Freakonomics Radio 555. New Technologies Always Scare Us. Is A.I. Any Different?
140 snips
Aug 31, 2023 
Ajay Khotra, an AI risk expert at Open Philanthropy, and Simon Johnson, an MIT economics professor, delve into the societal fears surrounding artificial intelligence. They explore how historical anxieties about technology mirror current sentiments towards AI, discussing its potential to disrupt jobs and transform the economy. The duo emphasizes the need for responsible technology management, addressing the alignment challenge of AI with human intent, and advocating for inclusive decision-making to navigate the complex landscape of technological evolution.
AI Snips
Chapters
Transcript
Episode notes
AI's Current Capabilities
- Current AI models like ChatGPT are essentially advanced prediction machines.
- They analyze vast datasets to predict desired outputs, but their overall quality is still subpar.
AI's Potential and Growth
- AI systems improve with larger datasets, bigger networks, and longer training times.
- AI models develop internal representations of the world, even without explicit training data, as shown by Othello GPT.
Biological Anchors
- Ajayi Kotra uses "biological anchors" to compare AI size to animal brains.
- GPT-2 was honeybee-sized, now GPT-4 is squirrel-sized; this helps predict AI's trajectory.