The Times Tech Podcast The humans training AI to be better
Dec 12, 2025
Edwin Chen, Founder and CEO of Surge AI, shares his journey from data science at Google and Twitter to leading a company that recruits expert annotators for AI training. He discusses the significance of Artificial General Intelligence (AGI) and the vital role of human expertise in enhancing AI models. Chen emphasizes the difference between high-quality annotations and cheap labeling, and warns about the societal impacts and motivations driving the AGI race. His insights unveil the unseen yet crucial work that shapes AI's future.
AI Snips
Chapters
Transcript
Episode notes
Million-Person Platform With PhD Experts
- Surge AI claims over a million contributors, including a large cohort of PhD experts.
- Edwin says Harvard professors and Stanford statisticians teach models advanced reasoning and research skills.
Probe Then Teach Specific Model Gaps
- Probe models to find specific weaknesses before designing targeted teaching interventions.
- Use demonstrations and varied teaching methods rather than only feeding more text.
AGI Likely, But Timelines Are Longer
- Chen expects AGI but warns timelines are longer than hype suggests and that bridging final gaps is hardest.
- Real-world ambiguity and the jump from '80%' to near-perfect reliability are major obstacles.