Inner Cosmos with David Eagleman Ep42 "Is there any such thing as true news?" (Truth Part 3: Artificial Intelligence)
6 snips
Jan 15, 2024 This podcast explores the role of AI in the future of fake news and misinformation, discussing its connection to our brain's internal models and voice passwords. It delves into the impact of generative AI on disinformation and truth and highlights the potential dangers of AI voice generation. The podcast also discusses AI manipulation and social influence, as well as the challenges and potential issues of building AI oracles.
AI Snips
Chapters
Transcript
Episode notes
LLMs As Fair Meta-Humans
- Large language models act like a metahuman, summarizing millions of written perspectives instead of giving binary answers.
- This fairness yields balanced syntheses that reflect diverse human opinions rather than forcing a single verdict.
Balkanization Of AI Models
- Eagleman predicts a rapid Balkanization of AI as groups train models on ideologically curated data.
- That fragmentation will produce many partisan AIs tailored to confirm their creators' beliefs rather than a single shared oracle.
Party Example On AI Fake News
- Eagleman recounts a party conversation about AI generating fake news and shows it mainly saves time over manual fabrication.
- He argues AI doesn't fundamentally change the need for corroboration from trusted sources to spread a story.