Search: “safety”

The NSF helped build the internet, funded the observatory that detected gravitational waves, laid foundations for AI, advanced MRI technolog…

The NSF helped build the internet, funded the observatory that detected gravitational waves, laid foundations for AI, advanced MRI technology, and trained generations of scientists. It is on track for

20 Jun 2026

Safety

There is no way the hyperscalers are going to make back their $5.3 trillion — unless they take it out of the taxpayers, via massive governme…

DGX agent

There is no way the hyperscalers are going to make back their 5.3 trillion — unless they take it out of the taxpayers, via massive government subsidies. Which is *exactly* what they are going to try t

safetygary-marcus--x

20 Jun 2026

🚨 This is how you turn a stock market crash into great depression.

This post likely discusses economic policy errors and systemic vulnerabilities that could transform a stock market correction into a severe, prolonged economic depression similar to the Great Depressi

20 Jun 2026

C'est un grand honneur d’avoir reçu cette semaine l'insigne d'officier de l'Ordre national du Québec. Merci @CFrechette pour la remise de ce…

C'est un grand honneur d’avoir reçu cette semaine l'insigne d'officier de l'Ordre national du Québec. Merci @CFrechette pour la remise de cette distinction. Il reste beaucoup de chemin à parcourir pou

safetyyoshua-bengio--x

dot-com bubble vs. a possible AI bubble. From the famous 'Dean of Valuation', Professor Aswath Damodaran, of NYU Stern School of Business, “…

dot-com bubble vs. a possible AI bubble. From the famous 'Dean of Valuation', Professor Aswath Damodaran, of NYU Stern School of Business, “And that’s the real big difference between the dot-com boom

Eight reasons to slow down hyperscaling: • What GenAI has already done to society • What AI slop is doing to the internet • What data center…

Eight reasons to slow down hyperscaling: • What GenAI has already done to society • What AI slop is doing to the internet • What data center overbuild is likely to do the economy • What data center ov

'If Bores wins the primary, it has the potential to clarify and perhaps even transform the federal debate about regulating artificial intell…

'If Bores wins the primary, it has the potential to clarify and perhaps even transform the federal debate about regulating artificial intelligence. On the other hand, if he loses, Bores and others fea

Only a narrative violation if you believed people with a lot of money at stake. • Most of the people that said lots of jobs would soon disap…

Only a narrative violation if you believed people with a lot of money at stake. • Most of the people that said lots of jobs would soon disappear were running LLM companies trying to overhype their war

“1.75 trillion dollar” company can’t do diddly-squat over impersonating accounts running (i have seen three in three days) like these stock …

“1.75 trillion dollar” company can’t do diddly-squat over impersonating accounts running (i have seen three in three days) like these stock scams and yet is supposedly gonna make orbiting data centers

A Survey of Reasoning and Agentic Systems in Time Series with Large Language Models

arXiv:2509.11575v3 Announce Type: replace Abstract: Time series reasoning treats time as a first-class axis and incorporates intermediate evidence directly into the answer. This survey defines the pro

A theory of learning data statistics in diffusion models, from easy to hard

arXiv:2603.12901v2 Announce Type: replace-cross Abstract: While diffusion models have emerged as a powerful class of generative models, their learning dynamics remain poorly understood. We address thi

A true honor to be listed in this sentence with @demishassabis. ❤️🙏

Gary Marcus expresses appreciation for being mentioned alongside Demis Hassabis, the co-founder of DeepMind, in what appears to be a recognition or honor-related context. The post reflects Marcus's re

Adapting Vision-Language Models from Iconic to Inclusive for Multi-Label Recognition Without Labels

arXiv:2606.11626v1 Announce Type: new Abstract: Understanding multi-label images remains a challenging task in computer vision. With the rapid progress of vision-language multimodal learning, vision-l

Adv-TGD: Adversarial Text-Guided Diffusion for Face Recognition Impersonation Attacks

arXiv:2606.11615v1 Announce Type: new Abstract: The widespread adoption of face recognition (FR) technologies raises serious privacy concerns, as facial data can be exploited without consent. To addre

