Humans Won't Be Able to Control Artificial Intelligence, Scientists Warn
Artificial intelligence is just a set of algorithms programmed to react to a set of inputs and outputs. But what if that algorithm was able to change on its own or become smart enough to learn without humans knowing how they do it? That’s exactly what scientists fear.
In a paper published in 2016 in the Journal of Artificial Intelligence Research, scientists warn superintelligence cannot be contained.
“Superintelligence is a hypothetical agent that possesses intelligence far surpassing that of the brightest and most gifted human minds. In light of recent advances in machine intelligence, a number of scientists, philosophers, and technologists have revived the discussion about the potentially catastrophic risks entailed by such an entity,” reads the paper.
The warnings issued by the authors is a familiar plot used in many apocalyptic films involving artificial intelligence: The Terminator, The Matrix, and Netflix’s recently released original film Outside the Wire.
Humans would do well to install fail-safe mechanisms on artificial intelligence, but that isn’t good enough. The authors of the study say well-intentioned control mechanisms can easily backfire.
“A superintelligence given the task of maximizing happiness in the world, without deviating from its goal, might find it more efficient to destroy all life on Earth and create faster computerized simulations of happy thoughts,” says the scientists.
Humans cannot control advanced AI. Period.
Scientists explored how humans could possibly contain artificial intelligence. They found that there is no algorithm to make that possible.
One way to control a superintelligent AI is to isolate it from everything so it would not learn. The problem with this solution is it defeats the purpose of creating AI.
Another suggestion is to design a “containment algorithm" (think about the Laws of Robotics) to ensure that artificial intelligence will not harm humans. With the current language of computing, it is impossible to create such an algorithm, according to the scientists.
In other words, no program or algorithm is good enough to determine whether artificial intelligence will kill off humans.