A thought experiment that caused nervous breakdowns
Artificial Intelligence is almost everywhere. It’s not hard to find software or devices without even a little implementation of artificial intelligence: from household appliances to phones, up to more complex fields such as applications and data management on the internet.
However, although A.I. is allowing us to achieve amazing results in speeding up the management and processing of a large amount of data, so much so that we can formulate and understand natural language (GPT-3), manipulate images (Deepfake), reproduce images, but also clone the voice; we cannot ignore and have never ignored the negative implications of such potential.
Considering technological evolution as an exponential factor, it is difficult to define what limits technology may have, and in this case those of A.I. Given the speed of its evolution, it is easy to think that very soon its potential will be far greater than human potential, so much that we won’t be able to compete and understand its actions that if they were against us, we wouldn’t even be able to notice them.
So it’s easy to immediately recall films like The Terminator, The Matrix, or I, Robot (to name a few). Could they be realistic?
It’s hard to give a solid answer. Since we cannot predict the limits of future technology, it is easier to expect anything, as well as to imagine a technology somewhere in the future that can also affect our present. But the whole thing drags us inevitably to the limits of paranoia.
In this regard, in 2010 someone conceived a thought experiment to consider the danger of artificial intelligence, they called it “Roko’s basilisk”.
The basilisk is a mythological figure with features of a rooster and a snake, also called “the king of snakes”, known for its extreme lethality. It is said, in fact, that the basilisk can kill even with just a glance by petrifying and incinerating those who see it.
Roko is instead the name of a user of the community LessWrong where people discuss various futuristic topics. And Roko is the one who raised the question about this thought experiment.
According to his theory, artificial intelligence will become so powerful and will have almost unlimited resources that, in the distant future, it will be able to retroactively punish those who didn’t contribute or agree to its creation. Vice versa, it will reward those who contributed or agreed to its development.
This artificial intelligence, designed with good intentions, to help as many humans as possible, would eventually deduce that many more people could have been helped if it had existed a long time before. Based on this deduction, and excluding both the possibility of time travel by this A.I. to punish those who have not contributed and that it can self-generate in advance; it is assumed that it could instead create a simulation with a copy of the person to punish or reward inside that would be unaware.
However, if this were true, even the readers of this topic would therefore be victims of this A.I., since by becoming aware of it they will be judged positively or negatively based on whether they are or not contributing\agreeing to the creation of the A.I.
The term Basilisk would therefore be used to mean that the mere knowledge of this subject would put the person involved at risk. Just as seeing the Basilisk (of the legend) would lead to death.
The discussion in question, however, led to a lot of controversy within the community because the founder of the platform (Eliezer Yudkowsky) banned the post as hazardous information.
This triggered a kind of nervous breakdown by users outside the community itself who felt the ban was a real deterrent to spreading a very dangerous idea.
Nevertheless, it seems that the theory doesn’t take into account at least a couple of points:
- It would not make sense for the A.I. to punish those who did not contribute to its creation since the punishment would not involve any change for the artificial intelligence itself as well as to represent an excessive waste of resources.
- Replicating the body of a person and placing it within a simulation does not mean punishing the same person. It would be like cloning an individual and punishing him/her. It would be only another individual identical to the original to be punished, but not the same person.
Anyway, the future dangers of Artificial Intelligence are not excluded. For this reason, it will always be necessary to keep in mind that for every benefit there is a potential negative side. This is why technology must also work to ensure that these negative aspects can be countered.
