Since mastering the art of walking upright, humanity has sought innovations to aid in their quest for survival. Where fire and primitive tools gave early adopters a leg up in hunting, health, and protection, subsequent iterations of progress have only grown more complex. Sprinkle in some heavy alien overtones, and this is more or less the plot of Stanley Kubrick and Arthur C. Clarke’s masterpiece 2001: A Space Odyssey, developed concurrently as a novel and film in 1968. For our ongoing discussions of the chatbot arms race, it felt only proper to revisit one of science fiction’s most iconic examples of AI gone rogue: HAL-9000. However, as we’ll learn, the cultural memory of HAL’s unyoking is more a product of human limitations than technological resistance.
On its face, the story documents humanity’s climb from the morass of the animal kingdom to creating AI in its own neurological image. More specifically, both the film and novel renditions posit a connection between extraterrestrial artifacts and the maturation of human intelligence. The film accentuates this point by accelerating humanity’s technological progression, jumping from bone clubs to interstellar space travel in record time. However, we’re going to sidestep the issue of alien lifeforms to focus on the relationship between the human and AI characters in the film. Like Mary Shelley’s Frankenstein, a similar horror lurks in the struggle to outsmart an opponent of one’s own making.
As we’ll discuss in this analysis of 2001, in which we find parallels with today’s rapidly advancing AI technology, science fiction’s portrayal of the future often omits the difficult process of experimenting with trial and error. Far from happening overnight, these advancements come in fits and starts, with periods of dreaming and refinement in between. However, this does little to quell our collective desire for the worlds we’ve imagined to be made readily available. We see this time and again with the pump of promising ideas that often only work on paper. But even more concerning, without proper vetting, regulation, and input from all parties affected, rushing the adoption of new, viable tools can result in delayed or pervasively harmful consequences.
“Everything is going extremely well”
The opening sequence of 2001: A Space Odyssey has become the stuff of legends and endless parody. When a monolith appears to a group of primates on an ancient Sub-Saharan plain, the experience has a profound effect on the animals, who all display visible agitation. It’s inferred that this experience leads to their use of discarded bones as tools for hunting and fighting off rival bands, as the group establishes dominance.
Fast forward to the film’s present, and another monolith has been discovered in a crater of the Earth’s moon. Inhabitants from a nearby lunar base are seen monitoring the site, excavating the surrounding perimeter. Suddenly, the monolith is hit with a beam of direct sunlight, and begins emitting a high powered signal. This disorients and overwhelms the workers, before spiraling outward into space.
The film’s next major movement cuts to Discovery One, a shuttle en route to Jupiter. Astronauts Dr. David Bowman and Dr. Frank Poole are transporting a group of scientists in suspended animation on an undisclosed mission. To assist them in their navigation, the shuttle’s AI-mainframe computer, HAL-9000, controls much of the ship’s functioning. These are some of the more calming, cinematic moments in the film, where the viewer is right to marvel at the wonders of human creation.
The journey is progressing smoothly until HAL reports the imminent failure of an antenna on the vessel’s exterior. Dr. Bowman quickly boards an extravehicular activity (EVA) pod to retrieve the device from outside the ship, only to discover nothing is wrong. While discussing with HAL whether to let the antenna fail again to diagnose the problem, the crew are interrupted by an inbound message. Mission Control’s HAL unit has detected a flaw in their own HAL’s programming. Discovery One HAL refutes the claim, and blames the irregularity on human error. As it turns out, this is clever foreshadowing.
HAL-9000; 2001: A Space Odyssey (1968)
At this point, the astronauts become concerned about their AI’s functioning, and retire to the pod bay to discuss their next move. Believing they are outside of HAL’s range, the pair agree to allow the antenna to fail, and disconnect HAL if it turns out that his programming was in fact on the fritz. Unbeknownst to the crew, but conveyed through clever editing, it’s revealed that HAL can read lips, and is now aware of their plot.
“I’m afraid I can’t do that, Dave”
The dynamism of HAL’s presence in the film is reinforced through a few cinematic techniques. Depicted as a single red light bulb nested within a fisheye lens, HAL’s image frequently fills the screen after moments of dialogue between crewmembers. Shots of HAL’s unblinking terminal signal the AI’s persistent omniscience, and help establish a sense of foreboding that mounts with each subsequent appearance. The hum and whirr of the ship’s interior often blanket these scenes, as if the audible churn is HAL’s continual processing of all available data.
