Ruben Hassid · 10 min read · Gen AI Tools
Can you detect AI?
Read the original on ruben.substack.com →
Substack's Pangram AI detector is featured, including anecdotes about trying to bypass it and a critique that AI detection is largely unreliable. The piece argues detectors are pattern-matching systems, AI and human writing can overlap, and writing style can be manipulated with prompts, making authorship hard to prove.
Key Takeaways
- Pangram exists and the author attempted to game it for hours, showing detectors can be bypassed.
- Detectors are pattern-matching tools with limited reliability; short texts provide little signal, and even OpenAI has questioned detection feasibility.
- AI-written and human-written text can be manipulated via prompts and stylistic changes, making true authorship hard to prove.
- Practical takeaway: use prompts and style controls to influence voice (e.g., ADS-STE100), but detectors should not be the sole measure of authorship.
Excerpt
Substack is at war against AI. The CEO of Substack just announced their new AI detector: Pangram is the leading AI detector. They want to fight Claudefishing - fun word. You can also detect AI slop instantly. What’s AI slop? It’s when you receive this from your colleague: This budget isn’t a number. It’s a statement of intent. Because at the end of the day, a budget doesn’t tell you what you can afford — it tells you […]. You don’t need a degree to detect it. If you and I can detect AI so easily, clearly Substack bot can…