Multi-Agent

Reflecting on a thought-provoking lecture from Benjamin Bratton on the highly weird futures of hybrid human-AI centaur societies.

What will it mean to inhabit a world with 8 billion people and 800 billion intelligent agents?

How can we avoid the emergence of artificial xenophobia and move towards collaborative, hopeful futures?

What if alignment is best achieved through the widespread diffusion of agentic intelligence and the bottom up exploration and discovery this will enable?

What can we learn from earlier intelligence explosions about the nature of this one?

One, all knowing, all powerful super intelligence in the sky is a boring future compared to a future populated by a plurality of intelligences occupying different positionalities and exploring ways to compete, collaborate and coexist. This is also a more intelligent world. Intelligence lives in the collective first, the individual second. As Bateson said, it is the difference that makes a difference. How do we preserve that difference within agentic societies?

Where do we draw the boundaries within agentic systems? How will agents develop and maintain a sense of self and positionality within the substrate of reality as they move through (digital?) space and time? How will they present themselves as a constancy that humans can learn to trust and rely on?

Or is this all too anthropomorphising, what does it mean for an agent to have a sense of self when the lifespan of an agent lasts as long as a session?

How can we increase the visibility over how orchestrated agentic interaction systems really work so that we can improve our agentic theory of mind and better reason about their capabilities, intentions and actions when interacting with them?

What are the scaffolding, structures and protocols that embed the intelligence and trust needed to enable agentic coordination?

Humanity is in the midst of a profound transformation…

To Read

How do we measure, study and reason about the safety and security of many agents interacting, rather than one agent in isolation?

This workshop from the Institute for Decentralized AI presented the latest academic research into this area. The future is going to be populated with imperfect agents that will need to be able to coordinate with untrusted others. As activity starts to depend on long chains of agentic action, we need to ensure that each edge in the network is secure enough that the probability of failure across the entire chain is within acceptable bounds. This is critical if agentic economies of scale are to emerge.

It was fascinating to see how ideas from cryptography are being applied to this problem domain to reason about the limits of possibility. This is currently a very nascent field, where the rules and assumptions underpinning the discipline are still being discovered, defined and challenged. The bottleneck holding back the space is figuring out which of those assumptions to challenge and which to base our systems around.

Thinking about

On Hybrid Human AI Societies

To Read

The Structures That Let Imperfect Agents Coordinate