The Imitation Game

The movie “Imitation Game” amazed me! It was exceptionally well-written. It had war, spy intrigue, friendship, romance, computer science, and addressed Alan Turing’s homosexuality. It is an enormous achievement for a script to pull together so many diverse threads in a way that makes sense, with just the right amount of each one.

So why isn’t this a blockbuster mega-hit? Because this is America, where we prefer shallow techno action thrillers. Sure, there was some eye-candy. A few of the war shots (particularly a flying fortress, well, flying) screamed computer graphics. But most of the movie showed people interacting in fairly normal environments.

It was clever to approach the subject of Turing’s homosexuality via a heterosexual relationship. He forms a strong platonic friendship with the woman on the team, and proposes marriage so they could keep working together. Later he confesses his homosexuality as a way of driving her away, specifically to protect her from the growing intrigue around them. She says she knew, and that she still wants to be with him for their intellectual connection. In a heart-wrenching twist, he succeeds in breaking the relationship anyway.

Can’t speak to the historical accuracy, which is apparently quite bad. As a computer scientist, I think more about Turing’s theories and philosophical questions rather than his personal life.

One of the most clever twists of the movie was to turn the classic “Turing test” into another kind of question: was Turing a criminal or a hero? Otherwise, the movie did very little to address the “Imitation game” or its meaning. Since that is lacking, let me tell you about it…

The fundamental question (also mentioned in the movie) is whether machines can think. Turing wrote a paper in 1950 called Computing Machinery and Intelligence. It is popular-level writing, so you don’t need to be a computer scientist to read it. I recommend it to everyone.

The Imitation Game was a party game in which a judge exchanges messages (think texting here) with two people, a man and a woman. The judge must determine their genders. The man’s goal is to trick the judge into the wrong choice, and the woman’s goal is to get the correct choice. This turns into a Turing test when the man is replaced by a machine.

This test is poorly constructed for several reasons. One is that gender is an unnecessary variable. In the modern version the judge texts two entities, a machine and a human, and must decide which one is which. Both the human and machine want to convince the judge they are human. This is still poorly constructed, because the judge doesn’t really take each entity on their own merits. Instead, the judge knows that one entity is lying.

A post-modern version appears in my novel SuSAn. There, the judge interacts with a humanoid robot (only one), which may be operated either by a human or by an artificial agent. There is a large pool of humans and agents, in equal number. There are also a large number of judges, including people who watch the interaction. I would argue this setup has far more validity than the recent highly-touted test run by the University of Reading (see my blog post on this).