On with Kara Swisher AI Ethics and Safety — a Contradiction in Terms?
169 snips
Jan 2, 2025 
Join Rumman Chowdry, the first U.S. Science Envoy for AI, Mark Dredze, a bias researcher from Johns Hopkins, and economist-turned-AI scholar Gillian Hadfield as they delve into AI's ethical minefield. They address the haunting question: can we create unbiased AI? The trio also discusses the frightening realities of AI under potential future administrations, the need for robust legal frameworks, and the risks of stifling innovation amid the AI hype. It’s a thought-provoking exploration of the urgent need for responsible AI development.
AI Snips
Chapters
Transcript
Episode notes
Hidden Bias
- LLMs exhibit hidden biases even when trained to avoid explicit bias.
- A study revealed that LLMs showed bias in relationship scenarios based solely on name changes.
AI: Tool or Weapon?
- AI tools can be used for good or bad, like any other tool.
- AI's impact depends on how it's used, whether for healthcare or profit-driven insurance claims.
AI and Children
- Children's brains aren't developed enough to handle hyper-realistic AI interactions.
- Consider 'AI right to repair' to give users more control over AI's impact on their lives.