#134 - Text-to-Speech, Gartner Hype Cycle, AI2 OLMo, AlphaStar Unplugged, China Regulations, AI Porn Marketplace
Last Week in AI
00:00
Advancements in AI Evaluation and Training
This chapter explores a generative assessment project for evaluating language models, highlighting issues like hallucination and the introduction of a novel instruction fine-tuning technique. It discusses the potential of synthetic data generation for enhancing model training and the strides made in offline reinforcement learning, particularly in robotics. The chapter also covers optimization strategies in language models, emphasizing the importance of larger steps in achieving better performance outcomes.
Transcript
Play full episode
