Student GPT-3 fake blog posts pass as human

Occurred: August 2020

A US college student fooled people into believing fake blog posts generated by the GPT-3 large language model were written by a human, raising concerns about the ease with which GPT-3 could be misused to produce clickbait content and misinformation and disinformation. 

Using the account of a PhD student to bypass usage restrictions imposed by Open AI, University of California, Berkeley, student Liam Porr used GPT-3 to post a fake blog under a fake name. 

The first post - titled Feeling unproductive? Maybe you should stop overthinking - was ranked first on Hacker News

Porr told MIT Technology Review that only a handful of people seem to have noticed that his posts were AI-generated, but they were quickly downvoted by Hacker News community members. According to Porr, 'it was super easy, actually, which was the scary part.'

Operator: OpenAI; Liam Porr; Substack
Developer: OpenAI

Country: USA

Sector: Multiple

Purpose: Generate text

Technology: Large language model (LLM); NLP/text analysis; Neural network; Deep learning; Machine learning
Issue: Mis/disinformation; Ethics

Transparency: Governance; Black box

Page info
Type: Incident
Published: April 2023