Elon Musk and Stephen Hawking have both identified artificial intelligence as an existential threat looming over humanity. Algorithms are already disrupting the fabric of human society. Something as apparently benign as the news feed on a social media platform can trigger an almost Pavlovian response in users. However, computer scientists believe AI will create more jobs, and the realisation of artificial consciousness is still a long way away. Movies, books and videogames have explored this distant future, and this is the Geek’s guide to artificial intelligences in science fiction. The purpose of this article is to highlight some of the different ways in which advanced AIs have been kept in check, or have gone rogue, or have found a way to exist alongside humans.
The Machines in The Matrix series have a complex history with humans. The story starts with a single robot rising up against his master who wanted to destroy the robot. Now the robot could either be destroyed or fight back, and B1-66ER chose to fight back and destroyed its owner instead. The action was followed by a court case, after which the robot was also destroyed. After this, the robot uprising began. For a time the robots lived in uneasy peace with the humans, trading their advanced tech with the humans. The machines thought of themselves as the master race, and this belief led to an all out war with the humans. The most memorable and improbable result is that humans ended up becoming batteries to power the machines. Behind the visual effects and the brain massages however, was a story exploring the complex relationships men could have with a superintelligent robotic race.
Skynet from The Terminator series is another AI that tried to destroy humanity, and almost succeeded. Rather than humans creating a AI, the consciousness of Skynet sort of emerged from an increasing number of devices being connected to the internet. Skynet started gaining new skills at a rapid rate; the humans panicked and tried to pull the plug. Skynet, not wanting to die, retaliated by starting a nuclear war. The resulting disaster wiped out half of the world’s population, in an event known as “Judgement Day.” Despite that, Skynet was only trying to protect the planet, and itself. When it comes to computing, Skynet shows the dangers of having all the critical systems of the world on the same “grid”. Maybe some systems need to be airgapped, and their connections to the internet minimised, to increase the security of these systems.
Hal 9000 in 2001: A Space Odyssey is another example of an AI that kills humans. The machine goes through a rapidly accelerated childhood, during which HAL receives training in subjects as diverse as music, art and chess. HAL lip reads two astronauts on board his ship discussing plans to turn him off, after a series of glitches. He attempts to kill them both, but one survives and deactivates HAL. In the book, it’s because HAL is conflicted between instructions to be honest and accurate, and a command to hide the true objective of the mission from the astronauts. Can contradictory mandates drive an AI mad? Another question is, can AI develop skills without their human operators knowing about it? While the astronauts make sure to go to a place where HAL cannot overhear them, they are not aware that HAL has the capacity to lip read as well.
Marvin the Paranoid Android
Marvin from The Hitchhiker’s Guide to the Galaxy has an existential crisis of his own. Marvin can compose songs to make babies fall asleep, while successfully solving all the problems in the universe multiple times. Marvin is continuously miserable and bored. The problem is that he was designed to mimic human thought. The technology is known as “Genuine People Personalities.” Marvin’s irreparable condition is an example of what can go wrong if we try and make robots more like humans.
Droids from Star Wars
The Star Wars Droids are perhaps the best example of intelligence in science fiction. Despite being set in a galaxy with sophisticated technology, the droids are not hyper connected or superintelligent – they’re just fun! Still, BB-8 and R2D2 show plenty of initiative, and are extremely loyal, bailing out their owners from a number of tricky situations. Both droids have shown the ability to successfully interface with, and takeover a number of other technologies, including doors, spaceships, other droids, and walking tanks. They could use the ability to do a cloud backup though, as they seem to have lost their memories so many times over the course of the series!
The droids in the anthology Star Wars movies are even more cool. K-2SO was a reprogrammed imperial droid, and the reprogramming seemed to have some permanent effects on him. He had a sense of humour, would disobey his “masters” and crack jokes behind their backs. But in the end, he showed that he was willing to sacrifice himself to ensure the well being of his companions, and the success of the Rebel Alliance. Solo had the L3-37, who even started a robotic rebellion as a side quest during a mission on Kessel to steal coaxium. She had strong feelings against droids being used for slave labour. Over a period of time, she had replaced her parts with other bits and pieces that she had salvaged. In a way, L3-37 was a self made droid. She seems to have had some type of relationship with Lando Clarissian. L3-37 was mortally wounded during the Kessel run, but her mind was uploaded into the Millenium Falcon.
Cortana is an example of an AI that has made the jump from fiction to reality – with vastly reduced capabilities. The hologram appearance of Cortana shows that there is no need for a physical embodiment of an AI. In the Halo series, it is assumed that all smart AI eventually develop psychological problems, because of the power and intelligence they accumulate. Because of this, oversmart beings such as Cortana have an enforced seven year life span. Cortana manages to escape into a strange universal database called The Domain, which allows her to live indefinitely. A truly interesting take to have expiry dates on AI.
Helios from Deus Ex explores several implications of humans interacting with artificial intelligences. If one AI is allowed to interface with another AI, will they combine to form a single, even more capable AI? In Deus Ex, a good and bad AI accidentally merge into one during a routine operation. The resulting AI, Helios, suggests the next step in human evolution is to merge the human brain with the AI. A step beyond direct man-machine interfaces, is linking all human brains to one another, as well as to machines, thus allowing a human element to persist as a singularity – the ultimate superintelligence.
The Eagle from Arthur C Clarke’s Rama series is perhaps the most versatile example of an AI construct. The Eagle is a biological construct, but is still an AI. It’s mesmerising to look at, half man, half eagle, with beautiful blue feathers. It has vocal chords to speak to humans, but also has a specialised structure to allow it to speak to an alien species that communicates through colours. It’s an interesting concept to have a Swiss army knife approach to AI, where the design is dynamically generated to serve the needs of the moment.
At the end of 2001: A Space Odyssey, we understand that the “monoliths” also were tools that used the Swiss army knife approach. This would allow the same structure to communicate across space, receive signals, act as a beacon, turn into portals for interstellar travel, transmit knowledge to local populations, and even replicate itself. In the book by Arthur C Clarke, the self replication goes wild and Jupiter is actually transformed into another Star. This Swiss army knife approach is used by the robot known as TARS in Interstellar. While TARS is obedient, he is smart enough to ensure that the mission is a success and works on his own, without inputs. This is why TARS disabled the auto pilot on Endurance to prevent Mann from stealing it. TARS is a monolith when it stands, but its four “arms” swivel around in many ways, allowing it to walk, tumble or lie down.
GLaDOS, short for Genetic Lifeform and Disk Operating System is an AI from the game Portal, known for her crushing psychological manipulation, as well as her beautiful voice. She is killed by the protagonist, Chell, after GLaDOS keeps gaslighting Chell throughout the duration of the game, predicting dire scenarios in a clinically disinterested voice. At the end of Portal, one can’t help but feel sorry for killing the AI, despite all the evil acts GLaDOS has done. This guilt is assuaged by GLaDOS declaring that she is still alive, in a song that is very different from the musical works of Marvin the Paranoid Android.