safety

arXiv:2605.22596v1 Announce Type: new Abstract: Robotic tasks are typically specified by a tuple of factors, such as the object to be grasped, the obstacles to be avoided, the color of the target, and

arXiv:2605.22340v1 Announce Type: new Abstract: Single-cell RNA sequencing (scRNA-seq) provides high-dimensional profiles of cellular states, enabling data-driven modeling of cellular dynamics over ti

@GaryMarcus If OpenAI gets bailed out, I will lobby with all of my effort to ensure whatever political faction was responsible gets voted out. I'm about to become a single issue voter over AI because

arXiv:2605.22507v1 Announce Type: new Abstract: We propose a new framework for generative modeling based on a discrete-time stochastic control formulation of measure transport. Adapting classic result

arXiv:2605.21516v1 Announce Type: new Abstract: Harness engineering has emerged as an important inference-time technique for large language model (LLM) agents, aiming to improve long-term performance

arXiv:2602.05286v3 Announce Type: replace Abstract: Healthcare facility visit prediction is essential for optimizing healthcare resource allocation and informing public health policy. Despite advanced

Here’s a key line in this mythos update. This is precisely an example of why engineers don’t go away, ever. We’ve made it far easier to create and find security issues, which means the new bottleneck

arXiv:2603.02604v2 Announce Type: replace Abstract: We introduce Heterogeneous Agent Collaborative Reinforcement Learning (HACRL), a new Reinforcement Learning from Verifiable Reward (RLVR) problem th

I am deleting a question I had about @theo in light of his clarification. I hope that he and others will reconsider his/their own slander in light of my own clarifications, such as the one below that

I love that this post, my most foulmouthed ever, has gotten over 100k views. You want to know why it has spread? Simple. Because a lot of people have become really fed up with OpenAI. Fuck this OpenAI

I once asked a bunch of e/accs how much damage was an acceptable risk relative to take (any) safety precautions. none answered. we may be about to find out. One implication of the below is that we rea

Gary Marcus expresses concern that AI-generated low-quality mathematical content ('AI slop') will proliferate, making the role of professional mathematicians increasingly important for quality control

it ain’t just me who sees the emperor has no clothes The AI bubble math doesn't add up. Anthropic spends 3 to make 1 and that’s before you include any and all other costs like staff or electricity. Mi

it’s truly astonishing to see the entire field move in the neurosymbolic and world model directions I advocated in 2019 and 2020 and then simultaneously see people claim (invariably without specific e

This essay discusses 'vibe slop,' a term describing low-quality AI-generated content that lacks substance but mimics the aesthetic and emotional tone of authentic creative work. The piece likely argue

arXiv:2605.22207v1 Announce Type: cross Abstract: Safety has been a major concern when deploying deep reinforcement learning algorithms in the real world. A promising direction that ensures that the l

arXiv:2605.22717v1 Announce Type: cross Abstract: Interactive streaming music generation promises the use of generative models for live performance and co-creation that is impossible with offline mode

arXiv:2605.22291v1 Announce Type: new Abstract: Long-term fairness algorithms aim to satisfy fairness beyond static and short-term notions by accounting for the dynamics between decision-making polici

arXiv:2605.22775v1 Announce Type: new Abstract: Real-time cognitive load assessment from eye-tracking signals could potentially enable adaptive human-centered-AI such as safety-critical applications s

arXiv:2603.19310v3 Announce Type: replace Abstract: Reinforcement learning has emerged as a powerful paradigm for improving large language model (LLM) reasoning, where rollouts are sampled from the po

arXiv:2605.21722v1 Announce Type: cross Abstract: Sampling from discrete distributions with multiple modes and energy barriers is fundamental to machine learning and computational physics. Recent disc

my guess as to why: the OpenAI crowd is suddenly coming after me relentlessly, but too chicken to actual face me in a moderated debate. here’s why: a. the IPO is coming, but OpenAI’s has lost their le

