The Neuron: AI Explained The Humans Behind AI: How Invisible Technologies Trains 80% of the World's Top Models
16 snips
Nov 7, 2025 Casper Elliott, a leader at Invisible Technologies, discusses the crucial human role behind AI training. He explains how supervised fine-tuning and reinforcement learning from human feedback are essential for developing top models. Casper emphasizes the importance of high-quality data over sheer volume and warns against the pitfalls of messy enterprise data. He highlights that specialized skills, including surprising ones like gaming, are beneficial for training AI. Lastly, he predicts that future AI deployment will rely on human oversight and develop in robotics and critical industries.
AI Snips
Chapters
Books
Transcript
Episode notes
Model Improvements Are Trade-Offs
- Improvements in one metric (creativity) can worsen others (hallucination or verbosity).
- Choosing which trade-offs matter is now a core evaluation challenge for models.
Turning Games Into Scout Data
- Invisible helped the Charlotte Hornets by turning video into structured scouting data using computer vision.
- The result gave teams near-real-time scouting insights without hiring thousands of human scouts.
Build Evaluations By Doing The Work
- Before deploying agents, perform the task manually many times to define clear evaluation metrics.
- Use human performance to build the evaluative dataset that agents will be measured against.
