# Divide-and-Conquer Dynamics in AI-Driven Disempowerment

**Authors:** Peter S. Park, Max Tegmark  
**arXiv:** [2310.06009](https://arxiv.org/abs/2310.06009)  
**Date:** 2023-10-09  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2310.06009v1/

## Key points

- AI companies are developing systems to outperform humans in economically valuable tasks, leading to the automation of jobs traditionally held by artists, actors, and writers.
- Internal discord exists within the community, with some focusing on current harms and others on future threats.
- Authors use a game-theoretic model of conflict to understand the causes and consequences of this disunity.
- Historical evidence shows that stakeholders facing a common threat historically unite against it for mutual benefit, while the common threat seeks to sow discord among adversaries.
- The study emphasizes the importance of current victims conveying the imminent threat to future victims for solidarity and support across affected groups.
- A cohesive movement stands a greater chance of success when members believe in the efficacy of their efforts rather than viewing them as futile.
- Unity and success are more likely when movement members prioritize long-term well-being over immediate gains.

## Summary

In their paper titled "Divide-and-Conquer Dynamics in AI-Driven Disempowerment," authors Peter S. Park and Max Tegmark delve into the complex dynamics surrounding the advancement of AI technology and its impact on various industries. They highlight how AI companies are striving to develop systems that can outperform humans in economically valuable tasks, leading to the automation of jobs traditionally held by artists, actors, and writers. However, a significant challenge arises from the internal discord within the community. Some individuals focus on addressing current harms while others prioritize future threats. To better understand this disunity, the authors employ a game-theoretic model of conflict that sheds light on the causes and consequences of such divisions. Through their analysis, they uncover a historical pattern where stakeholders facing a common threat have historically united against it for mutual benefit. Conversely, the common threat has often sought to sow discord among its adversaries to weaken their collective resistance. Under realistic parameter assumptions, the model generates several predictions that align with historical evidence. Firstly, it emphasizes the importance of current victims of AI-driven disempowerment in conveying the imminent threat to future victims, encouraging solidarity and support across affected groups. Secondly, it suggests that a cohesive movement against AI-driven disempowerment stands a greater chance of success when its members believe in the efficacy of their efforts rather than viewing them as futile. Lastly, unity and success are more likely when movement members exhibit less myopic behavior by prioritizing long-term well-being over immediate gains. Overall, this study provides valuable insights into how collaboration and strategic alignment among stakeholders can strengthen resistance against shared threats like AI-driven disempowerment. By understanding these divide-and-conquer dynamics, individuals and organizations can work towards a more unified front in addressing challenges posed by advancing technologies.