Martin Wattenberg

edjefferson.comLocal Wikipedia: The Encyclopedia For Local PeopleWikipedia, but only for things a half hour walk from you.

What a great idea! Got lost for a while exploring...

Ed Jefferson@edjefferson.com

Had this idea in lockdown, half-built it, forgot about it, remembered it again, though might as well finish it.

Play a game of Go against an old-fashioned AI, and watch as it thinks! The curves visualize spheres of influence for each potential future position it considers. Try it at www.bewitched.com/go

Just added a black-and-white mode, by popular demand!

Play a game of Go against an old-fashioned AI, and watch as it thinks! The curves visualize spheres of influence for each potential future position it considers. Try it at www.bewitched.com/go

This is the latest in a series of transparent game AIs created with Marek Walczak. The earlier pieces focused on chess. You can see some of these at www.bewitched.com/chess/ or at MoMA through mid-2027: www.moma.org/calendar/exh...

Reviewing an AI-written paper is like listening to someone's boring dream

I fully expect this will be different in a year, and then we will have a whole new set of problems

Four dimensions is too many! Let's unfold this 4D cube to three dimensions, then two, and then one. Don't worry, we'll put it back together again. The math is lovely, but it's the rhythm and dance of the lines that fascinates me here.

If you like this kind of thing, there's more at www.instagram.com/martin.images/

Baby Name Wizard author Laura Wattenberg

6 Sept

btw, I've started an art-only account: www.instagram.com/martin.images/ I'm posting work that's new, as well as old images and some projects that are "restored" (ported from ancient Java). If you're on Instagram, see you there!

Martin Wattenberg reposted

Alberto Acerbi

@acerbialberto.com

12 Aug

This pattern often noticed for cars seems to hold also for furniture... from: europeancorrespondent.com/en/r/how-col...

Martin Wattenberg reposted

Nicolas Lambert

@neocarto.bsky.social

24 July

Footbal projections observablehq.com/@mpkhinda/th...

Martin Wattenberg reposted

@babynames.bsky.social

8 July

The topic for this article literally came to me in a dream. And honestly, I think my unconscious was onto something. bit.ly/NameSpread

bit.lyWatch the Influence of a Celebrity Baby Name SpreadWhat started out as a single name became so much more.

2 July

The one paper I review that desperately needs an ethics discussion is, predictably, the one paper that omits one. Meanwhile, the other papers are utterly innocuous and have long statements that are basically "Air currents created by typing this paper may potentially lead to a hurricane in 2029"

@simongerman600.bsky.social

28 June

This is such a fantastic paper! We need more of this kind of empirical work on what people really do with AI. The idea of the "solipsistic reader-writer" seems important. Also, the paper is so well-written!

Melanie Walsh@mellymeldubs.bsky.social

Excited to share this. @neel2112.bsky.social, @mariaa.bsky.social, and I analyzed 500K anonymous ChatGPT convos (shared w/ consent from WildChat) to see if people were generating fiction. We found tons of stories, fanfiction & erotica. Many users iterated on the same stories for days and weeks.

Martin Wattenberg reposted

Simon Kuestenmacher

14 June

What an amazing way to visualize early human migration. Lovely map by @HarvardCGA. A great colour scheme and an appropriate map projection! Source: buff.ly/3lbxonJ

31 Mar

Everyone's talking about AI sycophancy and meanwhile ChatGPT just called my writing "very salvageable"

I want to talk about why AI-based mass surveillance is so dangerous, and why I would oppose it no matter which party or president is in office.

View full thread

That offers governments a vast, unprecedented level of power over their citizens. In evil hands, that’s obviously a disaster. But, like the framers of the US Constitution, I believe it’s also wrong to give absolute powers to democratically elected leaders, or people you think of as the “good guys.”

As AI capabilities increase, we need a broad, deep, society-wide discussion of what limits make sense, and how we can hold the government meaningfully accountable to citizens. For that reason, I stand with Anthropic and anyone else who is avoiding a rush toward mass AI surveillance.

