Professor shows minor tweaks flip AI-detector scores from 100% AI to 100% human
ipeirotis · x · 2026-10-29
In a debate with BlancheMinerva, professor ipeirotis presented direct evidence against deploying AI-text detectors in classrooms: minor tweaks to a text can flip the score from 100% AI to 100% human, showing the tool is not robust. He argues this is 'inappropriate deployment of a faulty product' and that responsible founders should be explicit about proper use cases, while refusing to do free QA for a company he says enables witch hunts for profit. BlancheMinerva counters that reliably detecting AI text is a real skill and questions why granting that power to people is bad. The core dispute: detector reliability, classroom misuse, and vendor responsibility.
More from Safety
- Oxford scholars slam frontier AI labs' 'doomsday talk while racing to build' contradiction — SandraWachter5 · 2026-10-29
- x401 Protocol Aims to Give the Agentic Web Verifiable Identity Proof at HTTP Level — csuwildcat · 2026-10-29
- DOJ Reportedly Probing Nvidia's $13B Hugging Face Acquisition Over Antitrust — Polymarket · 2026-10-29
- Voice AI Can Invent Drug Names Undetected — Rime's Prism Puts Guardrails Back In — aakashgupta · 2026-10-29
- Zvi calls out Google's RSI contradiction: execs tout it while the safety card denies doing it — TheZvi · 2026-10-29
- METR builds DC team to explain AI loss-of-control risk assessments to policymakers — RishiBommasani · 2026-10-29