The Machine in the Studio

Artists and architects are turning robots from instruments of repetition into partners in material research. Automation is not abolishing skill. It is moving skill into calibration, code, choreography, maintenance and the disciplined reading of matter.

A robot can repeat a movement with a consistency that exceeds the human hand. That fact made industrial automation powerful, but it does not explain why artists have become interested in robots. Repetition alone is rarely the point. The more consequential work begins when a machine encounters resistance: ink that repels another ink, concrete that must support its own wet weight, a sensor that translates a body imperfectly, or a toolpath whose digital elegance collides with gravity.

In the studio, failure is not always a defect to be removed. It can be information.

The artist Sougwen Chung has spent more than a decade developing robotic drawing systems that do not merely reproduce a pre-programmed line. Their 2026 work RECURSIONS places the artist beside two robotic arms linked to live electroencephalography data. A neural network draws from an archive of Chung's gestures while the artist responds in real time. Water-based calligraphic ink enters the linen; a silver composite remains on the surface and resists it. The resulting mark is neither purely manual nor autonomous. It is a record of several systems meeting under conditions none of them controls completely.

At a different scale, the Tor Alva in the Swiss village of Mulegns uses robotically printed, load-bearing concrete components to produce a 30-meter cultural tower. Designed by Michael Hansmeyer and Benjamin Dillenburger with ETH Zurich and the Nova Fundaziun Origen, the structure depends on custom software, new material formulations, robotic reinforcement and a plan for future disassembly. Its ornamental geometry would have been exceptionally difficult to form through conventional casting.

These projects belong to different fields, but they reveal the same change. Computation is becoming material practice. Code does not replace craft. It enters craft and redistributes where expertise resides.

The central answer

Robotics and 3D printing become artistically important when they do more than automate a finished idea. They allow artists to design behaviors, toolpaths and material conditions that could not be realized through direct handwork alone. The resulting practice still depends on embodied knowledge: how a substance flows, when a joint fails, how a sensor drifts, how a robot must be calibrated and which imperfection should remain. Computational craft is therefore not craft without hands. It is craft extended across people, machines, software and matter.

From tool to system

A traditional tool amplifies or concentrates a movement. A chisel transfers force. A brush carries pigment. A loom organizes repeated relations between threads. An industrial robot is different because it can execute a sequence of actions across several axes, integrate sensor data and change behavior through software.

This makes the robot less like a single tool and more like a configurable system of tools.

The system includes:

• the robotic arm and its physical limits;
• the end effector, such as a brush, nozzle, gripper or welding head;
• the code that defines motion;
• sensors that observe position, force, sound, vision or biological signals;
• the material being moved or deposited;
• the safety envelope around the machine;
• the person who calibrates, interrupts and interprets the process.

Authorship is distributed across this configuration. Yet distribution does not mean that responsibility disappears. Someone decides why the system exists, what it should be allowed to do, what data it receives and what counts as a meaningful result.

The most serious robotic art makes that distribution visible. It does not pretend that a machine has become an isolated creative personality. Nor does it reduce the robot to a theatrical assistant that confirms human mastery. It asks what kind of relation can emerge when human intention meets computational behavior and material uncertainty.

Drawing with memory rather than imitation

Chung's Drawing Operations Unit began in 2015 with a robot that observed the artist's gestures through a camera and drew alongside them. Later generations introduced a neural network trained on two decades of archived drawings, multi-robotic systems, biofeedback and spatial computation. The sequence is important because it shows that the artistic object is not only the finished drawing. The evolving relation between artist and machine is itself the long-term work.

In RECURSIONS, the robot does not simply reproduce the artist's current movement. It works with a model of past gestures, live brainwave data and the physical behavior of two incompatible inks. Memory becomes a technical and genealogical material.

This complicates familiar claims about automation. The machine is not replacing a repetitive task that had a stable definition. It is participating in the definition of the task. The artist builds a system whose responses alter the next human decision.

That feedback loop resembles musical improvisation more than factory execution. It demands preparation, but the preparation is not a script that predicts every mark. It establishes a range within which surprise can remain legible.

