The rise of Artificial Intelligence
I was inspired to write something on the state of artificial intelligence, or AI, after watching Joe Rogan’s podcast episode where he discusses robots with Elon Musk. When discussing a video that surfaced of a Boston Dynamics robot doing a backflip, Rogan expresses abject horror and fear at just how humanistic the robot is, and ponders how quickly technology is now advancing.
Musk does absolutely nothing that will make any of us feel any better, as he then tells Rogan that within a few years the robot will be able to move so quickly that we’ll need strobe lighting to be able to see it…gulp.
AI has often been the subject of fantasy and fiction, spawning films, novels and stories about the future and what to expect from machines once they become self-aware and able to improve themselves. The Terminator being a good example of what we don’t want to happen. Unsurprisingly Musk isn’t too confident on the matter and listening to him speak on the subject too much is likely to make you want to drive your car off a bridge.
Recent...concerning developments include an AI robot passing its medical examinations in China, a bot which beat the best poker player in the world, and most worryingly also learnt how to fly drones by itself.
Wired also reported that AI can now replicate video footage through CGI that’s almost impossible to tell from real footage. The implications of such a scenario are obvious, and there’s a very real and understandable risk of AI becoming rapidly advanced and powerful and falling into the hands of just a few tech company CEO’s who could become real life villains with very little effort.
Is iRobot going to come to life over the next few years? Probably not, but it’s not impossible to conceive. Musk himself thinks we have a 5 to 10% chance of controlling it once it becomes self-aware.
What are the creepiest things to happen in AI recently? Well I’m glad you asked. Stephen Hawking, just before his death, said of AI “I fear that AI may replace humans altogether. If people design computer viruses, someone will design AI that replicates itself. This will be a new form of life that will outperform humans.” Which isn’t exactly a ringing endorsement of new technology.
This year an AI robot called Sophia was given citizenship and rights in Saudi Arabia, making her the first robot to have a nationality and gender. This raised interesting questions about whether, as a woman, she would have more rights than human women in the Kingdom, where women have only recently been allowed to drive.
Creepy, but perhaps not as creepy as the fact that the NSA in America has an actual AI programme called SkyNet which uses mobile phone metadata to track and predict the movements of terrorists. Why have they called it SkyNet? Nobody actually knows, but WIRED reported that SkyNet looks for terrorist connections based on questions such as “who has travelled from Peshawar to Faisalabad or Lahore (and back) in the past month? Who does the traveller call when he arrives?” It also looks for suspicious behaviours such as someone who engages in “excessive SIM or handset swapping" or receives “incoming calls only."
A programmer who wrote an AI algorithm to get the highest possible score on NES games was taught how to see which moves or actions yielded the most points and repeat them as much as possible to get the highest possible score. It worked fantastically well for games such as Super Mario and PAC man, but failed miserably when it was asked to play Tetris. Why? You’re asking yourself. The reason was that the programme figured that because you scored a point when the shapes hit the bottom, the trick to scoring the most points was to land them as quickly as possible, but the problem was that the shapes quickly stacked up and then the bot failed and lost. Rather than trying to figure out a new way to look at it, the bot figured that it would just quit the game and paused it forever.
One of the most worrying stories came from the US military when they built a drone titled X-47B, tasked with using almost complete AI programming for flying and operations. This drone required the least human interaction in history, and did very well…for a while. X-47B made history when it became the first unmanned aircraft to land itself safely on an aircraft carrier. When tasked with going out on a third practice run the craft was asked to land once again on the aircraft carrier, but it never returned. It decided…on its own…that it would be safer to land on a runway on the mainland instead. Shortly after this the army defunded the project and shut it down.
Whether or not we’re going to be eradicated by super smart AI remains to be seen. Many aren’t particularly confident that we won’t be replaced almost immediately, but there is the possibility that self-aware machines may well want to care for us and look after us. – We hope.