About GPT-3 Covid-19

The article titled “I Asked GPT-3 About Covid-19. Its Responses Shocked Me.” by Thomas Smith on Medium explores the potential of generative AI systems, specifically OpenAI’s GPT-3, in guiding future pandemic decision-making. The author conducted an experiment where he introduced GPT-3 to the concept of Covid-19 and asked it questions about the pandemic. Despite GPT-3’s training data only going up to October 2019, before the virus started circulating broadly, the system’s responses were insightful and surprising.

Here are four key features highlighted in the article

  1. Powerful AI System: GPT-3, trained on billions of web pages and tens of thousands of books, can generate nearly any kind of text, from news articles to computer code to sea shanties.
  2. Limitations in Training Data: The current version of GPT-3 was only trained on data gathered through October of 2019, meaning it has no inherent knowledge of Covid-19.
  3. Experimental Use: The author used GPT-3 to explore its potential for future AI-assisted decision making, especially in evaluating embedded biases. It’s not a production-ready use of the system for medical purposes.
  4. Inference and Response Mechanism: To interact with GPT-3, one provides a written prompt. GPT-3 then infers the user’s intent and responds by “completing” the prompt. The author started with general prompts about the pandemic, yielding diverse responses in different formats.