I want to talk about why AI-based mass surveillance is so dangerous, and why I would oppose it no matter which party or president is in office.

Before now, there was always a natural barrier on the power of surveillance. Even in the limit, if everyone’s actions were recorded all the time, there wouldn’t be enough people and time to watch and analyze the entire footage. But AI threatens to make that natural barrier completely obsolete.

Martin Wattenberg reposted

Santiago Ortiz

@moebio.bsky.social

26 Feb

This is @garrykasparov.bsky.social versus Deep Blue (game 2). Explore and interact with other games here: moebio.com/chess/ (including fast Hikaru versus Magnus, longest, shortest and oldest games ever!)

bit.lyThe 2025 Name of the Year is ElaraShe doesn’t exist. Yet she’s everywhere—and she’s all of us.

16 Dec 2025

In 2025, AI left its imprint on everything—even names! If you've ever asked an AI to tell you a story, you've probably seen the Elara-Elena-Clara nexus...

Baby Name Wizard author Laura Wattenberg@babynames.bsky.social

Meet Elara, the Name of the Year. She doesn’t exist. Yet she’s everywhere—and she’s all of us. bit.ly/Elara-Name-o...

3 Dec 2025

I want to play with this clever book myself!

Global Museum@globalmuseum.bsky.social

A book from the 19th century that depicts the Rhine Valley by creating an impression of three-dimensionality and spatial distance. @MasayukiTsuda2 #globalmuseum #books #travel #19thcentury

22 Nov 2025

I asked for a caricature of myself in the style of Al Hirschfeld, and Gemini knew to hide a NINA in my hair 😲

22 Nov 2025

Trying this a few more times, it turns out to work only sporadically. But still, that is one observant neural network!

theatlantic.comWhat AI Thinks It Knows About YouWhat happens when people can see what assumptions a large language model is making about them?

22 Nov 2025

It will hide other names, if you ask!

Martin Wattenberg reposted

vrli.bsky.social

@vrli.bsky.social

4 Aug 2025

Charts and graphs help people analyze data, but can they also help AI? In a new paper, we provide initial evidence that it does! GPT 4.1 and Claude 3.5 describe three synthetic datasets more precisely and accurately when raw data is accompanied by a scatter plot. Read more in🧵!

Martin Wattenberg reposted

Jonathan Zittrain

@zittrain.bsky.social

21 May 2025

AI is often thought of as a black box -- no way to know what's going on inside. That's changing in eye-opening ways. Researchers are finding "beliefs" models are forming as they converse, and how those beliefs correlate to what the models say and how they say it. www.theatlantic.com/technology/a...

Martin Wattenberg reposted

Baby Name Wizard author Laura Wattenberg

@babynames.bsky.social

12 May 2025

The interactive NameGrapher is updated with 2024 baby name popularity stats! Come explore--and marvel that Oliver and Olivia have converged namerology.com/baby-name-gr...

12 May 2025

A wonderful visualization for those of us obsessed by sunlight and geography!

Olivia Vane@oliviafvane.bsky.social

This map shows the hour of sunrise globally through the year. It reveals time zones following national and, sometimes, regional boundaries, and slicing through the oceans.

transformer-circuits.pubOn the Biology of a Large Language Model

27 Mar 2025

An incredibly rich, detailed view of neural net internals! There are so many insights in these papers. And the visualizations of "addition circuit" features are just plain cool!

Chris Olah@colah.bsky.social

Can we understand the mechanisms of a frontier AI model? 📝 Blog post: www.anthropic.com/research/tra... 🧪 "Biology" paper: transformer-circuits.pub/2025/attribu... ⚙️ Methods paper: transformer-circuits.pub/2025/attribu... Featuring basic multi-step reasoning, planning, introspection and more!

24 Mar 2025

Now that we have vibe coding, we need vibe testing!

24 Mar 2025

I'd be curious about advice on teaching non-coders how to test programs they've written with AI. I'm not thinking unit tests so much as things like making sure you can drill down for verifiable details in a visualization—basic practices that are good on their own, but also help catch errors.

