ChatGPT's new voice mode is insane for learning a new language. Literally a private teacher who can correct your pronunciation and help you progress step by step. By far one of the best use cases. (Prompt I used below)

Como anécdota/juego es divertidísima, pero desde un punto de vista de safety es un gran ejemplo de usos peligrosos de la IA. En este case es un juego/prueba, y lo que no vemos es que se le pide a la IA que haga un rol playing. El riesgo es un uso similar con intereses ilícitos.

Estoy esperando que llegue a España hace meses. El modo de voz es un game changer en aprendizaje de idiomas, como puede apreciarse. Predicción: Aparecerán productos que guíen el camino de aprendizaje para obtener una certificación, y cada paso se practicará con el voice mode 🗣️

We’re thinking of all those impacted by the devastating flash floods in the region of Valencia, Spain. Apple will be making a donation to help with relief efforts on the ground.

Lo de los microplásticos es que ya es escandaloso uab.cat/web/sala-de-pr…

AI and soft power: DeepSeek R1 refuses to speak about the 1989 Tiananmen Square protests and massacre (I've placed OpenAI o1 next for comparison).

Our work on copyright and LLMs is finally out 😎 I know it's not DeepSeek or Qwen, but I still think is very interesting stuff!

In fact, analyzing the CoT of models like R1 is fascinating and an invaluable educational resource: we now know that intelligent reasoning can be learned (humans). It's just like when a tutor guides you step by step, instilling an appropriate CoT that you can learn to replicate.

Prepara directo Carlos Santana ! Yo voy preparando la actuación de nuestra plataforma… 💪🏼

Meet Helix, our in-house AI that reasons like a human Robotics won't get to the home without a step change in capabilities Our robots can now handle virtually any household item:

Con esto se empieza a desbloquear la magia... ✨🧙🏼‍♀️ La consistencia visual entre escenas es fabulosa. ¡Enhorabuena Oriol Vinyals y equipo!

DNA replication process

Now that o3 can so thoroughly answer any question I might have, I feel like the limiting factor is my ability to think of a question to ask.

The race for LLM "cognitive core" - a few billion param model that maximally sacrifices encyclopedic knowledge for capability. It lives always-on and by default on every computer as the kernel of LLM personal computing. Its features are slowly crystalizing: - Natively multimodal

Ya no se "programa" como antes. En mis inicios con Java recuerdo tener memorizado **todas** las funciones de los paquetes principales, porque no había autocompletado. El autocompletado paso a autocompletado inteligente de la IA. Y ahora el cambio es aún mayor.

What if you could not only watch a generated video, but explore it too? 🌐 Genie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt. From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵

Introducing Among AIs, a social reasoning benchmark where embodied models play Among Us to test social intelligence: deception, persuasion, and coordination. We put 6 SOTA models in a live arena and GPT-5 came out on top by leading in Impostor & Crewmate wins. Why did GPT-5 get