The Case for AI Pessimism in the Age of AI Optimism
summary
Claims vs evidence: a Mayo Clinic model flags pancreatic cancer on CT scans 475 days earlier than clinical diagnosis and a Swedish model finds early breast cancer 20% more often than radiologists, yet a support chatbot named Tessa gives harmful advice and a company‑approved LLM repeatedly recommends "drop table" during an SQL test. Detection of AI content averages only 51% accuracy, bias amplifies existing gaps 3–6x amplification, and energy use trends toward tenfold growth — concrete wins and concrete failures.
AI-written summary · about
chapters (15)
- 0:00AI pessimism introduction
- 1:28What AI truly is
- 4:23AI’s reasoning limits
- 8:40Marketing vs engineering in AI
- 9:23AI hype cycles history
- 14:21AI optimism risks
- 15:03Anthropomorphizing AI dangers
- 16:29AI vs communication and trust
- 17:32AI’s humor and nuance failure
- 18:15AI’s lexical bias and automation risks
- 19:39AI’s energy and environmental costs
- 20:43AI’s threat to democracy
- 22:28Critical thinking against AI risks
- 23:31Closing remarks
- 23:51Q&A: AI’s democracy risk
related talks
31mAvoid mistakes building AI products
38mKeynote: Thriving in the Age of AI
37mAI in medicine: where do we stand?
49mAn AI with an Agenda
28mThe Future of Healthcare – AI’s Promise and the Challenges Ahead
47mThe Past as a Compass for Systems Engineers in the Age of AI
56mArtificial Intelligence? - more like Artificial Stupidity!
46mAI for Society – Shaping a Positive Future