Tequehead @tequehead

My review of @odysseymovie in @nybooks now online nybooks.com/online/2026/07…

Real brains follow Dale's principle: a neuron can either excite its neighbors or suppress them, but never both. Standard deep learning ignores this and uses backpropagation. In our new paper, Diffusing Blame, we fix this disconnect. By introducing a routing method that broadcasts error signals directly to the hidden layers, we can train networks made of dedicated positive and negative neurons to strictly obey Dale's principle, all without relying on backprop! This method works surprisingly well on image recognition tasks despite the strict biological constraints. We also achieved competitive, backprop-free reinforcement learning on complex locomotion tasks and the open-ended Craftax environment. It is neat to see that representation learning remains possible even when we force deep learning to play by the rules of real neurons.

Sakana AI @SakanaAILabs

3 days ago

Introducing "Diffusing Blame": can a neural network learn competitively while strictly obeying Dale's principle, the rule that real neurons follow? We show it can, across both image classification and reinforcement learning. 🧠 Accepted at #ALIFE2026 arxiv.org/abs/2606.31700

9 136 845 194K 589

The most interesting thing about Agamemnon's death (he's stabbed in his bath while trapped in a net, meaning that he drowns, bleeds out and is strangled simultaneously) is that "threefold deaths" like this appear in the mythology of such a wide range of Indo-European cultures that it's generally assumed that there's some sort of proto-Indo-European ritual significance to this they're all drawing on and you start to wonder if there are any other much older elements that snuck into Greek mythology.

EdgyGookie 🇰🇷🇺🇸 @EdgyGookie

a day ago

Agamemnon really be getting unalived as soon as he brought glory and prosperity to his homeland 😭😭😭

8 0 58 159K 10

When Wikipedia is edited, real science is affected. The articles cited by it themselves get cited more, and it can even change the direction of innovation. Full paper below:

In intellectual stuff, most of the value is way out in the tail of unusually good talks. So I say: go for great, not mediocre.

Koenfucius 🔍 @koenfucius

2 days ago

“There is a limit to how charismatic you can appear when you are, at bottom, not charismatic.” A refreshing perspective by @rbnmckenna86—aim for mediocrity, and just give an average talk: buff.ly/Noh0iIh?

0 1 17 21K 11

.@DAMendelsohnNYC on Nolan’s unheroic Odyssey nybooks.com/online/2026/07…

What are the most cited reasons people have for opposing data centers in their area? The top three are water use, energy use, and utility bills. This suggests that the propaganda effort to worry people over nothingburgers has been successful!

Sustained molecular rejuvenation demonstrated in serially transplanted iHSCs (stem-cell-derived blood generating cells) -- a new preprint from @RetroBio_

The Geometry of Integration by Parts.

In many places in the English speaking world, an increasing share of 25-34 year olds live with their parents; especially people with a lower income. By @jburnmurdoch

Value creation and value capture are different things. Airlines transformed the world and never made money; the surplus went to passengers. Days after launching its best model, Anthropic includes it in existing plans. That is what competition does: the surplus goes to users, not to the labs. Profits will not justify these valuations.

Claude @claudeai

2 days ago

Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits. Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit. Demand for Fable has been challenging to

5K 6K 53K 25.5M 7K

An A+ manager increases salaries by 13%. And it turns out managers are best at boosting workers when the two are located in the same office. New research by @VirginiaMinni at U Chicago finds that workers with A+ managers are better at transitioning workers to positions that match their skillset, improving performance and pay. Many of the job switches A+ managers bring about involve real shifts in day-to-day tasks. That suggests these managers take the time to really get to know their employees, discover untapped abilities, and help them find a role that fits those skills. Remote work makes two parts of a manager's positive impact harder: 1. On a distributed team, it is harder for a manager to notice someone's talents in another area 2. Without being physically alongside other teams, it can be harder to understand their needs well enough to help place someone where they'd really thrive Indeed, Virginia finds the effects of great managers are strongest when managers and employees share an office.

Since nobody called it out in the comments, this “new way” to show 3D Gaussian splats is a basic 19th century construction, studied by Jakob Steiner and others, called a “sphere inversion” or “Möbius transformation”: en.wikipedia.org/wiki/Inversive… It looks good because it is an example of a so-called conformal map, which preserves angles, but distorts scales. However, when you look at something up close, it doesn't really matter what its absolute scale is (whereas angular distortion is far more noticeable). Conformal maps, namely the 16th century Mercator projection (old as it is) remains the projection of choice in Google Maps and other mapping tools, since it is the unique map that (i) preserves angles, and (ii) keeps north pointing "up" everywhere on the globe. The 3DGS scheme shown here is just bumping up the usual approach up one dimension: apply a conformal projection; keep the "up" direction fixed. The specific choice of conformal maps (inversions in the sphere) are in fact the only choice available in 3D, due to Liouville's theorem on conformal mappings. [Orange illustration below is from my PhD thesis on conformal geometry: cs.cmu.edu/~kmcrane/Proje…]

