Pangram AI detector flags 2016 human-written papers as 100% AI-generated
GaryMarcus · x · 2026-10-21
User @QualiaQuanta reports that Pangram, an AI-content detector, scored papers they clearly wrote themselves in 2016 and earlier as 100% AI-generated. Gary Marcus amplified the complaint.
The takeaway: don't trust AI detectors to judge whether text was written by AI — false positive rates are high enough to flag years-old human work, underscoring the risk of using such tools in academic-integrity settings.
More from Models
- OpenAI partners with Google Cloud to bring frontier models and Codex to Model Garden — AIFlow_ML · 2026-10-21
- 35B MoE at 20.5 tok/s on an iPhone: OpenBMB's Lookahead Expert Scheduling skips no experts — GasSmooth7439 · 2026-10-21
- ICA Lens in action: decomposing Qwen residuals reveals clean city/France features — furongh · 2026-10-21
- $200 subscription covers $655 of API usage — 3900% ROI on Claude Code/Codex — umeshonai · 2026-10-21
- NVIDIA Kumo Forecast Pairs Pretrained Time-Series Representations with Pattern Retrieval — Hugging Face Blog · 2026-10-21
- Reddit user suspects Gemini 6.1 Ultra usage limits quietly shrunk, seeks independent benchmarks — Reiszecke · 2026-10-21