HAL 9000: A Struggle Against Self Conflict
Caution, this content contains spoilers for the movie 2001: A Space Odyssey!
While the film 2001: A Space Odyssey, directed by Stanley Kubrick, contains many mysteries, such as the monoliths and the space baby, this blog will take a deeper dive into HAL 9000 to provide analysis on intelligent agents’ ethics and safety based on HAL’s actions throughout the movie. First, let us grasp a brief understanding of the plot and what HAL 9000 is.
HAL 9000 is an artificial intelligence computer equipped with the Discovery One spacecraft. It has a range of capabilities necessary to achieve mission success, such as speech recognition and image processing. However, there are actions that HAL cannot perform, thus requiring human intervention. David Bowman and Frank Poole are two of the five crew members that were assigned to Discovery One. Bowman and Poole will be the ones working with HAL. Simultaneously, the other three crew members, who are scientists, remain in stasis for the remainder of the flight until they reach Jupiter. The crew’s mission was to reach and investigate Jupiter with the aid of HAL 9000.
For this analysis, we will refer to HAL as our intelligent agent. But before we dive into the main problem regarding ethics and safety, we will first analyze and identify the performance measures that our intelligent agent abides by. We will then discuss how these performance measures ultimately turned HAL into the main antagonist of the film.
Performance measures are goals and qualities that we want our intelligent agent to possess. We will start identifying HAL’s performance measures based on what the movie’s plot provides us. In the film, HAL was first introduced to us through an interview with a news reporter. Through this interview, we discover that HAL 9000 can emulate almost all human brain activities. It was also emphasized that HAL has never made a mistake or has had an error since it came to service. It is known for its speed and the ability to produce accurate and reliable information. Later throughout the movie, it was revealed that the true mission of Discovery One was to investigate alien life in Jupiter. HAL was programmed to keep the actual mission a secret from Bowman and Poole. Therefore, we conclude that HAL’s performance measures are the following: fast, reliable, accurate, maximize the chance of arrival to Jupiter, maintain life support for crew members who are in stasis, aid the crew members of Discovery One, ensure the spacecraft is operational, and prevent Poole and Bowman from discovering the true mission.
Now that we have established HAL’s performance measures, we can move on to the root of the problem. It all stems down to the fact that HAL was instructed to keep the real mission a secret from Poole and Bowman. If Poole and Bowman were to ask HAL what the mission is, HAL would have to lie to them to maintain the secret. However, HAL was primarily programmed to provide fast and accurate information processing. The act of lying contradicts HAL’s primary programming. Therefore, this contradiction causes HAL to execute a sequence of actions that caused most of the crew members’ death, which led to its shutdown.
The takeaway regarding the case of HAL 9000 is that it is crucial to consider tradeoffs when two or more goals within the performance measure are conflicting with one another. The moment HAL decided to kill the three scientists/crewmembers in stasis, the mission was already jeopardized. Without those three scientists, it would be impossible to study and investigate alien life in Jupiter. Programming HAL to produce/process very reliable and accurate data is an example of the King Midas problem. We may think it is suitable for a machine to produce/process accurate and reliable data. However, it is not ideal in reality because there are actions such as lying that contradict reliability and accuracy.
Teaching a machine how to lie is difficult because there are multiple ways of creating a lie through manipulation and deception. What if we built a machine that is capable of lying? How can we tell whether it is lying or telling the truth? In the film, there were many occasions where the audience was unsure whether HAL was stating the truth or not. HAL’s role in the film is a great example of a scenario that we must be cautious of when creating an intelligent agent.
