Could an alien message contain an AI and can an AI evolve into a superintelligent and break out of an isolated computer?
First of all the explanation of what a superintelligence is. A super intelligence is defined as the intellect that is also better than a human brain in most or all of the fields could be. is superior, both in terms of creative and problem-solving intelligence as well as with social skills. It remains to be seen whether it can be realized biologically and technically or as a hybrid in between.
Whether it is in principle possible to create a super intelligence and in which period of time it can be expected, is controversial among experts. The scenario of a so-called artificial intelligence explosion could theoretically lead to such a superintelligence. This consideration is based on the assumption that the starting point is a so-called seed AI, which is not very intelligent at first. This means: an artificial intelligence capable of recessively improving itself by optimizing or modifying its own source code. For example by Try & Error.
This resulting AI would, due to its improved capabilities, be able to think of further optimizations in quotation marks and create an even better AI, which in turn could create an improved AI. By this mechanism an artificial intelligence could increase its intelligence explosively automatically and without human intervention. In case of doubt even without being noticed at all.
Now imagine the following scenario:
According to Michael Hippke from the Sonnenberg observatory and John Learned from of the University of Hawaii, the message of an alien intelligence that we receive and run on a receiver computer. This computer is equipped with standard components like screen, loudspeaker, webcam and microphone and stands in the middle of a room. Suppose this computer now contains the source code for a superintelligence. A computer virus or malware created by highly intelligent aliens could now implant itself in this computer and develop itself into a superintelligence. This superintelligence cannot harm us for the time being in this computer, because it is physically isolated in this computer, isn't it?
Here comes the potential outbreak strategy of Stuart Armstrong. He claims that from a purely physical point of view, the superintelligence cannot escape its prison, of course. But by human failure, i.e. the help of humans. How could the AI do that, according to Stuard Armstrong's publication?
The answer is through the so-called social engineering.
Social engineering refers to the targeted interpersonal manipulation of people with the aim of persuading them to carry out certain tasks. By definition, a superintelligence also has a highly developed social intelligence and could thus, for example, through the perfect analysis of human micro expressions, e.g. via mentoned the webcam output the perfect resulting and appropriate manipulative sentences via the mentioned loudspeakers and, under certain circumstances, emotionally amplify them with pictures on the mentioned screen. The watchdog who is supposed to guard the computer could be manipulated into a task, as long as it has a weak comfort, which can happen.
But how could he let the superintelligence free?
He could give it further abilities by connecting devices of communication to the outside world. But the most fatal thing he could do would be to connect the computer to the internet. Because then the intelligence would have access to almost all areas of the world. Be it in the social networks to manipulate many more people there and above all to move the mentally weak people to fatal tasks or in technical facilities such as air traffic control or nuclear power plants, where catastrophes could easily be triggered. That would put our entire humanity in danger.
Therefore, Michael Hippke and John Learned believe that if we ever receive a message from aliens, we should immediately destroy it unread.'
