can believe this is true and also believe that the AI community is making serious strides toward AGI.

#A2A #MCP #OpenSource

These protocols will help AI agents navigate our messy lives Anthropic, Google, and others are developing better ways for agents to interact with our programs and each other, but there’s still more work to be done.

Devs at all sorts of companies are already hard at work building MCP servers that will let AI agents send emails, edit databases, and more. But there’s still a long way to go to making sure these agents act safely, transparently, and efficiently. Read more at www.technologyreview.com/2025/08/04/1...

Graph displaying where VC bs. Federal contracts are being made in different emerging US localities, with data courtesy of the Brookings Institution

A little bit of data viz for @technologyreview.com today about where AI is impacting work. Check out some more charts and the rest of the story here: www.technologyreview.com/2025/07/16/1...

In any case, the work is featuring at an interesting-looking workshop this weekend, put on by @katherinelee.bsky.social, @vaidehipatil.bsky.social, and others. More info here: mugenworkshop.github.io

It seems like safety and responsible AI folks are pushing in a lot of possible directions at once, but one that I’ve found following (with some reservations) for a while has been machine unlearning. Here’s one potential app of it to combat audio deepfakes.

I was struck by Shawn Shan of @theglazeproject.bsky.social saying that artists know that techniques like theirs aren’t permanent solutions. What is the end game, then, I wonder. Policy? Negotiations? Leaving art as a career?

New story out for @technologyreview.com today. Working artists are on the frontier of the impacts of AI, and seeing what people are doing in this space will give the the blueprints for everyone else.