✨New Preprint ✨ How are shifting norms on the web impacting AI? We find: 📉 A rapid decline in the consenting data commons (the web) ⚖️ Differing access to data by company, due to crawling restrictions (e.g.🔻26% OpenAI, 🔻13% Anthropic) ⛔️ Robots.txt preference protocols

We’ve always seen shadows cast by clouds on the ground but this cloud is casting a shadow on the sky because the sun is “behind and under it” it. (Or atleast that’s what I think the streak is) Sunsets always present all sorts of interesting phenomena with light 🌄

📢 Who contributes most to politically polarized speech on Reddit and Twitter? As we come into the final stretch of the presidential election season, we present our work out of MIT Center for Constructive Communication a large-scale analysis (2.5 billion tweets/comments) of politically polarized speech.

My timeline is flooded with "hot" takes about Google's new quantum computer "killing Bitcoin." so here is a quick effort on my end to combat the fake news storm and why your crypto isn't about to vanish into the quantum realm 1/8 🧵

Project page: eyes.mit.edu Talk: youtube.com/watch?v=OuPwGk…

"What?" Reactions like this makes it all worth it🥲 Happy to answer any questions about our new work.

Yep ! We trained eyeballs from scratch, starting with just light-detecting photoreceptors. 🔬👁️ Why? To simulate vision evolution in-silico and understand why we perceive the world the way we do. 🌍✨ Check it out: eyes.mit.edu

An interesting AI hackathon is unfolding at MIT right now. Think Game of Thrones meets machine learning - where alliances and strategy matter as much as technical skills to build the best model.

GPS isn’t broken. It’s just vulnerable. Today, we’re proud to share a glimpse into what we’re building at Dirac Labs. A quantum sensor that listens to the Earth itself — no satellites required. Spoof-proof. Jam-resistant. Always on. 🌐 diraclabs.com

Reading a friend’s startup prompt logic felt like reading 2017 self-driving stacks—just layers of if/else rules and statements to steer the LM. The brittle edge-case handling lives on.

From my observations, it seems like Claude Code code exhibits strong variance suppression bias—errors are treated as unwanted variance, not diagnostic signals. This would explain the constant error/exception suppression, swallowing errors with pass, or print+pass to give the

Every lens leaves a blur signature—a hidden fingerprint in every photo. In our new #TPAMI paper, we show how to learn it fast (5 mins of capture!) with Lens Blur Fields ✨ With it, we can tell apart ‘identical’ phones by their optics, deblur images, and render realistic blurs.

Correct! Just as a reminder: this is what a Transformer found after looking at 10M solar systems

Should robots have eyeballs? Human eyes move constantly and use variable resolution to actively gather visual details. In EyeRobot (eyerobot.net) we train a robot eyeball entirely with RL: eye movements emerge from experience driven by task-driven rewards.

From what I remember my grandfather telling me, sometimes happens because multiple busses compete with each other on the same route and each bus is private. This creates a race condition on which driver reaches the stop first because they’ll get the most passengers. So some