Professor: using AI detectors as a static prior is just confirmation bias
ipeirotis · x · 2026-11-18
In a debate, ipeirotis criticizes blanket rejection of submissions flagged by AI-text detectors like Pangram:
- To "AI generates so much crap," he cites Sturgeon's Law—90% of everything has been crap for 70 years—but Pangram doesn't measure crap; an expert can spot actual junk in two paragraphs and reject it on the merits
- Using a detector as a prior that never updates on evidence "is called confirmation bias," e.g., rejecting everything on a "100% AI" screenshot
Core point: detector scores should not replace human review of content quality.
More from Models
- Microsoft ships MAI Code 2 and Code 2 Flash coding models into GitHub Copilot — marlene_zw · 2026-11-18
- GPT-6.2 Astra and Sol leak: 20% cheaper than 6.1, roughly on par with Claude Fable 5.2 — kimmonismus · 2026-11-18
- Matt Shumer pans a just-released model: "no idea why they released this" — daniel_mac8 · 2026-11-18
- OpenAI's GPT-6.2 Astra jumps 22 points on ExploitBench, ships tighter cyber safeguards — JiaweiLiu_ · 2026-11-18
- Matt Shumer: this model release needed 'a lot more iteration' — mattshumer_ · 2026-11-18
- GPT-6.2 Astra early review: best-in-class 3D, but frontend styling needs babysitting — flavioAd · 2026-11-18