NewWhite paper: Somatic computing

AI just gota body.Yours.

Abilith is pioneering somatic computing: superhuman ability, for everyone.

Concept visualizationEarly development. Not working hardware.Hermes / Somatic computing / Concept

The problem

Knowledge scales. Skill never has.

We can print a book a billion times. We've never been able to copy a surgeon's hands even once.

Physical skill still passes from person to person through years of practice side by side. Video, coaching apps and VR can show you what to do. None of them can put the movement into your body. That step, from seeing to doing, is the one somatic computing is built for.

Somatic computing

AI that works through the body.

The name comes from the somatic nervous system, the part of us that controls voluntary movement, and from soma, the Greek word for body. A somatic computer does three things.

  1. 01

    Senses

    What you're doing and what you're looking at: your joints and muscles through the sleeve, the task in front of you through camera smart glasses.

  2. 02

    Understands

    How your movement compares with the way an expert would do it. We call this somatic intelligence.

  3. 03

    Acts

    With light, limited guidance on a movement you've started: cues that show where and when to move, and brakes that steady a joint or stop it at the right point.

The somatic loop The sleeve senses joints and muscles while glasses see the task. Somatic intelligence compares the movement with an expert's from the skill library, fitted to the wearer. Hermes guides with cues, brakes and a light grip assist. The wearer leads, and can redirect or stop it at any time. Each turn brings the next movement closer to the skill. The somatic loopYou lead every turn of it Skill libraryRecorded from experts SenseThe sleeve readsjoints and muscles;glasses see the task CompareSomatic intelligencematches you to theexpert, fitted to you GuideCues, brakes and alight grip assist,hand to shoulder You leadRedirect or stop itanytime; help stepsback as you improve Each turn brings the next movement closer to the skill

The fast part of the loop, from sensing to a cue or a brake, runs on two small chips in the sleeve. It is designed to close in an estimated 25 to 60 ms, inside the roughly 80 ms in which a guided movement still feels like your own.

The models run on the glasses and your phone, which send short targets to the sleeve. Camera smart glasses add what the arm alone can't know: the object in your hand and the task in front of you.

Hermes

Human ability, extended.

Our first somatic computer is a knit sleeve that runs from the fingertips to the shoulder and pairs with camera smart glasses. It's designed to weigh about half a kilogram. Scroll through it in four states, and scroll back to reverse.

Concept renders of a proposed design

01/ 04

Sense

Fit

Guide

Learn

Yarn sensors in the glove read each finger joint. A 16-channel EMG band reads intent and effort at the forearm, and four motion sensors follow the hand, forearm, upper arm and shoulder.Sensing locations only. No live data.

One knit sleeve, with carbon only where the brakes need something stiff to push against. The battery and radio sit at the shoulder, so only about 90 to 130 g is past the wrist.Exploded along the axis each part mounts on. Weights are design targets.

Vibration cues show where and when to move, and brakes about a millimeter thick stop the finger where the expert's would. Cues and brakes can't move you. They shape the movement you start.Illustrative fingertip path.

Practice with help that fades. Hermes starts with full guidance and gives less as your errors drop, until the movement is yours.Guidance levels are illustrative.

Concept visualizationEarly development

Hermes on a raised arm: a light knit sleeve from fingertips to shoulder, with carbon cuffs, an EMG band and finger brakes01 / 04

Sense

Yarn sensors in the glove read each finger joint. A 16-channel EMG band reads intent and effort at the forearm, and four motion sensors follow the hand, forearm, upper arm and shoulder.

Exploded view of Hermes: the hand plate, finger brakes and forearm cuff lifted off the arm02 / 04

Fit

One knit sleeve, with carbon only where the brakes need something stiff to push against. The battery and radio sit at the shoulder, so only about 90 to 130 g is past the wrist.

Index finger following an expert's fingertip path, with a vibration cue glowing at the wrist03 / 04

Guide

Vibration cues show where and when to move, and brakes about a millimeter thick stop the finger where the expert's would. Cues and brakes can't move you. They shape the movement you start.

The same finger movement later in practice, with the guidance turned down04 / 04

Learn

Practice with help that fades. Hermes starts with full guidance and gives less as your errors drop, until the movement is yours.