The quality of such work depends on constraint. Without constraint, machine variation can become spectacle: endless movement, endless output, endless evidence that a robot can draw. With constraint, the machine's difference becomes perceptible. Its acceleration, hesitation, rigidity or sensor-driven drift can acquire expressive consequence.

The return of calibration

Industrial narratives often present digital production as a frictionless translation from file to object. Artists and fabricators know otherwise.

A robot must be calibrated to its environment. A tool has weight and wear. A camera introduces distortion. A surface may not be perfectly level. A nozzle can clog. A wet material changes with temperature, humidity and time. A digital path that is valid in simulation can produce an unstable or visually inert object.

Calibration is therefore not preliminary maintenance. It is a form of knowledge.

The person who calibrates learns the difference between the abstract coordinate system and the machine as it actually exists. That knowledge is often tacit. It accumulates through repeated observation of errors too small or variable to be captured fully in a manual.

Computational craft places this tacit knowledge beside mathematical precision. The practitioner must understand both. Code can specify a curve to a fraction of a millimeter, but material behavior determines whether the curve survives.

This is one reason robotic fabrication can create new forms of specialization rather than eliminate skilled work. It requires people who can move among design, programming, engineering, toolmaking and material testing. The craftsperson may no longer shape every surface directly, but they shape the conditions under which the machine can produce a viable surface.

Tor Alva and the ornament of structure

Tor Alva, or the White Tower, opened in Mulegns in May 2025. Its 30-meter height and load-bearing 3D-printed concrete made it a landmark in digital fabrication. Yet its importance is not simply that it is tall.

The tower brings computation, structural engineering and cultural narrative into the same form. Its branching white columns recall ornate confectionery and the history of Graubunden artisans who traveled through Europe as pastry makers, builders and stucco workers. A technological monument thus refers to a regional history of exported skill.

The columns were printed layer by layer without conventional formwork. A specially developed concrete had to remain fluid enough to be extruded while becoming stable quickly enough to support the next layer. Robotic reinforcement was inserted during fabrication so that the printed elements could carry structural loads. The geometry and the structural logic were developed together rather than treating ornament as an applied surface.

This is computational craft at an architectural scale. The visible form is inseparable from:

• software that defines geometry;
• material science that controls setting behavior;
• structural tests for layered concrete;
• robotic placement of reinforcement;
• assembly planning;
• tolerances among printed parts;
• future disassembly.

The tower was designed to stand in Mulegns for approximately five years and then be dismantled and potentially rebuilt elsewhere. That decision introduces circularity at the planning stage. It also raises a cultural question. If a structure can move, which part of its identity belongs to the original village, landscape and social purpose?

Digital fabrication makes relocation technically conceivable. It does not make context irrelevant.

Formwork, waste and the environmental claim

Additive manufacturing is often described as materially efficient because it deposits matter only where needed and avoids conventional formwork. Tor Alva demonstrates real advantages: non-standard load-bearing forms, reduced dependence on molds and a structure designed for disassembly.

But an environmental claim must examine the entire system.

Concrete remains emissions-intensive. Robotic equipment, testing, transport, failed prints and specialized additives also carry costs. A digitally optimized form can use less material while requiring an energy-intensive fabrication infrastructure. A unique component may be difficult to repair when the original software, robot or expertise is no longer available.

The correct comparison is not "robotic good, conventional bad." It asks:

• How much material was avoided?
• What was the carbon intensity of the mixture?
• How many prototypes or failures were required?
• Can the components be separated and reused?
• Can the design be repaired with locally available knowledge?
• Does the cultural value justify the specialized production?
• What happens after the planned service life?

For artists, the same discipline applies. A 3D-printed sculpture is not sustainable merely because it uses a bio-based filament or a mathematically efficient lattice. The feedstock, machine time, supports, failed iterations, finishing, transport and eventual disposal belong to the work's material account.

The Architecture Biennale and collective intelligence

The 2025 Venice Architecture Biennale was organized under the title Intelligens. Natural. Artificial. Collective. Its framework placed artificial, biological and collective intelligence into direct relation with architecture's response to a hostile climate.

