Holy smokes! Opus 4.6 set a new record on the Remote Labor Index! At 4.17%. Anyone who claims that we are close to AGI is either lying or lost.
RESEARCH
-
Face Patches Implement Domain-Specific and Domain-General Processing
By
–
We show face patches implement the following code through recurrent dynamics: Detect face If (face found) Discriminate face else Continue to detect face IMHO, our paper conclusively resolves a debate that has raged since I was a graduate student, about whether face patches are specialized for processing faces or not. It turns out domain-general folks were right early on, domain-specific folks were right later in the response. So proud of @Yuelin_Shi and the entire team! Yuelin Shi (@Yuelin_Shi) Our paper is now out! nature.com/articles/s41586-0… A big question: 1) Is IT cortex well described as a general-purpose feedforward DNN? OR 2) Are face patches genuinely specialized for processing faces? Read on to find out the answer. (1/N) — https://nitter.net/Yuelin_Shi/status/2039100718100185426#m
→ View original post on X — @berkeley_ai, 2026-03-31 22:17 UTC
-
Bio-Inspired Modular Robot Snails Team Up for Complex Tasks
By
–
#Robot Snails That Team Up: Bio-Inspired Modular Bots Form Arms, Climb Obstacles, and Tackle Big Tasks
— Ronald van Loon (@Ronald_vanLoon) 31 mars 2026
by @lukas_m_ziegler
#Innovation #EmergingTech #Technology #Tech pic.twitter.com/dupybZWGpd#Robot Snails That Team Up: Bio-Inspired Modular Bots Form Arms, Climb Obstacles, and Tackle Big Tasks
by @lukas_m_ziegler #Innovation #EmergingTech #Technology #Tech -
H Company Releases Holo3, Outperforming GPT-5.4
By
–
H Company released Holo3, a new series of SOTA "Computer Use" models that outperform GPT-5.4 and Opus 4.6 on OSWorld-Verified and other benchmarks.
-
3D Gaussian Splats Improvement with MRNF and PPISP
By
–
3d gaussian splats suddenly looking way less splatty! high frequency details look so much better with MRNF + PPISP https://t.co/hMBrdgWsyB
— Bilawal Sidhu (@bilawalsidhu) 31 mars 20263d gaussian splats suddenly looking way less splatty! high frequency details look so much better with MRNF + PPISP Tavius Koktavy 🌐💿🔆 (@TaviusKoktavy) In @lichtfeldstudio the new MRNF strategy plus PPISP is great. Compare MCMC (left) to MRNF (right) on a quick 20 second scan. Full-screen recommended Fine details pop, backgrounds are much more clear, and the color space is true-to-life. — https://nitter.net/TaviusKoktavy/status/2039061893722091533#m
→ View original post on X — @bilawalsidhu, 2026-03-31 21:00 UTC
-
Reinforcement Learning Limitations on Fine-tuned Model Prompts
By
–
No, RL doesn't fix it. It merely makes e smaller for prompts present in the fine-tuning set.
-
Experts underestimate AI’s potential economic growth impact
By
–
Fascinating work on how different groups of experts expect GDP to change in the coming years. Kudos to @connacher_ , @pawtrammell , @BasilHalperin of the Stanford @DigEconLab and the others who did this data collection and analysis. @PTetlock always does such amazing work. My take? Most of the experts surveyed (I was one of them) are not optimistic enough about growth. Forecasting Research Institute (@Research_FRI) We completed the most comprehensive study of how economists and AI experts think AI will affect the U.S. economy. They predict major AI progress—but no dramatic break from economic trends: GDP growth rates similar to today's and a moderate decline in labor force participation. However, when asked to consider what would happen in a world with extremely rapid progress in AI capabilities by 2030, they predict significant economic impacts by 2050: • Annualized GDP growth of 3.5% (compared to 2.4% in 2025) • A labor force participation rate of 55% (roughly 10 million fewer jobs) • 80% of wealth held by the top 10% (highest since 1939) 🧵 Here's what we found: — https://nitter.net/Research_FRI/status/2038965685431259520#m
-
Autoregressive Models Error Propagation in Discrete Sequences
By
–
That's a ridiculous argument. – all auto-regressive models diverge, whether they are generative (in input space) or not. – for discrete symbol sequences, the probability of correctness decreases exponentially with the sequence length, assuming independence of errors. – THAT
-
Clarifying the utility debate around large language models
By
–
I never said LLMs were not useful. We're discussing a different question here.
-
Error Recovery Impossibility in Autoregressive Models
By
–
You didn't understand either. Yes, the independence of errors is an assumption, which may or may not be reasonable. No, errors are NOT RECOVERABLE in an auto-regressive setting because the set of correct answers form a subtree in the tree of all possible sequences. Once you get