What are methods human actors use to imitate robots?

Question

Robot technology is usually thought from an engineering perspective. A human programmer writes a software this executed in a robot who is doing a task.

But what would happen, if the project is started with the opposite goal? The idea is, that the human becomes the robot by himself. That means, the human is using makeup to make his face more mechanically, buys special futuristic clothing which mirrors the light and imitates in a roleplay the working of a kitchen robot.

What are methods human actors use to imitate robots?

Answer 1

The great acting teacher Stella Adler wrote about mannerisms being a powerful tool for actors. Method acting in general focuses on natural performances based roughly on understanding the mindset of the character portrayed.

It's possible actors who have portrayed androids have observed industrial robots to inform their physicality, and many performances convey the idea, via movement, of a mechanical inner structure. (It is often said that an "actor's body is their instrument".)

What is more interesting is actors trying to convey the cognitive structure of the androids.

With Arnold, and Terminator robots in general, the baseline performance is decidedly robotic, to convey their inhumanity. But the more advanced Terminators are able to mimic naturalistic human mannerisms, and even established human characters, to trick humans.

Lieutenant Data often used head motions, such as cocking his head slightly, to convey computation. Here the character arc involved working to become more human, as this character draws heavily on Pinocchio, the wooden puppet that became a boy. Overall Data's performance conveyed a lack of emotion, a definite reference to the logic-oriented Mr. Spock, although I recall episodes where Data experimented with "emotional circuits" and "humor circuits", where the output was intentionally inconsistent with natural human behavior.

Blade Runner, where the Tyrell Corporation's motto was "More Human than Human", presented the cutting edge Nexus-6 androids as having emotions, but, due to their artificially short life-spans, were portrayed as childlike in trying to reconcile extremely powerful feelings. The Voight-Kampff Test, a form of Turing Test, used in the film to identify androids, relied on the emotional response to questions.

The key plot point of Do Androids Dream of Electric Sheep, the novel the film was based on, utilized what would be formalized as evolutionary game theory to hypothesize that empathy is a natural function of intelligence sufficiently advanced. Deckard, who may or may not have been an android, and Rachel, who definitely was, are both capable of love. This capacity informed their performances, to the extent that the androids came off as more human than the actual humans, due to the depth of their emotion. This is also reflected in Blade Runner 2049 via the girlfriend-bot Joi, who us the most limited android, but the most human character in the film per her capacity to love (or at least simulate it.)

In the recent HBO Westworld reboot, the Androids replicate natural human mannerisms when playing their designated roles, but reset to more mechanical mannerisms when acting under their own agency. This is reflected in Ex Machina, where the android mimics human emotions to pass a Turing Test and trick the human subject, only to revert to purely alien mannerisms after the android is free. ("Alien" here used in the sense of non-human--it's possible the android is sentient as it seems to convey some degree of emotion in regarding the simulated human skin it will wear.)

The most interesting recent android performance may come from the recent Alien: Covenant where Michael Fassbender plays two identical androids, David and Walter, which have two distinct neural structures. (David has the capacity to be creative, where Walter cannot. In the film it is mentioned that David made people uncomfortable, so the creative functions were removed from subsequent models.) The key difference in the performance seems to be that David demonstrates passion, and even emotions, where Walter is more clearly "robotic".

• In general, the underlying approach of actors seems to have been to show the androids being distinct from humans, drawing a clear, though sometimes subtle, contrast.

• Actors portraying androids have typically utilized robotic mannerisms to convey an artificial entity.

Answer 2

Disclaimer: The intent of this answer is to suggest a a parallel between methods of acting and machine learning, both in intent and application, and theory. A large number of links are included for the convenience of readers new to the field, and there is not an exact correspondence of AI concepts to acting preparation techniques.

In my prior answer, I mentioned the method acting technique, and Stella Adler's interpretation of Stanislavski's method. Bear in mind that the method is a post-empiricism approach, an attempt, in some sense, to create a science of acting in the sense of analysis, and an approach that is fundamentally algorithmic in the sense of process. (The original manual is titled An Actor Prepares.) Note that areas covered include action, imagination (creativity), units and objectives, emotional memory (accessing memory), and adaptation. See Also: Classical Acting.)

Note also that plays are aptly named. Drama and comedy arise out of interplay of individuals, and the process of refining performance is the process of play—searching within a rule-space for the most optimal outcome.

• Strong actors will rigorously research the character to create a mental model of the character's experience of the world, similar to a model-based agent.

• Modern actors seek objectives, sometimes referred to as motivations, similar to goal-based agents.

The model has many dimensions, and there may be multiple layers of objectives in the sense of the subconscious. (What does the character want? What does it really want? What does it really really want?) This also applies to the contexts for any choice, which are multiple (personal, societal, economic, etc.)

• Actors observe human behavior for the purpose of imitating it, commonly referred to as "people watching".

As you note, actors preparing for role of robot may observe machinery, with the purpose of indicating for an audience that quality. Actors may also observe other actors, although novelty in performance is typically understood to be optimal.

• Actors will access emotional memory, alternately referred to as "sense memory" & "emotional recall" (affective memory), either to produce a physical effect or or analysis. Output are signifiers. They create a state space which they can return to access on command.

Essentially it's a form of "memory palace" (method of loci) where events take the place of locations.

• Actors will improvise in preparation, to identify and test choices (actions and mannerisms), which involves decision theory. The choices of the other actors (rational agents) are factor, and influence each other.

• Choices are selected in a genetic process, for fitness in environment, here defined as audience response. The improvisation that leads to the performance is evolutionary, in that it optimizes via the rehearsal process, with director as audience, and later, in the case of live theater, in response to live audiences. (See also the Actor-critic model.)

It's not quite a monte-carlo, being more of an informed search, but does not exclude randomness.

Essentially, it's a process of analysis, trial-and-error, more analysis, repeat, similar to machine learning with heuristics. It wouldn't be far off to say that there is a convergence, leading to what is perceived to be the optimal set of choices, (although it is more typical to say a performance "gels" or "comes together".)

• It can be said that modern acting methods are themselves algorithmic processes, where the intent is maximizing utility, here audience response, which can carry significant economic consequences.

Modern actors are using methods similar to modern AI methods to imitate intelligent androids!

In the sense of Adler specifically, the technique involves simulating natural emotions to "trick" the observer, a form of Affective computing. In other words, via training, the actor is doing what AI's are being trained to do in the context of interacting with humans.

The underlying method can be understood as a form of applied psychology and neuroscience, where the actor is accessing emotion for the purpose of analysis, and accessing specific parts of the brain on command to create observable signs.