3 Mar 2025

The wind map at hint.fm/wind/ has been running since 2012, relying on weather data from NOAA. We added a notice like this today. Thanks to @cambecc.bsky.social for the inspiration.

David Ho@davidho.bsky.social

All websites that depend on NOAA data should display something like this. earth.nullschool.net

25 Feb 2025

Neat visualization that came up in the ARBOR project: this shows DeepSeek "thinking" about a question, and color is the probability that, if it exited thinking, it would give the right answer. (Here yellow means correct.)

github.comReasoning or Performing · ARBORproject arborproject.github.io · Discussion #11Research Question When asked the DeepSeek Distilled R1 models a challenging abstract algebra question, they often generated hundreds of tokens of CoT before providing the final answer. Yet, on some...

25 Feb 2025

You can see the model go down the wrong path, "realize" it's not right, then find the correct answer! To see more visualizations, or if you have related ideas, join the discussion here! github.com/ARBORproject... (vis by @yidachen.bsky.social in conversation with @diatkinson.bsky.social )

20 Feb 2025

Great thread describing the new ARBOR open interpretability project, which has some fascinating projects already. Take a look!

David Bau@davidbau.bsky.social

ARBOR aims to accelerate the internal investigation of the new class of AI "reasoning" models. See the ARBOR discussion board for a thread for each project underway. github.com/ArborProjec...

github.comGitHub - ARBORproject/arborproject.github.ioContribute to ARBORproject/arborproject.github.io development by creating an account on GitHub.

20 Feb 2025

Today we're launching a multi-lab open collaboration, the ARBOR project, to accelerate AI interpretability research for reasoning models. Please join us! github.com/ARBORproject... (ARBOR = Analysis of Reasoning Behavior through Open Research)

github.comARBORproject arborproject.github.io · DiscussionsExplore the GitHub Discussions forum for ARBORproject arborproject.github.io. Discuss code, ask questions & collaborate with the developer community.

20 Feb 2025

Take a look at some initial research projects, and see if there's one you'd like to work on: github.com/ARBORproject... Or propose your own idea! There are many ways to contribute, and we welcome all of them.

19 Jan 2025

Out: rectangular maps of the globe In: rectangular maps of butterfly wings!

The Public Domain Review@publicdomainrev.bsky.social

In 1897, Alfred G. Mayer created his butterfly wing projections, an attempt to gain new insights into natural patterns and laws. Vertical blocks denote individual wings, distorted and stretched mathematically to fill a tidy rectangular space. More here: publicdomainreview.org/collection/m...

17 Jan 2025

I'm realizing that many people hear "generative art" (a term going back at least to the 60s!) as synonymous with generative AI. Is this how cryptographers feel about the new meaning of "crypto"?

Martin Wattenberg reposted

Simeon Schmauß

@stim3on.bsky.social

16 Jan 2025

What a beauty! This is comet C/2024 G3 (ATLAS) passing through the field of view of the LASCO C3 coronagraph. It wasn't for certain whether it would survive it's closest approach to the sun on January 13th, but it did and delivered us a spectacular show! #comet #C2024G3 🔭

15 Jan 2025

Idle question: Are there any papers from legit math journals using emoji as notation? We already use every symbol on the keyboard, musical sharps and flats, and even weird made-up fonts (what is that Weierstrass P??). A smiley is easy to draw with chalk and put into LaTeX, so why not?

15 Jan 2025

(Ha, even Wikipedia punts on what the Weierstrass P is, just describing it as "uniquely fancy")

15 Jan 2025

like if you had to name that function today, the obvious choice is 🍩

I tried asking AI for writing advice and it was so sarcastic I'm never going to ask again

being treated this way by a machine really made me want to aggressively fold origami swans

Martin Wattenberg reposted

Andrew Lee

@ajyl.bsky.social

New paper <3 Interested in inference-time scaling? In-context Learning? Mech Interp? LMs can solve novel in-context tasks, with sufficient examples (longer contexts). Why? Bc they dynamically form *in-context representations*! 1/N