Navigating A World Where AI-Generated Misinformation Can Be Created At Scale
summary
Claims a practical detector and dashboard called GPTZero API that returns sentence‑level and document‑level probabilities and an overall “document generated” score. System reports 3.5 million users, ~400,000 daily API calls, ~1M monthly active users and integrations with 40 organizations. University study analyzes 100,000 submissions from 4,000 students, finding 1.5% likely AI overall, up to 20% in one course and a worst case of 14 flagged submissions from one student. Detections hit ~99% human recall at chosen thresholds but remain imperfect; plans include recaptcha‑style signals and source‑verification tools.
AI-written summary · about
chapters (12)
- 0:00Introduction to AI risks
- 2:08Automation diminishes writing value
- 5:55Data poisoning and hallucinations
- 8:24Uncanny valley of AI errors
- 10:32Scalable security attacks
- 12:15Responsible AI adoption
- 14:18Introducing GPT0
- 18:59GPT0's impact and use cases
- 21:08AI detection in education
- 25:45Future of AI detection
- 27:51Personal Wikipedia editor vision
- 28:59Q&A and closing
related talks
24mAI: Empowering Students, Not Replacing Universities
31mGenAI among us
26mBuild AI Systems for Discernment, Not Approval
12mTrusting AI with Code: How Secure is AI-written code?
32mHow to Make AI Work for You: A Deep Dive Guide into ChatGPT for IT Professionals
34mTrust in the Age of AI Agents | Building Trustworthy AI
47mThe Rise of AI: A Reality Check on Generative AI
20mThe State of Truth and Trust at Work