The Second Half – Shunyu Yao – 姚顺雨
ysymyth.github.io · 2,432 words · saved by 17 readers
tldr: We’re at AI’s halftime.
The Second Half tldr: We’re at AI’s halftime. For decades, AI has largely been about developing new training methods and models. And it worked: from beating world champions at chess and Go, surpassing most humans on the SAT and bar exams, to earning IMO and IOI gold medals. Behind these milestones in the history book — DeepBlue, AlphaGo, GPT-4, and the o-series — are fundamental innovations in AI methods: search, deep RL, scaling, and reasoning. Things just get better over time. So what’s suddenly different now? In three words: RL finally works. More precisely: RL finally generalizes. After se
saved by
- Aryan Naik
- Hangyul Lyna Kim
- Lydia Nottingham
- surya
- Jirat C
- Dex Ter
- Enrique Moran
- Yudhister Joel Kumar
- Eric Huang
- Nathan Chen
- Sarhaan Gulati
- Rohan Kanti
related reading
- As Rocks May Think | Eric Jangevjang.com
- Ilya Sutskever — We're moving from the age of scaling to the age of researchdwarkesh.com
- Training Imperativesdan.io
- The Only Important Technology Is The Internet - Kevin Lukevinlu.ai
- Generative AI's Act o1: The Reasoning Era Begins | Sequoia Capitalsequoiacap.com
- What will be left for us to work on?normaltech.ai
- I. From GPT-4 to AGI: Counting the OOMs - SITUATIONAL AWARENESSsituational-awareness.ai
- Ryan Greenblatt – What happens once AI can automate AI research?dwarkesh.com
- A Short History of Artificial Intelligenceevery.to
- Lossy self-improvement - by Nathan Lambertinterconnects.ai
- The nature of LLM algorithmic progress (v2) — LessWronglesswrong.com
- Your Evals Will Break and You Won't See It Coming - Lun Wangwanglun1996.github.io