A.I. design: concept art and problems of representation

A.I.’s design had to make a world visible and solve a central problem: how to represent an artificial child without losing the ability to recognise its desire. The drawings and tests were not yet the filmed solution.

Concept art makes it possible to discuss a world before building it

Kubrick’s archive preserves a visual pre-production section for A.I., with materials associated with artists such as Chris Baker, Christopher Moore and Chris Foss. The presence of several collaborators shows that developing the look was neither a single task nor necessarily a linear one. A design could explore a figure, a space or a possibility of scale without the whole package being finalised at the same time.

Concept art has an advantage and a limitation: it makes it possible to represent something before deciding how to realise it. It can combine forms that would later require sets, models, animation or compositing. Its value lies in making a visual proposal open to discussion. It does not prove that a process already existed that could produce exactly that image under the conditions of a particular shoot.

The writing of A.I. and the design could raise questions for each other. A narrative situation needs a place; an image can suggest a relationship or a difficulty that changes the situation. It would be too simple to imagine that everything was written first and someone then illustrated definitive instructions. The development documentation reflects a process in which solutions could be revised.

The UAL catalogue identifies Chris Baker’s material as SK/18/3/2/3/2. This information helps establish his involvement without automatically attributing every image circulating about A.I. to him. A reproduction needs an artist, date and provenance where these can be established. The project’s general name is not enough to determine whether a drawing belongs to Kubrick’s preparation, Spielberg’s production or a later publication.

The artificial child posed both a dramatic and a production problem

David’s appearance had to sustain a narrative paradox. The character is artificial within the fiction, but needs to relate to other people in a way the viewer can follow. Showing mechanisms can explain how he was made; it does not, by itself, create an experience of desire or dependence. The representation required a relationship between appearance, behaviour and the response of the human characters.

In 2001, Jan Harlan explained that during development they tried to create the child using mechanical means, but the result did not satisfy the team. He linked this search to the difficulty of maintaining a childlike appearance throughout a lengthy production. His testimony establishes a problem and an attempt; it does not prove that there was an autonomous robot capable of performing the role, or finished footage that was later replaced.

Spielberg’s film realised David through Haley Joel Osment’s performance, integrated with make-up and effects work. This result should be studied as a decision made during the 2001 production. It is not correct to attribute all its details to Kubrick just because the dramatic problem came earlier. A project can pass on a need without passing on a definitive technical solution.

The question connects with machines and dehumanisation as a fictional theme, but is not limited to it. The design has to decide what information becomes visible and when. A character can be recognised as artificial from the context before any bodily alteration is shown. This arrangement of cues is narrative as well as technical, and depends on the order in which they are presented in the film.

Continuity between the materials does not erase the team that made the film

According to Harlan’s accounts of the transfer of materials, Chris Baker remained involved in the subsequent production. This continuity is a concrete connection between stages. It does not mean the drawings went straight to the screen without interpretation by other departments. A drawn space has to become something that can be photographed, animated or combined with the performers’ bodies and actions.

The film’s credits identify Rick Carter for production design and Dennis Muren and Scott Farrar for visual effects, alongside Stan Winston’s work on practical effects and character make-up. Identifying the roles helps us recognise an actual production, distinct from the archive of proposals. There is no need to choose between an earlier origin and a later team: both belong to the story, but they did different work.

The analysis of spaces and sets can examine how a completed environment organises movement, gazes and positions. A preparatory drawing does not contain all those relationships in motion. For example, the scale of a figure in relation to a building may be planned graphically, while the duration of its movement and the way it is revealed depend on the scene as filmed. The difference affects how each piece of material should be assessed.

The comparison with the effects in 2001: A Space Odyssey does not authorise an automatic assumption of technical continuity either. The projects belong to different production contexts and use solutions that need to be documented separately. The need to make a world believable can be compared, but specific processes used in A.I. cannot be inferred from what Kubrick used decades earlier.

Design publications offer insight into a selection of the earlier work, not a transparent inventory of everything that was done. An image selected for a book may have visual impact independent of its use in production. To understand A.I., it is worth keeping both readings in mind: appreciating the graphic proposal and asking what documented relationship it had with the film. This lets us recognise continuity and transformation without turning every drawing into an effect already completed by Kubrick.

References: University of the Arts London, Stanley Kubrick Archive, A.I. pre-production, SK/18/3/2 and Chris Baker materials, SK/18/3/2/3/2. Jan Harlan, interview with Quint, Ain’t It Cool News, 29 October 2001; interview with Geoffrey Kleinman, DVD Talk. Amblin, A.I. Artificial Intelligence (2001) technical credits. Jan Harlan and Jane Struthers, A.I. Artificial Intelligence: From Stanley Kubrick to Steven Spielberg, Thames & Hudson.

View all articles in A.I.: Kubrick’s development and Spielberg’s film