ASH: New AI System Learns Complex Tasks from Unlabeled Internet Video Without Expert Guidance
Researchers introduced ASH, an AI agent that learns to play complex video games by watching unlabeled internet videos and improving itself through a self-honing loop, without requiring hand-engineered rewards or expert demonstrations. The system uses an Inverse Dynamics Model to extract learning signals from relevant video footage and maintains long-term memory of key moments to tackle multi-hour planning tasks. ASH significantly outperformed existing baselines on two demanding games—Pokemon Emerald and Legend of Zelda—suggesting that self-improving agents could be a scalable approach for embodied AI learning.
ASH is an agentic system designed to address a fundamental challenge in embodied AI: learning long-horizon tasks without relying on hand-engineered rewards or expert-annotated demonstrations, which do not scale well. The system operates through a self-improvement loop where, when stuck, it learns an Inverse Dynamics Model from its own trajectories and uses this model to extract supervision signals from relevant internet video. ASH employs unsupervised learning to identify and retain key moments from large-scale internet video as long-term memory, enabling it to handle extended planning problems. Evaluated on two complementary environments—Pokemon Emerald (turn-based RPG) and Legend of Zelda: The Minish Cap (real-time action game)—ASH sustained progression across 8-hour evaluations, reaching 11.2/12 and 9.9/12 milestones respectively, while the strongest baselines plateaued at 6.5/12 and 6.0/12. The results demonstrate that self-improving agents represent a scalable recipe for learning complex embodied tasks from unlabeled data.
What's missing
The paper does not discuss potential limitations of the approach, such as failure modes when the Inverse Dynamics Model produces poor supervision signals, computational costs of the self-improvement loop, or generalization to real-world robotic tasks beyond simulated game environments. The study also does not address how the method would perform on tasks where internet video examples are sparse or unavailable.
What different sources said
- arXiv cs.AICenter
ASH: Agents that Self-Hone via Embodied Learning
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.