Practical AI Analyzing the 2021 AI Index Report
Nov 10, 2021
This discussion dives into the latest insights from the AI Index Report, showcasing the surge in investments for AI-driven drug development. The hosts explore the realm of natural language processing, highlighting breakthroughs like GPT-3 and its industry applications. Ethical considerations of AI, particularly its implications on misinformation, are critically examined, alongside innovative models like Delphi for integrating morals into AI. The conversation underscores the evolving landscape of AI expertise, especially in hiring trends in academia versus industry.
AI Snips
Chapters
Books
Transcript
Episode notes
AI Learns Ethics
- Researchers trained an AI called Delphi on ethical judgments.
- It achieved 92.1% accuracy, outperforming models like GPT-3.
Source of Ethical Training Data
- Delphi's training data includes examples from sources like "Dear Abby".
- This raises questions about the age and nature of Abby.
Extreme Examples in Ethical Judgments
- Delphi makes moral judgments on everyday situations, including extreme ones.
- An example includes judging situations involving killing bears and nuclear bombs.
