Inteligencia Artificial Semanal

Robotaxis y drones, Chatbots en Retail, Ciberseguridad post-Mythos

Robotaxis and drones, Chatbots in Retail, Cybersecurity post-Mythos

Google SL2T, Watermarking de Anthropic, Publicidad en ChatGPT

Google SL2T, Anthropic Watermarking, Advertising in ChatGPT

AI en fabricación de Pringles, Persona 8B, World-Action Models

AI in Pringles manufacturing, Persona 8B, World-Action Models

Caso real de video-avatares, PixelRAG, LLMs muy humanos

Real case of video avatars, PixelRAG, very human LLMs

Flux-mimic de renderizador a simulador, De agentes y sandboxes, Medios y GoogleAI

Flux-mimic from renderer to simulator, Of agents and sandboxes, Media and GoogleAI

Kimi K3 y open weights vs closed weights, PrismML Bonsai 27B, Thinking Machines Inkling

Kimi K3 and open weights vs closed weights, PrismML Bonsai 27B, Thinking Machines Inkling

Training incluyendo fracasos, Mano robótica de 1X, Oak Lab de Richard Sutton

Training including failures, 1X robotic hand, Richard Sutton's Oak Lab

Metacognición, Series temporales probabilísticas, AdaJEPA

Metacognition, Probabilistic Time Series, AdaJEPA

Arquitecturas multimodelo, El problema de la memoria, Robots entrenados con videojuegos

Multimodel architectures, The memory problem, Robots trained with video games

AI en empresas, Errores humanos vs. errores AI, GLM-5.2

AI in companies, Human errors vs. AI errors, GLM-5.2

McDonals insiste con ArchIQ, DiffusionGemma, Manos robóticas de Agilink

McDonals insists with ArchIQ, DiffusionGemma, Agilink robotic hands

Fei-Fei Li sobre World Models, Gemma4 12b encoder-free, Cerebro humano y next token prediction

Fei-Fei Li on World Models, Gemma4 12b encoder-free, Human brain and next token prediction

Más emergencia por escala, Training de transformers por difusión, Los datos no se agotan

More emergency by scale, Training of transformers by diffusion, Data does not run out

Gestión de cambios de estado, Hierachichal Recurrent Model, Memory as a Model

State change management, Hierarchical Recurrent Model, Memory as a Model

Unitree IPO, Multi-Stream LLMs, Soluciones para grandes contextos

Unitree IPO, Multi-Stream LLMs, Solutions for Large Contexts

Subquadratic Sparse Attention, Natural Language Autoencoders, Coordinación de 2 robots Helix-02

Subquadratic Sparse Attention, Natural Language Autoencoders, Coordination of 2 Helix-02 robots

RLCR, Razonamiento en espacio latente, Bitter Lesson para agentes

RLCR, Reasoning in latent space, Bitter Lesson for agents

Ineffable con puro RL, Vision Banana, Thinking vs. pass/maj@k

Ineffable with pure RL, Vision Banana, Thinking vs. pass/maj@k

Gated Memory, Aprendizaje por descripción, Ising OS para computadores cuánticos

Gated Memory, Description-based Learning, Ising OS for Quantum Computers

Neural Computers, Saturación agéntica, Anthropic Mythos

Neural Computers, agentic saturation, Anthropic Mythos

PrismML modelos de 1 bit, Aprendizajes sobre LRMs, GEN-1 robots que improvisan

PrismML 1-bit models, Learnings on LRMs, GEN-1 robots that improvise

LeWorldModel, Máquinas de Darwin-Gödel universales, TurboQuant

LeWorldModel, Universal Darwin-Gödel Machines, TurboQuant

CompACT para robots, Epistemic Neural Networks para RLHF, Online Experiential Learning

AttnRes, Trusted Executor Dilemma, Autoresearch y AI recurrente

AttnRes, Trusted Executor Dilemma, Autoresearch and Recurrent AI

Training multimodal, Chip con neuronas humanas aprende a jugar a Doom, Hastío agéntico

Multimodal training, Chip with human neurons learns to play Doom, Agentic boredom

Personalidades de GenAI, C-JEPA, Máquinas de Gödel-Darwin para workflows agénticos

GenAI Personalities, C-JEPA, Gödel-Darwin Machines for agentic workflows

Nadie al volante en empresas, Experiential RL, Robots con tejidos blandos

Nobody at the wheel in companies, Experiential RL, Robots with soft tissues

ALMA: meta-agentes y conexionismo, Impacto laboral de AI, Primer desarrollo original de una AI

ALMA: meta-agents and connectionism, Labor impact of AI, First original development of an AI

SaaSpocalypse, Large Tabular Model, RLVR en dominios abiertos

SaaSpocalypse, Large Tabular Model, RLVR on open domains

Errores en GenAI, Helix 02 con Locomanipulation, OpenClaw y MoltBook

Errors in GenAI, Helix 02 with Locomanipulation, OpenClaw and MoltBook

Societies of thought, Sesgos de GRPO, Aprendizajes del uso de agentes

Societies of thought, GRPO biases, Lessons from the use of agents

Nada más práctico que una buena teoría: un ejemplo ilustrativo

Nothing more practical than a good theory: an illustrative example

Industrialización de ciberseguridad de ataque, Bitter Lesson para agentes y memoria, Habitabilida...

Industrialization of offensive cybersecurity, Bitter Lesson for agents and memory, Habitabilida...

AI en Retail y CPG, Competencia entre agentes evolutivos, DroPE y RLMs para contextos grandes

AI in Retail and CPG, Competition among evolutionary agents, DroPE and RLMs for long contexts

DeepSeek mHC, Cómo implantar AI en empresas con sentido común, Estructura de la materia y Bitter ...

DeepSeek mHC, How to implement AI in companies with common sense, Structure of matter and Bitter ...

MACLA memoria y aprendizaje continuo, Transferencia humano-robot, Problema y dato del problema

MACLA memory and continuous learning, human-robot transfer, problem and problem data

Simulador vs. oráculo, ARTEMIS y ciberseguridad, Conectoma artificial

Simulator vs. Oracle, ARTEMIS and Cybersecurity, Artificial Connectome

RL en pre-training, Subespacios universales en DNNs, Unsupervised Learning

RL in pre-training, Universal subspaces in DNNs, Unsupervised Learning

Ventajas de IBM Small Models, AI motores y caballos, Cuestionando la cross-entropy

Advantages of IBM Small Models, AI motors and horses, Questioning cross-entropy

Estudio MIT sobre AI y empleo, Estrategias Evolutivas para GenAI, LatentMAS

MIT study on AI and employment, Evolutionary Strategies for GenAI, LatentMAS