When machines 'act' they may take our action away

The promise of AI agents is seductive: machines that bridge conception and execution. Yet through Hannah Arendt's lens, we see why AI 'action' fundamentally differs from human action.

When Manus AI launched some months ago, it did boldly claiming it was "the first general AI agent" capable of acting autonomously without human instruction. An AI that finally bridges "the gap between conception and execution."

Beneath this techno-optimistic claim lies a shift in how we conceptualize the relationship between humans and machines. What happens when machines don't just assist our actions but replace them entirely?

Through Hannah Arendt's political philosophy, I argue that AI "action" fundamentally differs from human action. While Manus AI may execute tasks independently, its actions lack what Arendt identifies as the three essential conditions for meaningful human action: plurality, political engagement, and responsibility.

This is the heart of the problem: not just that machines act differently, but that their 'action' threatens to displace ours. The issue lies in the false equivalence we create when we grant machines the status of "actors" in our shared world. By accepting machine operations as “action”, we unwittingly create a false equivalence that undermines our own agency.

Manus AI: The Seduction of Autonomous Agency

Manus is the Latin word for hand — and a revealing name choice. The company positions it as "a general AI agent that turns your thoughts into actions," a seemingly bening extension of your body, another hand. But Manus represents something unprecedent: an AI that doesn't just assist humans but replaces their agency entirely.

Unlike other chatbots, Manus maintains coherent goal-directed behaviour across extended periods and contexts: researching topics, drafting documents, scheduling meetings, and managing workflows without constant human prompting.

Beyond the impressive technical capabilities, it is the redefinition of action what is consequential.

Why AI Action Differs from Human Action

Hannah Arendt’s work helps us to understand that action isn't merely doing things.

Action is the fundamental activity that defines human existence.

Action, in her framework, is what creates meaning and constitutes our political reality: it is how we appear to each other as unique beings in a shared world.

Arendt's conception of human action rests on three pillars, each of which revealing why AI systems like Manus cannot truly act in the world.

1. AI action lacks plurality

Plurality is the condition of human action because we are all the same, that is human, in such a way that nobody is ever the same as anyone who ever lived, lives and ever will be.

Source: Arendt, Hannah. The Human Condition. University of Chicago Press, 1958, p. 8.

Plurality is what makes us universally unique. We exist not in isolation but in a web of relationships to others who are both like us and distinct from us.

Our actions gain meaning through this plurality: through being witnessed, recognized, and responded to by other unique individuals, like us.

When Manus "acts," it does so without appearing before others like it. It has no stake in outcomes, no reputation to maintain, and no unique viewpoint to contribute.