Superintelligent systems are already being created and disseminated across the globe due to the technological race. For this reason, some questions arise: is it possible control artificial intelligence how far? How ready are we to fight a AI that works beyond human capacity? How ethical is it to create a program whose consequences and impacts cannot yet be measured? See now more information about these questions asked by scientists.
Read more: 7 applications of artificial intelligence: some already replace human labor
see more
Alert: THIS poisonous plant landed a young man in the hospital
Google develops AI tool to help journalists in…
Artificial intelligence with superintelligence
It sounds like a science fiction movie plot, but the truth is that an artificial intelligence can indeed create autonomy to learn commands that can later turn against humans. It is not guaranteed, for example, that they will follow rules such as “do not cause harm to humanity”, since that they will be programmed to have autonomy and can circumvent the limits imposed by programmers.
Alan Turing halting problem
Called the “halting problem”, this problem gives two alternatives for a command given to a computer: either it will come to a conclusion and either respond to it by stopping, or it will continue in a loop trying to find a solution – until find her. For this reason, it is impossible to know whether the artificial intelligence will stop, or whether it will find and store all possible conclusions in its memory.
The problem with this is that it is not possible to measure, after a while, which solution or alternative the machine will take for a given problem. This way, its behavior will be unpredictable and it will be pointless to create a containment algorithm.
Divergences between opinions of scientists
Some scientists argue that all advances in technology must be designed to promote the common good in a safe and controlled way. In this way, a supermachine capable of storing any and all types of intelligence, with inefficient containment powers and a still unmeasured capacity for destruction, seems like something dangerous.
With that in mind, computing researchers claim that superintelligent machines may have harms that outweigh the benefits, and their construction should be avoided. Many programmers constantly deal with the fact that their machines have learned to do things they never could. taught or created commands for them to learn - and this, depending on what they learn, can symbolize a risk.
ethical issues
Limiting superintelligence capabilities may be a viable alternative. To do this, it is possible to cut it off from parts of the internet or from certain networks. These would be examples of containment mechanisms. However, it is not yet known whether these alternatives are viable – or whether the machine would have access to these contents in a way that has yet to be revealed.
Therefore, if humanity continues to advance with artificial intelligence, it is necessary to be very careful with the extent to which it must be autonomous.