safety

arXiv:2606.16272v2 Announce Type: replace Abstract: Human hand-object demonstrations provide dense reference motions for training dexterous manipulation reinforcement learning (RL) policies through re

arXiv:2606.20748v1 Announce Type: new Abstract: Humanoid robots are emerging as co-workers in smart manufacturing, yet their dynamic, human-like movements introduce safety risks that differ fundamenta

arXiv:2512.02581v3 Announce Type: replace Abstract: Reinforcement learning (RL) faces a persistent tension: policies that are stable to optimize (e.g., Gaussians) are often too simple to represent the

arXiv:2606.22550v1 Announce Type: new Abstract: Autoregressive visual models (AVMs) based on next-scale prediction have emerged as a prominent paradigm for image and video synthesis. However, decompos

arXiv:2606.23496v1 Announce Type: new Abstract: Discrete text-trigger optimization -- searching for text sequences that, when ingested by a model, steer it toward a specified objective -- underpins mo

arXiv:2606.13714v2 Announce Type: replace Abstract: Unsupervised video object-centric learning aims to decompose dynamic scenes into temporally persistent entity representations. Existing recurrent vi

arXiv:2606.20971v1 Announce Type: new Abstract: We introduce UNITY, a Universal-to-Specialized adapter for efficient and scalable composite conditioning in diffusion based image generation. Unlike pri

arXiv:2606.21287v1 Announce Type: new Abstract: Unsupervised domain adaptation (UDA) enables robust transfer of knowledge from simulated to real environments while exploiting a subset of unlabeled tar

arXiv:2602.11944v2 Announce Type: replace Abstract: When building AI systems for decision support, one often encounters the phenomenon of predictive multiplicity: a single best model does not exist; i

arXiv:2606.22870v1 Announce Type: new Abstract: Recent advancements in chain-of-thought (CoT) reasoning have shown promise in enhancing video understanding and reasoning capabilities of multimodal lar

arXiv:2606.21600v1 Announce Type: new Abstract: Multi-task robot manipulation policies are challenging to learn from demonstration because traditionally a single network must select among qualitativel

arXiv:2508.03058v2 Announce Type: replace Abstract: Reinforcement Learning (RL) in real-world environments often suffers from ambiguous or incomplete reward supervision, which undermines policy stabil

arXiv:2606.22136v1 Announce Type: new Abstract: Scaling dexterous manipulation requires generalization across objects, scenes, and tasks, yet existing data sources face a trade-off between scale and s

arXiv:2606.21126v1 Announce Type: new Abstract: Despite the success of deep learning, training deep networks in biologically plausible and hardware-efficient ways remains an open challenge. Feedback a

arXiv:2603.06445v2 Announce Type: replace Abstract: Situated reasoning often relies on active exploration, yet in many real-world scenarios such exploration is infeasible due to physical constraints o

arXiv:2606.16196v2 Announce Type: replace-cross Abstract: Deep neural networks have achieved remarkable performance across medical imaging tasks, yet their tendency to overgeneralize under distributio

arXiv:2603.28919v2 Announce Type: replace Abstract: As robots increasingly enter the workforce, human-robot interaction (HRI) must address how implicit social biases influence user preferences. This p

arXiv:2606.22145v1 Announce Type: new Abstract: Reinforcement learning (RL) is a powerful and convenient tool to modernize controller design. In this work, we study the zero-shot transfer of RL-based

arXiv:2606.19340v2 Announce Type: replace Abstract: We present ZeroDex, a zero-shot framework for long-horizon dexterous manipulation that grounds language instructions into executable 3D task plans f

An interesting new paper by my recent PhD graduate on how AI agents' greed for visible incentives can lead them to abandon their safety alignment. You can read it here: https://arxiv.org/abs/2606.1691

An update. A US official tells me that Sen. Warner misunderstood the NSA director Gen. Rudd in this case. Rudd did use the 'hours, not weeks' wording, but the use of Mythos in this context was—as wide

'Because I'm going to be smarter than you' This clip is from a new short film by @ForegoneFilms on the race to develop superintelligence. Please watch the full 16 minute film (link below) to help unde