NEW: Google claimed this week that a team of agents had built an entire operating system, based on a single prompt, costing only about $900 in tokens. We fact check this claim and analyze what it mean

arXiv:2605.21561v1 Announce Type: new Abstract: Unsupervised feature selection is commonly formulated as a multiobjective optimisation problem that jointly optimises subset quality and subset size. Ye

arXiv:2605.21834v1 Announce Type: new Abstract: Aligned models can misbehave in several ways: they are often sycophantic, fall victim to jailbreaks, or fail to include appropriate safety warnings. Con

arXiv:2605.21763v1 Announce Type: new Abstract: We study risk-sensitive reinforcement learning in finite discounted MDPs, where a generative model of the MDP is assumed to be available. We consider a

arXiv:2605.22156v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has become a promising paradigm for scaling reasoning capabilities of Large Language Models (LLMs)

Gary Marcus humorously highlights the massive valuation gap for OpenAI, suggesting that the company would need to generate approximately 5 trillion dollars in profit to justify a 4 trillion dollar val

OpenAI’s “Roon” is officially a chicken, just as I thought. And just like his boss (who has also ducked debates with me). If either actually thought they could dismantle my arguments, it certainly *wo

arXiv:2605.21752v1 Announce Type: new Abstract: Recommender systems trained on user interaction data are susceptible to behavioral intensity imbalance--a systematic distortion arising from heterogeneo

arXiv:2605.22731v1 Announce Type: new Abstract: Large language model post-training methods such as supervised fine-tuning (SFT), reinforcement learning (RL), and distillation are often analyzed throug

arXiv:2605.22738v1 Announce Type: new Abstract: Shapley and Banzhaf interactions capture the complex dynamics inherent in modern machine learning applications. However, current estimators for these hi

New York Times: Q&A with Sundar Pichai on the future of Google Search, Google's place in the AI race, public skepticism toward AI, AI agents, AI safety, TPUs, and more — After a busy Google I/O, the c

Quarter million views and counting. OpenAI has no idea how many people they have pissed off. Fuck this OpenAI employee, seriously fuck him. Also read this quantitative study, which I had nothing to do

arXiv:2603.03454v2 Announce Type: replace Abstract: Offline Reinforcement Learning (RL) is an emerging field of RL in which policies are learned solely from demonstrations. Within offline RL, some env

receipt confirming the chicken part of the above: OpenAI’s “Roon” is officially a chicken, just as I thought. And just like his boss (who has also ducked debates with me). If either actually thought t

Gary Marcus references a Nautilus article arguing that deep learning is encountering fundamental limitations, suggesting readers can find supporting evidence and detailed arguments for this perspectiv

Retweeting this because there appears to be a coordinated and intellectually dishonest campaign to defame me by misrepresenting my beliefs. The case against me below is completely intellectually disho

arXiv:2602.10894v2 Announce Type: replace Abstract: Two-player games such as board games have long been used as traditional benchmarks for reinforcement learning. This work revisits a policy optimizat

Sam Altman is rushing to an IPO for OpenAI... ...but if the public markets treat @OpenAI like a traditional business instead of a religion, it's game over. 💀💀💀 @sama #openai #chatGPT credit: @Stylosa

serious escalation in Trump’s growing challenges with inhibiting his behaviors. neurologists know what that means. Trump posts AI video depicting him throwing Colbert in a dumpster and dancing https:/

six years ago world (cognitive) models were the centerpiece of my essay The Next Decade in AI. their time is finally coming. Demis Hassabis on the limit in today’s AI: language can describe the world,

arXiv:2511.04838v2 Announce Type: replace Abstract: Molecular property regression struggles with cases in chemically relevant target ranges that are underrepresented in datasets. Standard average erro

arXiv:2605.21736v1 Announce Type: cross Abstract: Logged advertising auctions make offline reserve-price evaluation attractive but risky. Replay tables can identify policies with large apparent yield

arXiv:2605.22263v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) is an emerging LLM post-training paradigm in which the model serves as its own teacher: conditioned on privileged inf

arXiv:2605.22376v1 Announce Type: new Abstract: Cross-domain offline reinforcement learning (CDRL) aims to improve policy learning in a target domain by leveraging data collected from a source domain.