Since AI has turned into a buzzword, many misperceptions created too many myths we`d like to share with you and bust together just like Adam Savage and Jamie Hyneman. Even though we are living right within the Secondary Information Age, the lack of understanding of what actually is AI and where it can bring the society we are all living in, it is a foundation of quite a few misconceptions. Some of them funny, other rather scary.
AI robots killing us
Well, thank you Hollywood and sci-fi literature! Such a statement is a simple verbalization of a natural apprehension of what a machine can do to a human being. If we put the subject in a broader perspective, we should have in mind that the main concern is going to be the misaligned intelligence with an internet connection. It doesn’t need an actual body to kill us.
AI controlling people
Well, let us take elephants for instance. They are rather smart animals, they are quite big respectively stronger than us. Why then the elephants are domesticated, playing in circuses, used as working animals in rural areas, etc. It is because we are smarter. Not bigger, nor stronger – just smarter. The truth is that intelligence enables control.
AI cannot have a goal
We`re sorry, it can if you made it to. Just think of the heat-seeking missiles, they have a specific goal and it is that plane they are chasing after being fired by humans. So once again it seems like the AI is not the invisible menace we are afraid of. It is the humans behind it.
AI turning evil
Continuing the abovementioned myth, this one is pretty close. AI can`t behave badly unless the people controlling it ask it to do so. The actual worry, as stated by Future of Life publication should be AI turning competent with goals misaligned with ours.
AI learns on its own
Not at all or not yet if you prefer. AI is an algorithm performed by machines. Learning skills in AI domain are possible but they require human intervention on multiple levels. Any system that has to deal with hidden information or uncertainty of any kind cannot be understood by AI, which still needs to be fed with input data by humans. One more thing – you can teach AI to define whether there is a kitten on a picture. If you ask, it to do this by feeding it with dogs pictures and after multiple iterations, AI will forget how the kittens look like.