A fixed-wing uncrewed aircraft flying straight and level above mountain ridges

Autonomy

The future of machines is autonomous.

Why machine autonomy is under way in every field at once, and where Machine Autonomy takes it.

01 / Under way

It has already begun.

Aircraft hold formation and fly their routes. Production lines run through the night. Robots move goods across warehouses the size of towns. Software agents answer, route and decide. Machine autonomy is not a forecast. It is under way.

02 / Every field

It is every field.

Machine autonomy is not robotics, and it is not software. It is the same shift arriving in manufacturing, logistics, aerospace, energy, medicine and the office at once: machines that carry the work through on their own.

A robot arm placing a circuit board in a tray, a driverless pallet truck, a small delivery robot on a sidewalk, aircraft in formation, a drone lowering a parcel on a tether and a humanoid robot carrying a tote

03 / Pace

AI changed the pace.

Language models gave machines the ability to reason about a task. What took a decade of hand-written rules now takes a model and a prompt. Every year machines take on more of the work, and the rate is rising.

04 / People

Why it matters for people.

Fewer people in harm's way: in the air, underground, at height, at sea. More done by every skilled hand, in a world short of skilled hands. Work carried to places people cannot reach and hours people cannot keep.

A quadruped robot climbing stairs at an industrial plant

05 / The industry

How the industry is building it.

Four directions, each one real progress.

  1. 01

    Larger models.

    More reasoning in the machine before it starts.

  2. 02

    More data and simulation.

    Wider coverage of what it might meet.

  3. 03

    Better hardware and sensors.

    Stronger, sharper bodies.

  4. 04

    Deeper autonomy stacks.

    Perception, mapping and planning that hold up in the field.

06 / Our direction

Our direction.

All four build what a machine knows before it starts work. We build what it learns once it is working. Autonomy is finished at runtime.

A robot arm halts over a part it has not seen, an operator's gloved hand guides it through the motion once, and the arm continues alone

07 / Why this direction

Why this direction.

  1. 01

    The world changes after deployment.

    No training covers every case a machine will meet, so part of the learning has to happen on site.

  2. 02

    Retraining for every change does not scale.

    An operator showing it once does.

  3. 03

    Control belongs at the moment of action.

    Rules, the record and the reverse switch have to be there when the machine decides.

08 / A layer

A layer, not a replacement.

The runtime works on top of the model, the sensors and the hardware a machine already has. Every one of them becomes more autonomous with it.

An autonomous haul truck on a dusty mine road

09 / Where we are going

Where we are going.

Hardware and software, combined to give every machine autonomy at runtime. That is the system we are building.

Capability comes from training. Judgment comes from experience, and experience only happens at runtime.