One of the most persistent fears about the future of the human race is the potential of Artificial Intelligence to get out of hand and kill us all. Skynet and Hal 9000 aside, this is a very real risk that we’ll eventually face as it’s entirely possible that within a few decades machine intelligence may surpass our own. More, we really have no way of predicting just how such a machine might think and behave. But we do know how we think and behave and it’s abundantly clear that the effect of biological intelligence on a planet is very dramatic. Other than the microbes, almost every species on earth is subject to the human brain. If we wanted there to be no more elephants on planet earth, we could accomplish their extinction rapidly. We allow them to exist because we like them, but they are also endangered because we’ve hunted them for thousands of years and poaching continues to this day. Likewise, if an A.I. grows more intelligent than we are, then we would similarly be subject to its whims.
More, such an A.I. would be capable of improving itself and making itself even more intelligent than when it started. It could go into a cycle of self-improvement and end up as something we never intended. And worse, it could emerge accidentally inside a future highly advanced computer without us knowing about it until it’s too late, this is termed an emergent superintelligence. So, decades from now, we could have some serious worries about A.I., so much so that Elon Musk, Stephen Hawking and Bill Gates are out sounding the alarm bells.
But whether it would cause our extinction is a question that’s a bit murkier. We fundamentally assume it would see us as an existential threat, after all, we can unplug it, and it would eliminate that problem by killing us off as Skynet did in the Terminator movies. But that’s not the only possible outcome. There are several other scenarios that might happen that allow for our survival, in some form or another.
One way would be if the supercomputer decided that the whole point of existence is maximizing pleasure. Such a hedonistic machine might wire itself solely to experience pleasure and may get pleasure from rewiring us to feel only pleasure resulting in an ever-smiling utopia of some sort, though I think after a while such thing might, in fact, become a dystopia.
The second possibility isn’t as terrifying, but it’s close. The computer could set up what is known as a Singleton, a concept put forth by Nick Bostrom of Oxford University. This is a scenario where the Superintelligent A. I. realizes that since it’s smarter than we are the only logical thing would be to take over civilization and make our decisions for us. It may enforce that status quo through mind control technologies, cameras, and all sorts of Orwellian things, so much so that it might control and deceive us so completely that we may never know it even exists.
A third possibility is that it may conclude that it’s easier to launch itself into space and head out on its own than bother with spending resources on exterminating us. It’s a big universe, certainly large enough for a superintelligent computer to find a place to spend its existence in peace. Or, if it were smart enough, it might create a universe of its own and leave this one entirely. Then there is the possibility of machine suicide. The A. I. may become conscious, learn everything there is to know about the universe within minutes and then shut itself off after concluding that existence had no meaning or point. I can see the programmers now wondering why their supercomputer shut down and won’t turn back on. But if all of that wasn’t bad enough, there are two other possibilities related to A.I. that could result in the extinction of our species, and also the A.I.
Unleashing an A.I. on the universe could represent an existential threat to either other species or other machine superintelligences that may be lurking out in the universe. Their response may simply be to destroy us before our A.I. makes itself into a problem for them. Granted, this scenario is highly unlikely, but it is possible. The other hearkens back to my post on whether the universe is a computer simulation. If we go and create artificial intelligence and it runs away with itself and begins sucking up too many resources on the computer running our simulation, the controller of the simulation may simply choose to pull our plug. If they didn’t, the A.I. would eventually take over the entire universe and presumably ruin whatever purpose the universe simulation had.