Pangram CEO Max Spero on why human touch is about to become the scarcest thing on the internet
Summary
Pangram CEO Max Spiro joins Ben Smith and Max Taneay to explain how their AI detection tool identifies machine-generated text by analyzing subtle stylistic patterns and model-specific biases. The discussion covers the escalating cat-and-mouse game between AI generators and humanizers, the impact of AI slop on platforms like Substack, and the evolving role of human authorship in media. Spiro argues that while AI assistance is common, detecting fully autonomous or deceptive AI content is crucial for maintaining trust, with human touch becoming the primary scarce value in a future of abundant …
Topics discussed
Introduction to Pangram and its CEO Max Spro
Sponsor segment: Frontier CMO podcast
How Pangram detects AI text using weak signals
Measuring false positive rates on pre-ChatGPT data
Common linguistic hallmarks of AI writing
Why LLMs use specific tropes and reinforcement learning
User demographics and business model of Pangram
Origin story: From essay detection to general text
Detecting fake reviews and mode collapse in AI
Distinguishing voices of different AI models
Pangram's rapid growth and literary scandals
The role of AI detection in media and journalism
Authorship, credibility, and the slippery slope of AI
Sponsor segment: Rethink Retail and Google Marketing
Partnership with Substack and AI disclosure tools
Challenges of detecting pre-2022 and mixed content
Future scenarios: Autonomous AI agents and security
The cat-and-mouse game of AI humanizers
Reactions from major AI labs and tech companies
Ethical stance on humanizers and industry response
Expanding beyond text: Audio, video, and granularity
The value of human touch in an age of abundant AI
Sponsor segment: Think with Google
Ben Smith's perspective on AI policing and utility
Closing remarks and credits
Listen ad-free on Castria