🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

I think it's kinda cool that humans have computationally intractable preferences like "wanting to know the value of BB(5)"!

Beginning in the spring of 2022, collaborators tackled a stubborn theoretical computer science problem called BB(5). This is the moment when mxdys, an anonymous user on the group’s Discord server, solved the problem using a proof assistant program. quantamagazine.org/amateur-mathem…

A minor delight among the many fascinating things about Turing machines: the degree to which the central diagram here looks like a mountain in the background of an SNES game.

In February, four computer scientists set out to develop an algorithm for simulating quantum systems. Along the way, they accidentally proved that entanglement in those systems vanishes completely above a certain temperature. My latest for Quanta Magazine: quantamagazine.org/computer-scien…

New reigning champ in this strain of bot names. I'm imagining it being uttered by a grizzled police sergeant in a 90s TV show set in Chicago.

We’ve spent months putting together an immersive special issue of Quanta Magazine that explores the ultimate scientific quest: the search for the fundamental nature of reality. Don’t miss it! quantamagazine.org/the-unraveling…

Ask us anything you want about physics! Our senior editor Natalie Wolchover and writer Charlie Wood are hosting a Reddit AMA today at 1:30 PM. reddit.com/r/IAmA/comment…

For Quanta Magazine, I wrote about a new synthesis of the two most remarkable proof methods in computer science. The new technique keeps the details of a correct proof absolutely secret, yet instantly broadcasts any error in a flawed one: quantamagazine.org/computer-scien…

John Hopfield and Geoffrey Hinton have won this year’s The Nobel Prize in physics for developing methods that have shaped modern powerful machine learning algorithms. The unusual choice highlights the way that physics inspired some of the earliest neural networks. 🧵

Far too soon to declare 'quantum victory' in the quantum-classical war, but it does seem like there is at this point a clear winner in the battle over random circuit sampling—and better yet, and understanding of the conditions needed for victory. nature.com/articles/d4158…

My latest in Quanta Magazine: the best known method for finding shortest paths through a network is even better than researchers thought. It’s a surprising new discovery about an algorithm that's nearly 70 years old. quantamagazine.org/computer-scien…

Out of all possible Turing machines with n rules, the 'Busy Beaver' is the one that runs longest. BB(4), discovered in '66, runs for 104 steps. It's taken >50 years to prove BB(5) runs for 47,176,870! Ben Brubaker with one of 2024's best maths stories quantamagazine.org/amateur-mathem…

Today in Nature, the first experimental demonstration that scaling up a quantum error-correcting code makes it better. It's a critical step on the path to useful quantum computers. Read more in Quanta Magazine: quantamagazine.org/quantum-comput…

I was thrilled to hear about the surface code when Alexei Kitaev visited Caltech in April 1997. It's heartening to see the hardware catching up with the theory of quantum error correction. There's still far to go but encouraging progress Google Quantum AI. quantamagazine.org/quantum-comput…

This year, researchers got a better look at chatbots’ thoughts, amateurs learned just how complicated simple systems can be, and codes became expert self-fixers. Read editor Bill Andrews’ (Bill Andrews) list of the year’s biggest developments in theoretical computer science:

My latest in Quanta Magazine: why theoretical computer scientists like to pose questions to imaginary black boxes: quantamagazine.org/why-computer-s…

Can neural networks learn languages that are impossible for humans? My latest story for Quanta Magazine explores how making language models worse at learning might make them better tools for linguists: quantamagazine.org/can-ai-models-…

In this episode of Polylogues, Ben Brubaker and Boaz Barak discuss Boaz’s path TCS, theory and practice in cryptography and AI, and specific topics in AI, including generality, alignment, worst- vs. average-case performance, and watermarking. youtu.be/rM1PeAWDNWQ?si…