Training as we know it will end | Vintage Data
vintagedata.org · 1,860 words · saved by 1 readers
Old data, new models
Training as we know it will end | Vintage Data Training as we know it might end Pierre-Carl Langlais, May 8, 2025 This post was originally supposed to be about synthetic data. It is finally about a vibe shift. Over the last months, there is an accumulated amount of evidence that models are changing and synthetic data is roughly at the center of it. Too many developments have suddenly challenged firmly held assumptions: the rise of tiny reasoners, the incredible data efficiency of reinforcement learning, the rapid expansion of agentified training run into entire simulated environment. All this
related reading
- Synthetic Pretraining | Vintage Datavintagedata.org
- There Are No New Ideas in AI… Only New Datasetsblog.jxmo.io
- The Only Important Technology Is The Internet - Kevin Lukevinlu.ai
- Training Imperativesdan.io
- The Extreme Inefficiency of RL for Frontier Models - Toby Ordtobyord.com
- Sweatshop data is overmechanize.work
- The Ai Decoupling | Vintage Datavintagedata.org
- Self-Fulfilling Misalignment Data Might Be Poisoning Our AI Modelsturntrout.com
- You can’t imitation-learn how to continual-learn — LessWronglesswrong.com
- A History of the Future, 2025-2040 — LessWronglesswrong.com
- How AI Is Learning to Think in Secretnickandresen.substack.com
- Ilya Sutskever — We're moving from the age of scaling to the age of researchdwarkesh.com