The title matters because it rejects the idea that intelligence belongs to a single technical system. A building can encode climate knowledge accumulated through vernacular practice. A material can respond through its physical properties. A community can organize collective expertise. A computational model can search a design space. None of these forms of intelligence is sufficient alone.

For computational art and architecture, this suggests a more mature ambition than technological display. The robot should not be present merely because a robot is impressive. The system should clarify which intelligence it contributes and which intelligence it depends on.

A robotic arm can execute complex geometry. It cannot determine whether the geometry serves a place, a community or a climate. A generative design can identify structural efficiencies. It cannot decide whether efficiency is the most important cultural value. Human and collective judgment remains necessary precisely because technical optimization narrows a question to the variables it can measure.

The hidden labor of the robotic studio

A robotic artwork often appears to perform itself. The audience sees an arm moving with exactness and may infer autonomy. Behind the movement lies a dense labor chain.

It can include:

• artists and creative directors;
• roboticists and software developers;
• mechanical engineers;
• fabricators and welders;
• material scientists;
• safety supervisors;
• technicians who transport and install the system;
• conservators who determine how it can be preserved;
• assistants who clean, reset and document the machine.

Crediting this labor is not merely ethical housekeeping. It helps audiences understand what the medium is.

When a robotic work is acquired by a museum, the institution may not be acquiring a stable machine. It may be acquiring a network of dependencies and permissions. The original robot model can become obsolete. Replacement parts may disappear. Software libraries may no longer run. A networked service may close. The artist's studio may need to define which substitutions preserve the work and which create a new work.

The public choreography rests on invisible maintenance.

Conservation after the robot stops

Traditional conservation often seeks to stabilize material change. Robotic and software-based art can require the conservation of behavior.

A museum may need to preserve:

• source code and compiled software;
• robot programs and coordinate systems;
• models, training data and calibration files;
• end-effector specifications;
• sensor mappings;
• safety procedures;
• approved replacement hardware;
• video documentation of movement and timing;
• acceptable variation in output;
• the artist's account of what must remain perceptible.

A literal preservation strategy can fail. Keeping the original machine untouched may prevent the work from operating. Replacing it with a newer arm may alter speed, sound and motion. Emulation may preserve software but remove the physical risk or scale that shaped the original experience.

The artist and institution therefore need a hierarchy of identity. Is the exact robot essential? Is the relationship between human and machine essential? Is a particular gesture or output essential? Can code be translated? Can the work remain authentic if its hardware changes?

These questions should be answered before crisis, not after the last replacement part disappears.

Seven stages of computational craft

A disciplined robotic practice can be understood through seven stages.

1. Cultural and conceptual framing

The artist defines why automation is relevant. Technology is not accepted as the subject merely because it is available.

2. Material study

The practitioner tests flow, resistance, curing, adhesion, wear and failure. Material behavior enters the concept.

3. System design

Robot, sensors, tools, code and human roles are configured as a relation rather than a collection of devices.

4. Calibration and rehearsal

The abstract design is confronted with actual tolerances, safety limits and environmental conditions.

5. Performance or fabrication

The system produces an event or object. Human intervention and machine response are documented.

6. Evaluation and refusal

Not every technically successful result becomes art. Selection, editing and destruction remain necessary.

7. Preservation and accountability

The artist records dependencies, collaborators, environmental costs, acceptable substitutions and the future conditions of display.

This framework restores craft to a field often described only through innovation. It recognizes that skill exists before, during and after the visible movement of the machine.

What remains handmade

The phrase "handmade" usually refers to direct physical contact. Computational craft asks for a broader understanding.

A work can be physically fabricated by a robot while remaining deeply shaped by human attention. Conversely, a hand-produced object can follow a standardized process with little meaningful decision. The relevant distinction is not simply hand versus machine. It is whether knowledge, judgment and responsibility remain perceptible in the process.

The hand survives in the system as measurement, touch, correction, maintenance and refusal. It appears in the selection of a material, the design of a tool, the response to an error and the decision that one output carries meaning while another does not.