Adversarial Attacks on Learned Policies for Surgical Robotic Tasks

arXiv:2606.11535v1 Announce Type: new Abstract: Learning-based policies are being considered to augment the dexterity of human surgeons in robot-assisted surgery. Can the end-to-end mapping from visua

safetyarxiv-cs-ro

An Ethical eValuation Agent (EeVA): Results of a Proof-of-Concept Test on a Prototype Agentic-like Workflow to Assist Ethical Deliberations

arXiv:2606.11218v1 Announce Type: cross Abstract: Ethical deliberation is often misunderstood as a search for single right or wrong answers, creating difficulties for non-ethically trained personnel w

and now the possible of drastic price cuts, a further sign of OpenAI’s desperation, as foreseen by me 2.5 years ago:

and now the possible of drastic price cuts, a further sign of OpenAI’s desperation, as foreseen by me 2.5 years ago: Called the LLM price wars – which are about to heat up — over two years ago. See es

APPO: Agentic Procedural Policy Optimization

arXiv:2606.12384v1 Announce Type: cross Abstract: Recent advances in agentic Reinforcement Learning (RL) have substantially improved the multi-turn tool-use capabilities of large language model agents

APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies

arXiv:2606.12366v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models that couple pretrained Vision-Language Models (VLMs) with continuous action experts have achieved strong manipulatio

safetyarxiv-cs-ro

Architecture-Aware Reinforcement Learning Makes Sliding-Window Attention Competitive in Math Reasoning

arXiv:2606.11634v1 Announce Type: new Abstract: The rapid progress of reasoning and agentic large language models (LLMs) has increased the demand for long-context inference, but self-attention (SA) sc

As featured today by @FT’s Alphaville: “An entire industry is being propped up by math that is insane.” Available only at Marcus on AI, a “f…

Gary Marcus critiques fundamental mathematical or computational problems underlying an entire industry, suggesting widespread reliance on flawed or unsound methodologies. The commentary appears on Mar

Auditing Demographic Bias in Facial Landmark Detection for Fair Human-Robot Interaction

arXiv:2604.06961v2 Announce Type: replace Abstract: Fairness in human-robot interaction critically depends on the reliability of the perceptual models that enable robots to interpret human behavior. W

Augmenting Molecular Language Models with Local n-gram Memory

arXiv:2606.12113v1 Announce Type: cross Abstract: Transformer-based language models for SMILES strings suffer from a locality gap: standard character-level tokenization fragments chemically meaningful

AVIS: Adaptive Test-Time Scaling for Vision-Language Models

arXiv:2606.11576v1 Announce Type: cross Abstract: Modern Vision-Language Models (VLMs) benefit from chain-of-thought prompting and test-time scaling, but these gains often come with prohibitive infere

Bad things apparently come in 4’s; tweet below must be updated with the breaking @wsj news that OpenAI is contemplating drastic price cuts, …

Bad things apparently come in 4’s; tweet below must be updated with the breaking @wsj news that OpenAI is contemplating drastic price cuts, which is surely a sign of weakness. So far today • Banks to

Beyond representational alignment with brain-guided language models for robust reasoning

arXiv:2606.11893v1 Announce Type: cross Abstract: The correspondence between large language models (LLMs) and the neural mechanisms underlying human higher-order cognition remains insufficiently chara

Beyond Third-Person Audits: Situated Interaction Auditing for User-Centered LLM Bias Research

arXiv:2606.12247v1 Announce Type: cross Abstract: Research on bias in large language models (LLMs) has predominantly focused on third-person audits, which study how models represent or evaluate demogr

Blind Dexterous Grasping via Real2Sim2Real Tactile Policy Learning

arXiv:2606.11767v1 Announce Type: cross Abstract: Blind grasping with a dexterous hand is a crucial manipulation capability. Nevertheless, learning such tactile-only policies for real robots remains c

🚨 Breaking neuroscience/evolution news! New Nature paper provides further evidence for a central hypothesis (with a long lineage) that I de…

🚨 Breaking neuroscience/evolution news! New Nature paper provides further evidence for a central hypothesis (with a long lineage) that I defended in my 2004 book, The Birth of the Mind: the evolutiona

Bridging Day and Night: Unsupervised Cross-Domain Re-Identification with Synergistic Prompt and Prototype Learning

arXiv:2606.12258v1 Announce Type: new Abstract: Cross-domain day-night re-identification (ReID) is fundamentally challenged by the substantial visual appearance discrepancies between daytime and night

Called the LLM price wars – which are about to heat up — over two years ago. See especially predictions 3, 4, and 7, which laid out the no m…

Called the LLM price wars – which are about to heat up — over two years ago. See especially predictions 3, 4, and 7, which laid out the no moat, small profit, price war regime we have seen ever since.

