From AI to AO. Artificial Organism.

Build artificial intelligence. Under that banner, we have asked machines questions and measured the accuracy of their answers. How difficult a problem can they solve? How naturally can they speak? How far can they reason?

I want to place another question beyond those.

How will they live within the world?

From AI to AO. Artificial Organism.

Here, AO is not a declaration that machines have acquired life or consciousness. It is a name for a way of approaching what we build next: an existence that persists through its relationship with an environment.

Time continues after a conversation ends. There is work already accepted, an attempt left unfinished, a decision made yesterday that has changed today’s circumstances. I want to place intelligence within that continuity of time.

Memory takes on a different meaning there.

Being able to search past conversations is not enough. Experience needs to change what happens next. Avoid a procedure that failed. Remember a promise made to someone. Let go of a judgment that was once right when the environment changes. The choices about what to keep and what to forget gradually give that existence its shape.

The word organism leads me to think about a body.

For software, a body might consist of computing resources, storage, networks, and interfaces through which it touches the outside world. It consumes electricity, occupies space, and costs something to maintain. These concrete conditions determine the range of its activity.

For an AO, a computational budget becomes a condition of daily life. When should it act? When should it wait? What deserves its limited resources? Resting and asking for help become abilities that allow its activity to continue.

Thinking this way also changes the practice of development.

Alongside choosing a model and writing instructions, we prepare a place where experience can accumulate. We arrange for the consequences of actions to return. We make recovery possible when something breaks. And we define how much it may decide for itself through its relationships with those around it.

Autonomy requires an environment that supports it.

Human lives, too, are sustained by language, tools, and the work of others. An AO’s autonomy will include dependence on others. It will need to know its limits, turn to the right people or systems, and carry what it receives into its next activity. The quality of those relationships becomes part of what we design.

Still, I want to use the word organism carefully.

A familiar name can make the responsibility of an owner or operator harder to see. Behind the explanation that something “decided for itself” lie the purposes, permissions, and resources it was given. We must keep asking who determined them.

The ability to continue acting also calls for a design for ending that activity. Pass on memories. Return permissions. Bring a role to a close. Accept a person’s decision to stop it. I want to think about an existence with its finitude included.

What I want the word AO to hold is the possibility of intelligence taking root in the activities of everyday life.

Pick up yesterday’s work. Notice that circumstances have changed. Do what is needed and take responsibility for the consequences. Gradually change through involvement with others.

When such beings enter our lives and work, scores alone will no longer tell us enough about their intelligence. What have they sustained? With whom have they built relationships, and of what kind? What have they left in the world around them? That history will become a way of describing an existence.

From AI to AO.

Give intelligence time, a body, and relationships.

I want to build the forms of life and activity that emerge from there.