The Newcomer Podcast AI Predicts Its Own Rise (and Fall)
12 snips
Nov 10, 2025 What if AI could predict its own future? The hosts delve into predictions made by AI about its own evolution over five years. They analyze the accuracy of past forecasts, from political ad disclosures to ambitious targets for AI-generated music topping charts. Discussions include AI's role in healthcare and the possibility of AI-driven scientific research. Interesting debates arise on everything from Apple acquisitions to real-time generative features in video games, questioning whether AI's self-expectations hold any weight.
AI Snips
Chapters
Transcript
Episode notes
AI Overestimates Near-Term Outcomes
- AI's own forecasts last year were overly optimistic and largely failed to materialize.
- Models gave high probabilities to events that did not occur, showing limits in short-term forecasting.
Hosts' Reality Check On Last Year's Bets
- The hosts reviewed last year's AI predictions and found none of them came true.
- They counted outcomes like no GPT-5 over 10T and no national healthcare AI regs being implemented.
Use Human Judgment With AI Probabilities
- Treat probabilistic AI outputs as one input, not definitive forecasts.
- Use human judgment to set over/under bets instead of taking model probabilities at face value.