History of AI

What is Blade Runner about, and why do people bring it up when they talk about AI?

· By Robert Scales · 5 min read

Short answer

Blade Runner is Ridley Scott's 1982 film about Rick Deckard, a police "blade runner" in a rain-soaked Los Angeles of November 2019 who must find and kill four escaped replicants, artificial people almost impossible to tell from humans. Released on 25 June 1982 and based on Philip K. Dick's novel Do Androids Dream of Electric Sheep?, it keeps coming up in AI debates because it asks how we would tell a machine from a person, and whether the difference would still matter.

Blade Runner is set in Los Angeles in November 2019, a date that has just slipped into the past. The real city never got flying police cars or towering video billboards over every street, and nobody is hunting artificial people. Yet the film's central puzzle feels closer now than it did in 1982. When a machine can talk, write or look like a person, how do you tell the difference, and what do you owe it?

The film opened in American theaters on 25 June 1982. It was directed by Ridley Scott, written by Hampton Fancher and David Peoples, produced by The Ladd Company with backing from Sir Run Run Shaw, and distributed by Warner Bros. Its stars were Harrison Ford, Rutger Hauer and Sean Young.

What happens in the story

An opening text sets the scene. The Tyrell Corporation has built "replicants," artificial beings almost identical to humans, stronger and at least as intelligent as the engineers who designed them. They are used as slave labor on colonies in space. After a bloody revolt, replicants are banned on Earth, and special police units called Blade Runners track down any who return. The film's chilling euphemism for killing one is "retirement."

Rick Deckard (Ford) is a retired blade runner pulled back for one more job. Four replicants of the newest Nexus-6 model have come back to Earth, led by Roy Batty (Hauer). They were built with a four-year lifespan, and they want their maker, Dr. Eldon Tyrell, to give them more time. While on the case, Deckard meets Rachael (Young), Tyrell's assistant, who turns out to be a replicant who believes she is human because she was given implanted memories.

The film's emotional peak is Batty's final scene. Having fought Deckard across a rooftop, he saves the man sent to kill him, and then, as his own life runs out, speaks about the things he has seen that will be lost, in his words, "like tears in rain."

How the film was made

The novel by Philip K. Dick, Do Androids Dream of Electric Sheep?, came out in 1968. Producers had tried to adapt it through the 1970s. Ridley Scott was attached to direct in 1980. Working titles included Dangerous Days and Mechanismo; the final title was borrowed from a book by William S. Burroughs. Reported budgets range from $22 million to $30 million.

Much of the look came from the designer Syd Mead. He was hired to draw the futuristic cars, but his background sketches won over Scott and helped shape the whole city. Crews redressed an old New York street set on the studio lot in Burbank into a crowded, decaying metropolis. Real Los Angeles buildings appear too, including the Bradbury Building downtown and Frank Lloyd Wright's Ennis House. Philip K. Dick died on 2 March 1982, about three months before the release, and the film is dedicated to him.

Reviews were mixed. Critics admired the images, Jordan Cronenweth's cinematography and Vangelis's electronic score, but many found the story confusing. It earned $6.1 million on its first weekend in 1,290 theaters, against stiff competition from E.T. and The Thing, and it disappointed at the North American box office. Its reputation grew on home video. A director's cut without Deckard's voice-over narration appeared in 1992, and Scott's own Final Cut in 2007. In 1993 the Library of Congress added it to the National Film Registry, and a sequel, Blade Runner 2049, came out in 2017.

How you tell a replicant: the Voight-Kampff test

The film's most interesting idea for anyone curious about AI is its test. Blade runners use a device called the Voight-Kampff machine, a sort of lie detector. It asks questions designed to provoke an emotional reaction, many about cruelty to animals, while measuring breathing, blushing, heart rate and the movement of the eye. Replicants, the theory goes, are slower or flatter in their empathy.

It helps to compare this fictional test with the most famous real one, Alan Turing's imitation game of 1950, now called the Turing test:

QuestionTuring test (1950)Voight-Kampff test (film, 1982)
What is being judged?Conversation, typed through a screenThe body's reactions to emotional questions
What counts as "passing"?The judge can't tell the machine from the personThe subject reacts like a human, with empathy
What does it assume?Thinking shows in what you sayBeing human shows in what you feel
What happens to a "machine" that fails?Nothing; it's a gameIt is "retired"

Notice how the film flips Turing's question. Turing asked whether a machine can seem intelligent. Blade Runner takes that for granted and asks whether a machine can feel, and whether its makers care.

Try it: Write three questions you think would separate a person from a very good chatbot. Then ask yourself honestly whether each question tests intelligence, feeling or just knowledge of human habits.

Why it matters for AI

Replicants are not computer programs. They are engineered biological beings, closer to Mary Shelley's creature than to a robot. Still, the film became one of the standard reference points in discussions of artificial intelligence, for three reasons.

First, it puts the burden on detection. As fake voices, faces and text get better, the practical question of how to recognize a machine is no longer fiction. Language models such as OpenAI's GPT-2, released in full in November 2019, can write paragraphs that many readers take for human writing.

Second, it raises the question of rights. The replicants want what any living creature wants, more life, and the humans treat them as property. Philosophers and ethicists continue to ask what we would owe a machine that could truly suffer.

Third, it blurs the line. The film never settles whether Deckard himself is human; Scott has said that in his vision he is a replicant, while others disagree. If the hunter cannot be sure what he is, the test itself becomes the problem.

Could we build a real Voight-Kampff test for today's AI?

Not in the film's form. Present-day AI systems have no heartbeat, no pupils and no blush to measure. A voice assistant or a chatbot is software running on servers; it has no body to react, and there is no evidence that it has feelings to detect.

What researchers do build are detectors for machine-made content. When OpenAI released the full GPT-2, it also published a detector that estimates whether a passage came from that model, and it caught about 95 percent of such text, which OpenAI said was not enough on its own. Detectors like this look for statistical fingerprints, patterns of word choice or pixels that machines tend to leave, rather than for missing emotions.

The deeper question the film asks is still open. A test can tell us whether something behaves like a human. It cannot tell us whether there is anyone inside. That is a question for philosophy and science together, and Blade Runner remains one of the most watchable ways to start thinking about it.

Sources

  1. Blade Runner (1982) — AFI Catalog of Feature Films (American Film Institute)
  2. Blade Runner — Wikipedia (archived 26 January 2020), January 26, 2020

Get the free Guide to Artificial Intelligence

Sign up for the weekly newsletter and receive the free guide: what changed in AI this week, one lesson, one prompt to try. Unsubscribe in one click.

How we handle your address

Back to top

Display

Your choices are saved in this browser only. No data is sent.

Text size
100%
Colours
Reading
More settings
Accessibility statement