But the most striking feature of HAL’s on screen presence is the dulcet and detached narration of Douglas Rain who voices the AI. The polished, aloof tone lends the impression of servitude, yet the power HAL wields on Discovery One should be familiar to anyone who’s received directions from Siri, or posed a question to Alexa. The slippage in these moments where the product, or AI, is subsumed with authority by its privileged relationship to both action and information. A user, or crewmember, is at once in command, and deferential to their device’s operating system, or ship’s central control. In essence, this creates a two-way power dynamic that often goes uninterrogated.
This relationship is forced to the surface once HAL realizes the present threat posed to him by David and Frank. At this point in the film, none of the crewmembers are aware of the mission’s ultimate purpose. In turn, HAL was programmed to only reveal this information at the appropriate time, a detail that is unpacked with more clarity in the story’s novel form. However, this is not made explicitly clear to viewers of the film. This often results in a misreading of HAL’s reaction to discovering the intent of the crewmembers. Namely, that HAL’s decision is not an act of rebellion, but rather the result of conflicting presets in the AI’s programming. With tensions rising on Discover One, HAL is forced into an ultimatum: reveal the details of the mission prematurely, or neutralize the threat of having to do so?
At the possibility of being disconnected, HAL enters survival mode, a turn of events that leaves Frank dead and also puts David on the defensive. With HAL in command of the vessel’s primary controls, Dr. Bowman is forced to outsmart the AI and manually override his way into the ship’s mainframe. From here, he begins the difficult task of decommissioning HAL.
Dr. David Bowman in the heart of Discovery One; 2001: A Space Odyssey (1968)
Upon entering the womb-like motherboard of Discovery One, David floats through anti-gravity as HAL’s monotone protests fill the chamber. The hissing sound of an air leak, punctuated with David’s helmeted breathing, adds a heightened note of anxiety to the scene. As Dr. Bowman locates the memory console, HAL continues to plead with him to halt his actions. With each successive cartridge that’s disengaged, another piece of the AI’s functioning is deactivated. HAL laments throughout this process, “my mind is going, I can feel it” and repeats the phrase like a refrain, before essentially defaulting to factory settings.
This regression back to a childlike state is exploded into relief as HAL offers to sing David a song. Now a fraction of its former self, HAL’s recital is laborious, with each word becoming slightly more elongated. Then, slowed and diminished, HAL goes offline. Where the film’s initial portrayal of technology seemed to suggest that human beings had superseded their innate frailty, at this point, such illusions have been shattered for the viewer. Rather, David’s struggle to reassert control on Discovery One is underpinned by the film’s opening sequence in the Sub Sahara. Only now, the chief aggressor is a simulacrum of humanity’s own creation.
Dave and HAL; 2001: A Space Odyssey (1968)
More lessons from the recent past
At first glance, 2001 appears to be another archetypal story of a wayward creation. However, a more careful understanding reveals the banal truth that even in this future, we’re still grappling with our own shortcomings. HAL-9000, like many wrecked aircraft that came before it, is just a stepping stone en route to a more perfect machine. It’s in that recalibration after failure where the clearest hindsight can come into focus. However, our tendency to marvel at the fruits of our labor often supersedes the painstaking effort required to work out all its kinks. And even then, anything we create is subject to bias or human error that can play itself out in a variety of cruel and unexpected ways.
In her book Algorithms of Oppression, Dr. Safiya Umoja Noble explores how search engines perpetuate racial and gender stereotypes. Highlighting how women and minorities are being marginalized by these tools, Dr. Noble argues the technology itself helps diffuse and depersonalize who’s ultimately responsible for these outcomes. This line of questioning is especially prescient when considering the persistent pay and hiring gaps that exist in tech spaces for women and people of color. Additionally, the notorious “anti-diversification” manifesto circulated by a former Google search engineer in 2017 added further credence to the persistence of Dr. Noble’s inquiry. Where such traits already cause incalculable damage through free internet services, their potential for destruction would be made exponential if paired with the agential capabilities of a HAL-9000.
In these early days of consumer-grade AI, it’s tempting to feel a certain excitement about the potential that awaits. And yet, if classic films and ongoing societal problems can offer any guidance, we have a lot to learn about ethically integrating new technology into our messy human existence. Like the sleek and eventually flawed Discovery One and its HAL unit, anything we build will inherit our strengths as well as our weaknesses. But that doesn’t mean we shouldn’t always strive to fail with a greater sense of purpose.
Note: Exchange Plus is currently not available in the U.S. Check the list of supported jurisdictions here.