Wooden Acting

The roles that AI agents hold in the movies are largely the same as those that humans might hold, e.g.:

• Comic relief - Star Wars (R2D2 & C3PO)
• Background - Red Planet (the spaceship)
• Antagonist - The Terminator, The Matrix
• Protagonist - Bicentennial Man
• Companion - Knight Rider (the car)
• Assistant - Silent Running (gardening robots)

There is nothing particularly special about how an AI agent might contribute to a movie. In fact, the plots and character roles of movies containing AI agents are hardly different than mainstream movies: e.g. action movies structured around a crisis and its resolution, rite of passage movies where the main character extends his/her status, abilities or outlook, extreme behavior movies that explore the interaction between abnormal behavior and society, etc.

Are these movies realistic in their depiction of AI? This depends a lot on what one believes is realistic, but at least several criteria are applicable:

1. The abilities that AI agents have are physically realistic by current standards. This allows different agent sizes, broader and faster communication and increased strength. Unrealistic abilities would include invisibility, mind reading, time travel, etc.
2. The agent's sensing and information gathering processes may be more widespread and distributed, have extended perceptual ranges and sensitivities, and gather more information, but are not omnipresent.
3. The agent's reasoning processes may be faster, more thorough, incorporate more information, but are not omniscient.

The key justification for believability is that the technologies used to construct the AI agents are plausible, but significant extensions of existing technologies. There might be some theoretical limits on the language abilities, personalities, intellectual capabilities of a plausible AI agent, but these seem largely without practical limitation. By these criteria, most AI agents presented in the cinema are close to being realistic. The mindless killing machines and advanced household and driving appliances are easily plausible AI agents, but the AI agents with `minds' might be contentious. My opinion is there is nothing obvious that excludes the AI agents that are presented - although maybe additional criteria of realism will be found as research progresses..

Thus, what we see in most AI movies are plausible agents and largely achievable ultimately with enough scientific and engineering research.