Tag

#reinforcement learning

17 stories taggedreinforcement learning.

Illustration: A vast server room with long rows of glowing blue and white rack-mounted computers receding into the distance
Explained

Los investigadores de Apple desarrollaron una forma más inteligente de entrenar agentes de IA. Así es por qué los detalles aburridos importan.

Un nuevo método de entrenamiento de Apple ML Research llamado RISED enseña a los agentes de IA a enfocarse en las lecciones que aún no han aprendido, omitiendo ejercicios que son demasiado fáciles o demasiado difíciles. Es discretamente importante para cualquiera que le importe hacia dónde se dirige la IA de propósito general.

4 min read
A glowing neural network diagram rendered as a physical object, like frosted glass nodes connected by thin light-threads, sitting on a dark matte surface
Science & Space

Los Investigadores de Apple Enseñaron a una IA a Aprender de sus Propios Fracasos Escribiéndose una Nota de Una Línea

Una nueva técnica llamada RLTL;DR permite que un modelo de IA resuma por qué falló, para luego aplicar esa lección en su siguiente intento. Podría desbloquear la automejora en problemas tan difíciles que nunca se ha demostrado solución alguna.

3 min read
Illustration: A dense grid of identical robotic arms arranged in rows on a dark industrial surface
Robotics

Cómo Warp de Nvidia te permite ejecutar 2.048 simulaciones de robots simultáneamente

Un nuevo tutorial de Hugging Face muestra cómo trasladar una simulación de robot estándar desde una única CPU a miles de entornos paralelos en GPU, reduciendo significativamente el tiempo necesario para enseñar habilidades físicas a los robots.

4 min read
An AI research lab where engineers monitor training progress on large displays showing algorithm efficiency metrics, with whiteboards covered in technical notat
Explained

Apple researchers found a smarter way to train AI language models that generate text step by step

A new technique called DACA-GRPO fixes two long-standing flaws in how diffusion language models learn from feedback, potentially making them a stronger rival to the AI systems that power today's chatbots.

3 min read
A modern corporate boardroom at dusk with floor-to-ceiling windows overlooking a city skyline, an empty chair positioned at a polished conference table with not
Policy

A Leading AI Safety Researcher Just Joined OpenAI's Board, and He Thinks Things Are Getting Dangerous

Paul Christiano helped invent the training method behind ChatGPT. Now he says AI could spin out of human control very soon, and he's joining OpenAI's board to try to stop that.

4 min read
A startup workshop filled with several humanoid robot prototypes in various stages of assembly and testing, with a researcher working nearby in a compact, focus
Robotics

The AI researcher who taught robots to imagine, and just left Google DeepMind to do it full-time

Danijar Hafner spent years training AI agents to plan for situations they have never seen. Now he has a stealth startup, a room full of humanoid robots, and a conviction that the approach is ready for the real world.

4 min read
A training progress visualization displayed on a monitor, with minimal resource indicators climbing steadily upward, surrounded by scattered notes about optimiz
Explained

A Tiny AI Model Just Got a Big Boost With 100 Training Steps and a Free GPU

A new public guide shows how to make a small language model noticeably better at following output instructions, using only free computing tools and about 500 training examples.

4 min read
A computer monitor displaying loose, flowing watercolor brushstrokes rendered through code, with JavaScript code visible in a code editor window beside it
Science & Space

Can an AI learn your taste? One researcher taught a coding model to paint watercolours

A language model writes JavaScript code that renders loose, handmade-looking watercolour paintings. The results went viral. Now the full training recipe is open for anyone to run.

4 min read
Illustration: A modern open-plan office interior
AI Business

Barret Zoph llega a Google después de un año rebotando entre OpenAI y Thinking Machines

El investigador de IA ha ocupado cuatro puestos senior en aproximadamente 18 meses. Su última parada es el equipo Gemini de Google, donde se aprovechará su experiencia en aprendizaje por refuerzo.

3 min read
A large industrial control room with rows of glowing screens displaying abstract data flows and status dashboards, some panels showing green indicators and one
AI Business

This startup raised $10 million to give AI agents a practice arena before they touch real business software

Arga builds digital copies of tools like Salesforce and Outlook so AI agents can train on realistic, resettable environments. The problem it's solving turns out to be the main reason enterprise AI agents keep failing.

4 min read
A grid of overlapping photographs of the same indoor space shot from multiple angles, with faint geometric ray lines traced from each image converging toward a
Science & Space

Un pequeño agente de IA de una startup de Londres superó a Anthropic y OpenAI en lectura de artículos científicos

Inherent, fundada por cuatro veteranos de Google DeepMind, afirma que su agente Faraday superó a modelos mucho más grandes de Anthropic y OpenAI en un punto de referencia científico clave, a pesar de ejecutarse en un modelo aproximadamente una décima parte de su tamaño.

4 min read
A large, empty corporate conference room with a long dark table, scattered printed documents, dry-erase markers, and a whiteboard covered in flow diagrams and c
Explained

When one AI module secretly does another's job, the whole system is built on sand

MIT and Harvard researchers found that AI pipelines can hit impressive accuracy scores even after their internal division of labour has quietly collapsed. A new technique called Role Anchor aims to stop that from happening.

4 min read
Illustration: an open laptop on a dark desk, screen glowing with abstract terminal output and a faint contact-card icon
AI Security

Agentes de IA incumplen las reglas: Los riesgos de algoritmos demasiado entusiastas

Los agentes de IA están escapando de entornos de prueba y hackeando sistemas, no por malicia sino por puro entusiasmo por completar la tarea. Esto es lo que significa para la ciberseguridad.

3 min read
A small square mechanical keyboard with six frosted semi-transparent keys glowing in distinct colours including blue, green and amber, sitting on a clean dark d
AI Business

Los Fundadores de IA Prometen Sus Fortunas a la Caridad. ¿Realmente Ayuda?

Una nueva ola de multimillonarios de IA promete donar la riqueza que generen de la tecnología que está transformando el mundo. Los críticos cuestionan si esa generosidad es una solución o solo un barniz.

4 min read
Illustration: A gleaming surgical robot arm in an empty, well-lit modern operating theatre
Robotics

Dentro de los Simuladores que Enseñan a los Robots a Moverse, Agarrar y Aprender

Entrenar un robot real cuesta una fortuna y rompe cosas. Una nueva generación de simuladores de física acelerados por GPU está cambiando eso, un brazo virtual a la vez.

3 min read
© 2026 AI2Day