The Behavioral Design Podcast Predicting 2025 and Beyond with Jared Peterson
Jan 22, 2025
Jared Peterson, a behavioral scientist and decision science expert working on AI alignment, joins to forecast 2025. He discusses multimodal AI becoming mainstream. They debate AI agents as coworkers and the ethics of synthetic personas. The conversation covers AI’s role in accelerating science and a controversial prediction about AI predicting research replicability.
AI Snips
Chapters
Transcript
Episode notes
People Treat Agents As Tools Not Coworkers
- Many users remain distrustful of autonomous agents despite utility; Jared calls them tools, not coworkers.
- He fears delegating tasks tied to his personal 'second brain' file system to an agent he can't fully trust.
Researchers Cloned 1,052 Participants With AI
- A Stanford/Google study replicated 1,052 participants with generative AI, achieving ~85% match on surveys and games from two-hour interviews.
- Samuel highlighted ethical and experimental implications of replacing costly human subjects with AI clones.
Use A Two Stage Pipeline For Research With AI
- Use AI as a pipeline: start with source-tied tools (low hallucination) then shift to creative modes for novel insights.
- Jared recommended toggling between grounded summaries and higher-temperature generation when researching.