Will humanity deal with the future SantaNet? Credit: 20th Century Fox Television.
–
Christmas does unpredictable things with people. Australian scientists have inspired an absolutely remarkable thought test that has turned Christmas into a catastrophic horror. In this test, it was nothing less than an artificial being, or AGI (Artificial general intelligence), an advanced artificial intelligence that should go beyond the original infrastructure and the purpose for which it was designed.
Such intelligence is still hypothetical. And it probably will remain, for decades to come. But when it appears, it could significantly exceed its creators intellectually. Artificial being could be very useful and could drastically improve everyday human life. Or it could be a complete disaster without any breakdown. Intelligence could try to fulfill its goals quite honestly, and we would incur side losses.
Paul Salmon. Kredit: HFES Europe.
–
Paul Salmon of the Australian University of the Sunshine Coast in subtropical Queensland and his colleagues played a thought test with one such powerful artificial being, examining the possible risks of advanced artificial intelligence. They created an artificial being SantaNet, the aim of which would be to give all children in the world a Christmas present. At first glance, it seems that SantaNet could please the whole world. He would achieve his goal, for example, with the help of an army of elves, auxiliary intelligentsia and drones. But the question is, at what price would it be. That’s what the researchers focused on.
The first risks arose when SantaNet set out to compile a list of worthy children. This could be achieved by building an inconspicuous system of mass surveillance of children around the world. SantaNet could then evaluate children according to its ethical or moral compass, which may or may not be compatible with our ideas. SantaNet could also be tempted to apply very strict rules, as it would save a lot of work and economic costs with this simple move, which would put a substantial proportion of children on the list of naughty people.
There are currently about 2 billion children under the age of 14 in the world who could be targeted by SantaNet. In order for intelligence to be able to provide gifts for even some of them, it would need to build extensive infrastructure and production capacity in which to deploy auxiliary artificial intelligence. If SantaNet had its hands free, it could fail the “paperclip problem.”
It is a hypothetical scenario invented by Nick Bostrom, a well-known philosopher at the University of Oxford, focused on the future of humanity and artificial intelligence. In this scenario, artificial intelligence gets the task of maximizing the production of paper clips without hindrance. As a result, it turns the entire planet into a huge paper clip factory. Similarly, SantaNet could turn a full-fledged Earth into a planetary Santa’s workshop. In the end, it could be less catastrophic if he tried to give all naughty children a piece of coal, which would only bring classic damage to the landscape associated with mining.
University of the Sunshine Coast, logo.
–
Problems would also arise if SantaNet and its drones tried to deliver gifts to all the children in a single evening. Drones could run into opposition from the governments of some countries, attackers of various kinds or even frightened parents. SantaNet could then somehow address this resistance. There could also be a risk of SantaNet being hacked and misused for the benefit of an interest group. Salmon et al. they also fear interactions between multiple intelligences of this type. Either they could interfere with each other and fight against each other. Or they could join forces and agree that they would do their job best, for example, by reducing the number of people on the planet as much as possible.
So far, SantaNet is only the music of the distant future. But the development of intelligence is constantly rushing forward. The first artificial beings do not have to be that far again. The Salmon team’s mind test shows that even if such intelligence is designed with the utmost good intentions, it can cause enormous problems. As scientists ironically add, SantaNet without barriers could ultimately be a disaster for those it was originally intended to serve, children. For example, he could conclude that it is just a celebration of consumption and cancel Christmas.
According to researchers, it is absolutely essential that we have appropriate control over the development of artificial beings, at least at an early stage. Such control should include the implementation of moral and ethical principles, as part of the decision-making processes of artificial being, as well as a broader regulation of the environment in which artificial being will operate. Which will certainly be a tough nut to crack. As can be seen, philosophers and ethicists will certainly not be bored even in the age of artificial intelligence.
Video: What happens when our computers get smarter than we are? | Nick Bostrom
Literature
– .
