ARTIFICIAL INTELLIGENCE So, what is the sentient or ‘universal’ artificial intelligence that science fiction has frightened us so much about? What exactly is ‘narrow’ AI? Take, for example, the voice-activated assistant on a smartphone. Though your iPhone’s Siri assistant appears to be clever, it isn’t capable of true thought - at least not beyond the limited capabilities that machine learning and deep learning have enabled you to encode into it. It’s an example of ‘narrow’ AI, or more specifically, a hybrid AI that uses a bunch of ‘narrow’ AI abilities (mainly voice recognition) to tap into the vast reams of constantly updated data in the cloud to create the appearance of superhuman intellect.
Siri’s ability to instantaneously access all of the internet’s knowledge does not imply that it is capable of creative thought or reasoning; it is just performing a duty that it has been trained to do exceptionally well. Only when Siri is capable of performing any action that a human can perform will it be called a true AI. And no artificial intelligence has ever been able to do so — at least not yet. So, how far-reaching is AI going to be? In this case, there are two schools of opinion. The first group includes people like Bill Gates and Stephen Hawk-
ing, who are concerned about the rise of superintelligence. They see a day when the development of a ‘general’ AI triggers a ‘intelligence explosion,’ in which AIs are capable of teaching themselves new skills at a breakneck pace, eventually surpassing human capabilities to the point where they enslave humans. This bleak future image, dubbed the ‘technical singularity,’ is mitigated by the fact that it is more of a philosophical worry than a practical one — there is no factual proof that AI will reach this point. Likewise, there is no undeniable proof that it could not happen. The second group sees AI’s advancement in a significantly more practical light. The intelligence explosion will not happen, but today’s ‘narrow’ AI will continue to dominate. That isn’t to imply that ‘narrow’ artificial intelligences won’t be extremely beneficial in the future. Their abilities will expand at an exponential rate, and while they won’t be able to think creatively, they will become extremely proficient and efficient in hyper-specific tasks where context is limited and nuanced interpretation is unnecessary. It’s here that the colliding realms of AI and robotics will have the greatest impact on people. 46