AI Agents and the Future of Human-AI Collaborative Intelligence
summary
Generative agents built from language models simulate people and teams, achieving about 85% replication of survey responses when trained on two‑hour interviews for 1,000 subjects, but multi‑agent setups reproduce implicit stereotypes in role assignment (women get scheduling, men get troubleshooting). Resource games show communication and explicit norms raise cooperative outcomes, with moral rules boosting resource survival increases by four months in some runs. Multiple agents also speed ideation and outline generation for grant work, yielding higher novelty when paired with humans rather than used alone.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:07NLP progress retrospective
- 4:37Single models to multi-agent systems
- 8:10Bias in role assignment
- 12:46Collaboration in resource management
- 18:11AI agents in grant writing
- 22:06Human-AI collaboration opportunities
- 22:50Agent AI vs generative AI
- 23:55Resources and future cases
- 25:22Discrimination in agentic AI
related talks
28mAgentic AI
16mBuilding Applications with AI Agents
19mBeyond agents: AI as a creative partner
01h 01mKeynote Panel: Future of AI agents
40mHuman-AI Collaboration Models for Better Decisions and Faster Workflows
47mHuman Robot Agent: Scouting AI Beyond Hype and Doom
29mPairing with the machine: How AI Agents are rewriting the rules of software development
35mAI Agent and the Promotion You Didn't Ask For