VoxelKei @VoxelKei

4 days ago

3D Gaussian Splattingの新しい見せ方を作りました。その名も Inversphere. 2Dのリトルプラネットと違い、空間ごと球面反転しています。ガラス玉に入れる従来の見せ方は球で「切り取る」のに対して、これは球の中へ「折り込む」ので360度の景色が無限遠まで立体で入っています。 #3DGS

147 915 7K 503K 3K

Only surprises carry new information. Yet most papers are written to confirm what their readers already believe. New in @ScienceMagazine with Casey Petroff and @kinggary: why scientists should design and write for surprise, and how to make framing part of formal inference. doi.org/10.1126/scienc…

Chunking enhances working memory capacity by compressing information directly within the current cognitive buffer rather than transferring data to long-term storage for retrieval. osf.io/preprints/psya…

Until 1991, India was the License Raj: If you wanted to operate a business there, you needed to obtain a license, and those took a long time and cost a lot. Perhaps one of the legacies of that era is that India is still the top-ranked country for workforce licensure.

Apropos nothing, here are some charts and graphs showing mean GRE scores by major, by intended graduate major, and average standard deviations from the mean by major. 🧵

keyboard_arrow_left Previous keyboard_arrow_right Next

All AI transfer learning in Go is within the first ~40 moves, and it's mostly memorization. Either engines didn't discover better fundamentals or humans can't learn from them if they did.

Noam Brown @polynoamial

2 months ago

After AlphaGo, the skill of human Go players noticeably improved. I suspect we will see a similar pattern in math.

187 962 9K 801K 2K

AI has helped resolve an important question in statistics. In the area of multiple hypothesis testing, the goal of controlling the false discovery rate (FDR) has been introduced in a seminal paper by Benjamini and Hochberg (1995). They also introduced a method (the Benjamini-Hochberg or BH method) and proved it controls the FDR. This method has been widely adopted in modern high-throughput science, including in genomics, astronomy, economics, etc. The paper has has garnered more than 130,000 citations to date. However Benjamini and Hochberg showed FDR control only when the data for the individual tests are *independent*. In practice, these data are often dependent; a good example is data on genetic variants due to linkage disequilibrium. Later work has focused on extending the validity of the BH procedure, e.g., to a form of positive dependence by Benjamini and Yekutieli (2001). The question of when the BH procedure controls the FDR has remained open. Over the last twenty years, many authors, including Reiner-Benaim (2007), Kim and van de Wiel (2008), Benjamini (2010), Sarkar (2023), Sarkar and Zhang (2025), have conjectured that the BH procedure controls the FDR for two-sided tests using any correlated Gaussian data. These authors have presented both theoretical and empirical evidence supporting, but not directly showing, the conjecture. With the help of AI (specifically GPT-5.6 Sol Pro), I have settled the question in the negative: The Benjamini-Hochberg procedure does *not* generally control the false discovery rate at the desired level for correlated two-sided Gaussian tests. This was done by exhibiting a Gaussian factor model for which, at a nominal level alpha=0.01, the false discovery rate is proved to be FDR>0.0104. There is a lot of interesting commentary to be made: 1. This result should be of interest to everybody in the field of statistics. Emmanuel Candes of Stanford University once called the false discovery rate and the Benjamini-Hochberg procedure "one of the two most important developments in statistics after 1950" (the other being James-Stein shrinkage). The present conjecture is probably the most central question about FDR/BH that was unresolved to date. 2. GPT-5.6 one-shot the problem after 90 minutes of reasoning, whereas with 5.5 I was not able to solve it even after iterating with multiple parallel agents for perhaps 20 hours. So the capability improvement is quite real. Exciting times to live in! 3. The argument is not especially surprising, but it does combine an asymptotic approach (standard for FDR analysis, see e.g., Genovese and Wasserman, Efron, etc) with a numerical certificate in a way that would be pretty non-standard in the field. Once we have the specific example, then straightforward simulations also support that the false discovery rate is indeed higher than the nominal value (see attached fig). 4. The current degree of violation over the nominal level is relatively small (0.104 vs 0.1). So the importance of this result is mainly conceptual. The practical implications remain to be determined. Overall, an exciting development! Preprint is available here (faculty.wharton.upenn.edu/wp-content/upl…) and will be on arxiv tonight; supporting code is here (github.com/dobriban/BH).

keyboard_arrow_left Previous keyboard_arrow_right Next

Researchers have calculated what would happen if a mirror were to vanish suddenly while a single photon’s wave packet was in the process of reflecting from it. The surprising answer: a quantum state of countless daughter photons. go.aps.org/4fgZIAx