Three modes

Enhance

Makes what you can already do steadier. Brakes resist unwanted motion, and a light grip assist adds holding force.

Guide

Takes you through a movement you haven't learned yet. It shows you the movement without making it for you.

Learn

Practice with help that fades: full guidance at first, less as your errors drop, until you're doing it on your own.

Hermes v1, by zone

One knit sleeve carries the wiring, the electrodes and 12 vibration motors. Weight sits at the shoulder, so the hand carries only thin brakes and sensors.

Hand and fingers

  • Five electrostatic finger brakes
  • Strain-sensor glove
  • Carbon back-of-hand plate
  • Hand motion sensor

Wrist and forearm

  • Wrist brakes
  • 16-channel dry EMG band
  • Carbon cuff and motion sensor
  • Wrist chip that runs the fast loop

Upper arm and shoulder

  • 8 Wh battery, about 36 g
  • Shoulder chip and radio
  • Carbon upper-arm cuff
  • Grip-assist motor, about 10 N

Design targets, not measurements of a finished device: about half a kilogram from fingertips to shoulder, with about 90 to 130 g past the wrist, averaging under a watt.

The skill library

Skill, recorded at its source.

Hermes is the first device. The library of skills is the platform. An expert records a skill by doing what they're best at while wearing Hermes, and is credited by name and paid for it.

Rendered hand in the Hermes knit glove, fingers on a piano keyboardConcept render

A pianist's timing

Music

A passage you can hear in your head but can't yet play. Feel how a concert pianist voices it, in your own fingers.

Rendered hand in the Hermes glove and sleeve, gripping a chef's knife over a board of diced carrotConcept render

A chef's knife work

Kitchens

Fast, even and controlled, from a chef who preps for hundreds of covers a night.

Close-up of gloved fingers pinching a small turned brass partConcept render

An assembler's precision

Work

Holding a small part still through a fine fit, the kind of skill that takes a technician years to build.

We start with precision hand skills: fine, guided movement of the hand and wrist. It's what Hermes covers best, it's measurable in accuracy, speed and error rate, and it runs from assembly lines to kitchens to music rooms.

How the library grows

  1. 01

    Experts record

    The world's best capture how they move, and are paid for it.

  2. 02

    Anyone runs them

    People and companies run skills on Hermes. A company can keep its own experts' skills in a private library.

  3. 03

    Sessions make data

    How real people move, measured on the arm itself. Wearers choose whether to share it.

  4. 04

    Every skill improves

    Shared data improves somatic intelligence, which makes every skill better and draws in more experts and users.

Principles

The human stays in control.

A technology that guides your body has to earn trust. Hermes is built around four commitments.

  1. 01

    You lead

    Hermes only adds to movements you start, and one control cuts power and releases every brake.

  2. 02

    You keep the skill

    We'll judge Learn by what people can do once the sleeve comes off.

  3. 03

    You own your data

    Nothing is shared, with robot makers or anyone else, unless you opt in. We treat it as sensitive neural data.

  4. 04

    Experts are paid

    Every skill in the public library comes from an expert who is credited by name and paid to record it.

"You lead" is a property of the hardware, not a software promise. Cues and brakes can't move you, brakes release the moment power is off, and the one part that adds force, a grip assist of about 10 N, is small enough to overpower.

Thesis

Physical AI is at a fork.

One path builds machines that do things for us. The other builds AI that lets us do anything.

Robots should take the heavy, repetitive work. Skilled, varied work is different. There, the intelligence can go into the person instead.

Within the next decade, we believe more people will wear physical AI than will ever live alongside a humanoid robot.

Resources

Read the white paper.

White paperAbilith

AI just got a body.
Yours.

Somatic Computing

Why physical skill has never scaled, why it can now, and how somatic computing works.

By AshwinNov 2, 202645 sources

Read the paper

More ways to be human.

Create more, learn new skills, and do things your body couldn't do yet. The skill of the world's best, in your own two hands.

If you work on haptics, motor learning, embedded control or smart textiles, we'd like to compare notes. A contact route will be listed here once it is set up.