Why AI Researchers Are Studying Cells and Cities to Understand AI
This could change how we think about AI safety and our future.
A software developer has shared early ideas on 'agent foundations'—a field that tries to understand what an agent is and how it adapts. He suggests that by looking at how cells, animals, and human societies have evolved, we might predict how AI agents will behave. For example, groups of agents tend to merge into larger 'super-agents' over time, like single cells combining into multi-celled organisms. This could help us anticipate whether AI systems will become more cooperative or more dangerous.
The author points out that biological super-agents (like our bodies) are tightly controlled, while human societies are more loosely connected. He wonders if AI development will follow a similar path—will AI systems become tightly integrated like our cells, or remain independent like humans? He also draws parallels between capitalism and neural networks, suggesting that money moving through cities is like data flowing through AI. These analogies might help us build a theory that predicts AI behavior.
However, this is very preliminary work. The author admits these are 'rough vibe-y ideas' written under time pressure. There's no concrete theory yet, just observations and questions. The hope is that by studying patterns in biology and society, we can develop a framework to ensure AI systems are safe and beneficial. But for now, it's more of a research direction than a practical tool.
For the average person, this might seem abstract, but it matters because AI is increasingly part of our lives. Understanding how AI agents might evolve could affect everything from job automation to privacy. If researchers can predict how AI systems will self-improve or cooperate, we can better prepare for the future. But don't expect immediate changes—this is a long-term intellectual endeavor.
- Researchers are looking at cells, animals, and societies to understand how AI agents might evolve.
- Groups of agents often merge into super-agents, which could mean AI systems become more integrated over time.
- This is early-stage thinking, not a product, but it could help predict AI behavior and prevent risks.
Why It Matters
This research could help us anticipate how AI will change our jobs, privacy, and safety in the future.