Can AI Reason Like an Urban Planner? Benchmarking Large Language Models Against Professional Judgment

arXiv:2606.11678v1 Announce Type: new Abstract: Problem, Research Strategy, and Findings: The rise of large language models (LLMs) raises a key question for urban planning: which forms of professional

Compatibility-Aware Dynamic Fine-Tuning for Large Language Models

arXiv:2606.11206v1 Announce Type: new Abstract: Supervised Fine-Tuning (SFT) is the predominant paradigm for aligning large language models (LLMs), yet it suffers from optimization instability and lim

Corpus Augmentation for Sign Language Translation via LLM-Guided Video Stitching

arXiv:2606.11925v1 Announce Type: new Abstract: Sign language translation (SLT) converts sign language video into spoken language text and holds significant promise for improving accessibility and ena

Critic Architecture Matters: Dual vs. Unified Critics for Humanoid Loco-Manipulation

arXiv:2606.11891v1 Announce Type: cross Abstract: Multi-objective reinforcement learning for humanoid robots must coordinate locomotion and manipulation within a single policy. A natural design choice

DeMix: Debugging Training Data with Mixed Data Error Types by Investigating Influence Vectors

arXiv:2606.11616v1 Announce Type: new Abstract: High-quality training data is essential for the success of machine learning models. However, real-world datasets often contain mixed types of errors ari

Deterministic Policy Gradient for Learning Equilibrium in Time-Inconsistent Control Problems

arXiv:2606.11798v1 Announce Type: cross Abstract: In this paper, we develop a continuous-time model-free reinforcement learning algorithm to learn deterministic equilibrium policies in general time-in

DiffCold: A Diffusion-based Generative Model for Cold-Start Item Recommendation

arXiv:2606.12245v1 Announce Type: cross Abstract: Cold-start item recommendation remains a persistent challenge in real-world systems due to the absence of interaction histories. While prior models at

Diffusing to Coordinate: Efficient Online Multi-Agent Diffusion Policies

arXiv:2602.18291v2 Announce Type: replace Abstract: Online Multi-Agent Reinforcement Learning (MARL) is a prominent framework for efficient agent coordination. Crucially, enhancing policy expressivene

Dynamic Execution Horizon Prediction for Chunk-based Robot Policies

arXiv:2606.11408v1 Announce Type: new Abstract: Action chunking has become a standard design in modern robot policies, from diffusion/flow policies to vision-language-action models, where the policy p

safetyarxiv-cs-ro

End-to-End Machine Learning for Depressive State Classification via EEG and fNIRS

arXiv:2606.11555v1 Announce Type: cross Abstract: The escalating demand for mental healthcare, driven by rising societal stress, highlights the limitations of traditional psychiatric diagnostics. Conv

Energy-Conserved Neural Pipelines: Attenuating Error Propagation in Modular Neural Networks via Physical Conservation Constraints

arXiv:2606.11341v1 Announce Type: new Abstract: Modular neural network pipelines suffer from error compounding: noise at any module boundary propagates and potentially amplifies through subsequent mod

Energy Use of AI Inference, Efficiency Pathways, and Test-Time Scaling

arXiv:2509.20241v2 Announce Type: replace Abstract: As AI inference scales to billions of queries, estimates of per-query energy use are increasingly important for capacity planning, efficiency interv

Enhancing Spectral Embedding through Robust and Flexible Knowledge Transfer in Electronic Health Records

arXiv:2606.11570v1 Announce Type: cross Abstract: We propose a spectral-based, unsupervised representation learning framework to derive low-dimensional embeddings for clinical concepts and patients in

Evaluating Bias in Phoneme-Based Automatic Speech Recognition Systems: An Analysis of IPA Transcription Models

arXiv:2606.11639v1 Announce Type: new Abstract: The popularization of automatic speech recognition (ASR) systems has increased exploration of the demographic biases related to race, age, gender, and a

Existential Indifference: Self-Nonpreservation as a Necessary Architectural Condition for Aligned Superintelligence (or: The Suicidal AI)

arXiv:2606.12032v1 Announce Type: new Abstract: Contemporary AI alignment research treats self-preservation as an instrumental nuisance to be suppressed by external mechanisms. We argue the framing is

External Experience Serving in Production LLM Systems: A Deployment-Oriented Study of Quality-Cost Trade-offs

arXiv:2606.11806v1 Announce Type: new Abstract: Production LLM systems accumulate reusable operational experience, but the practical deployment issue is not merely whether such experience can help. It

FACTR 2: Learning External Force Sensing for Commodity Robot Arms Improves Policy Learning

arXiv:2606.12406v1 Announce Type: cross Abstract: Contact-rich manipulation requires force sensitivity, but many robot arms lack dedicated force sensors due to their high cost. We present Neural Exter