By the end of the decade and maybe a lot sooner, we will wonder why we ever treated these folks like they were gods. “The more I listen to AI company CEOs, the stranger they sound,” says another tech

Great report on LLM agent communication protocols. Communication is a huge bottleneck in multi-agent systems. (worth bookmarking) The report builds a five-dimensional taxonomy (counterparty, payload,

have been thinking a bunch about model routing and related things current thoughts here, would love feedback: 1/ there is a difference between 'model routing' and 'model council' 'model routing' = rou

This newsletter issue covers three major AI topics: techniques for making AI systems more persuasive ('superpersuasion'), the concept of self-sustaining or self-improving AI systems, and various theor

LLMs can’t be trusted to follow rules. Which means they can’t be trusted, period. If we are going to solve alignment we must move on. Large language models can be persuaded to break their own rules. N

New piece in @TheAtlantic! We always hear that AI will cure cancer, and I would immediately benefit if it did. But I argue that racing ahead on generalist AI models creates unclear benefits for cancer

Nivida Corp. today announced Halos for Robotics, the industry’s first full framework for robotic safety systems that encompasses building, testing and managing complete artificial intelligence robotic

Ian King / Bloomberg: Nvidia unveils Halos, a safety-focused OS developed from autonomous vehicle tech and designed to run on IGX Thor hardware for humanoid robots, and opens a lab — Nvidia Corp. is w

Gary Marcus compares two countries' economic risk-taking, suggesting one is engaging in dangerous speculation while the other is not. The post likely discusses contrasting approaches to emerging techn

Our new incident response report exposes a coordinated network of 340 Facebook pages, largely operated from Vietnam, flooding feeds across Canada, the US, the UK, and Europe with AI-fabricated news, a

Amrith Ramkumar / Wall Street Journal: President Trump signs two executive orders aimed at speeding the development of advanced quantum computers and mitigating the security threats they present — Adm

Professor @GaryMarcus is an American psychologist, cognitive scientist, and author, known for his research on the intersection of cognitive psychology, neuroscience, and artificial intelligence. What

Sober minded folks need to understand the cash out happening in this bubble peak. Watch the @WSJ discussion 👇 How big tech hides the true cost of AI @wsj https://youtu.be/YrJzjC4kKCY?si=o1CiKxLaSVWOhK

The largest LLM-as-a-Judge reliability audit yet. Researchers ran 21 judges from nine providers over roughly 541,000 judgments on MT-Bench, JudgeBench, and RewardBench. Findings: Validating a judge wi

Gary Marcus posted a critical commentary on X regarding SPCX stock, sarcastically offering 'thoughts and prayers' to investors who purchased the stock at 225 in the previous week, implying the investm

This post by cognitive scientist Gary Marcus likely critiques the widespread credibility given to a particular public figure's predictions, questioning the basis for their influence or accuracy track

Absolutely insane revelations about Tulsi Gabbard. Throughout her public life, she's been a puppet for a Hindu cult. Washington Post got access to 25,000 pages of documents including directives from T

This video from the Wall Street Journal examines how major technology companies obscure the substantial economic, environmental, and societal costs associated with developing and operating artificial

'If curing cancer were the only result of building ever more powerful AI systems, I would cheer for their arrival. But the problem is that their impacts are much broader, and we are moving too quickly

Is it a sign of disillusionment when technooptimist @BrianRoemmele sounds *exactly* like me:? Here he writes that we risk an AI winter from “Over-promising relative to capabilities: Ambitious public c

this claim around creativity is too strong, IMHO. truly novel ideas from LLMs are surely rare but i don’t think that any math proves they are impossible. note that the objective function and the outpu

This post likely discusses how certain contemporary concepts, practices, or ideologies are actually rooted in or aligned with Buddhist philosophy, possibly challenging common misconceptions about thei

This now widely circulated claim is based on a line I wrote last week (https://economist.com/briefing/2026/06/14/donald-trumps-blocking-of-anthropic-is-capricious-and-chaotic). I accurately quoted Mar

about fucking time. they have always shown a disregard for humanity. 42 state attorneys general are investigating OpenAI over alleged harmful practices. Read again: 42 state attorneys general are inve