Robotics can make form more complex. Art must make the relation more intelligible.

The future of computational craft will not be decided by the machines capable of the most astonishing movement. It will be decided by artists and makers who can show why a movement belongs to a material, a place and a human question.

Six questions readers may ask

1. Does robotic fabrication make traditional craft obsolete?

No. It changes which skills are central. Direct hand shaping may be reduced in some stages, while calibration, toolmaking, material testing, programming, assembly and repair become more important. Traditional knowledge can also guide digital processes by revealing how materials behave over time and under local conditions. The risk arises when institutions treat computational production as a complete replacement for embodied knowledge. The strongest practice combines measurable precision with tacit expertise rather than forcing one to displace the other.

2. Who is the author of an artwork made by an artist, a robot and a technical team?

Authorship depends on contribution and context. The named artist may define the concept, constraints and final form, while engineers and fabricators make essential creative or technical contributions that deserve explicit credit. The robot is part of the medium, not a legal person who assumes responsibility. Institutions should document roles rather than hiding collective production behind a single name. Credit and authorship are related but not identical: a collaborator can deserve prominent recognition even when the artwork remains attributed to the artist who established its governing artistic decisions.

3. Is 3D printing automatically more sustainable than conventional fabrication?

No. Additive manufacturing can reduce formwork, enable material-efficient geometry and support disassembly, but environmental performance depends on feedstock, energy, failed iterations, transport, durability, repair and end-of-life treatment. Concrete printing remains connected to the emissions of cement and specialized infrastructure. A credible sustainability claim requires a defined comparison and a life-cycle perspective. Artists should document what was saved, what was consumed and what can be reused rather than treating digital optimization as proof of environmental virtue.

4. Why use a physical robot when software could generate the image directly?

A physical robot introduces scale, resistance, timing, sound, risk and material feedback. It can make the negotiation between computation and matter visible. In a robotic drawing, the path is not merely rendered; it encounters paper, ink, friction and the artist's body. In architecture, a toolpath becomes subject to gravity and structural performance. The physical system matters when those encounters are part of the work's meaning. If the robot contributes only spectacle, the same concept may be communicated more clearly through another medium.

5. How can museums conserve robotic artworks when hardware becomes obsolete?

They need a behavior-oriented preservation dossier. It should include source code, software environments, robot programs, calibration files, sensor mappings, technical drawings, safety procedures, video documentation and an artist-approved list of possible substitutions. The museum and artist should identify what is essential: the exact machine, the quality of movement, the relationship to a performer, the generated output or the larger system. Conservation may involve migration to new hardware, emulation or partial reconstruction. These choices should be documented as interpretive decisions.

6. What should an artist learn before working with industrial robots?

The artist needs more than basic programming. They should understand safety, coordinate systems, tool design, material behavior, tolerances, emergency procedures and the limits of simulation. They also need a clear artistic reason for using the system. Collaboration with engineers and fabricators is often essential, but the artist should remain able to evaluate technical proposals and recognize when convenience is narrowing the concept. The most durable skill is the ability to connect code, material and cultural meaning.

1. Sougwen Chung, RECURSIONS, 2026.
2. Sougwen Chung, artist and research practice, accessed August 2026.
3. Sougwen Chung, MIMICRY, 2015.
4. Sougwen Chung, MEMORY, 2017.
5. Sougwen Chung, ASSEMBLY LINES, 2022.
6. ETH Zurich, From confectioners to robots: Tor Alva is unveiled, May 2025.
7. ETH Zurich, Towering feat in an alpine village, March 2025.
8. ETH Institute of Technology in Architecture, Tor Alva nears completion, April 2025.
9. ETH Sustainable Digital Construction, Tor Alva, accessed August 2026.
10. ETH Zurich, robots assemble reinforcing steel, June 2025.
11. La Biennale di Venezia, Biennale Architettura 2025: Intelligens. Natural. Artificial. Collective, 2025.
12. Art Basel, Zero 10 in Basel, June 2026.
13. Art Basel, digital artists were right all along, Trevor Paglen, May 2026.