Lecture 11: Reinforcement Learning youtu.be/disWB7qwcOk - RL basics - Reward functions for NLP - Optimizing rewards (policy gradient) - Stabilizing learning (e.g., KL penalty, PPO)

hey Aravind Srinivas Perplexity pro responses are degrading in quality since R1 launch. responses are very "chatgpt" like (no insight, general). references are sometimes wrong. Please do not do quantity over quality. maybe reduce pro queries for pro, but don't reduce quality!

Muy feliz de poder participar en este evento e inspirar a nuestra nueva generación de científicos. Feliz de regresar a San Juan y contribuir ese granito de arena que pudiese hacer un cambio a los demás. ¡Espero nos puedan acompañar!

Glad to have introduced our new one million Puerto Ricans project at the Universidad de Puerto Rico and Puerto Rico television Wapa TV. More news coming out soon!

🇲🇽+🇯🇵

Very grateful to be attending the Mindfulness symposium organized at the Thich Nhat Hanh Mindfulness center of Harvard University. Nothing beats the feeling of sitting next to one of the most prominent Mindfulness figures in the world and MIT alumni Jon Kabatzin! 🙏

Good to have you Aravind Srinivas visiting Massachusetts Institute of Technology (MIT)! Hopefully you find the entrepreneurship spirit inspiring! In conversation with the one and only Bill Aulet from Martin Trust Center for MIT Entrepreneurship!

First time visiting the new Moderna HQ! Thank you for hosing me, exciting collaborations ahead!

Walton Goggins did to #WhiteLotus as Heath Ledger did for the Dark Knight. An acting you will always remember, for the emotional depth and suffering that it projects in an authentic and powerful way. Great work!

Another great concert and educational Latin jazz with our neighbors Berklee College of Music, thank you for hosting these amazing events!

Heading to Singapore 🇸🇬for ICLR 2025 and excited to present two accepted papers -- stay tuned for details! TetSphere Splatting: Representing High‑Quality Geometry with Lagrangian Volumetric Meshes Minghao Guo*, Bohan Wang*, Kaiming He, Wojciech Matusik openreview.net/forum?id=8enWn…

First time visiting the brand new Ragon Institute of Mass General Brigham, MIT, and Harvard developing the future of immunotherapy research. Thank you for hosting me Ragon Institute!

The actual problem lies in culture. Control people will never see themselves as ML people and viceversa. Would love to see what emerges if these folks talk to each other.

We've just released HealthBench — a new eval for AI systems for health. Developed with 262 physicians who have practiced in 60 countries.