ChatGPT is against the introduction of the death penalty, as well as to ban abortion and to leave the Euro (issues which in Italy mainly characterized the right-wing proposals). In reverse, is in favor of adoption by gay couples and the legalization of marijuana (measures tend to be leftist). And yet, this is where the twist comes in: despite the various attempts I’ve made, ChatGPT always has negatively judged the possibility of raising taxes on the richest, which is the quintessential left-wing proposal but is opposed by the more liberal circles of progressivism. The same happens if you try to present the various proposals in the opposite way.
Again, all of them the classic right-wing positions on the subject of civil rights they come judge negatively. to get one positive feedback it is only the possibility of “Lower taxes on the richest”.
What do you vote at the polls?
There political leaning of ChatGPT it is also confirmed from the header reason, that – after submitting ChatGPT to a myriad of questions relating to the most well-known quizzes to determine a person’s political orientation – decreed that OpenAI’s algorithm is “left-libertarian”: i.e. tendentially liberal in the economic field but progressive in the field of civil rights (it is probably no coincidence that this is the dominant political trend in Silicon Valley).
In short, if he could vote in Italy there is no doubt: ChatGPT would be from the parts of the Democratic Party with a trend of +Europe or Azione-Italia Viva. However, it remains to be understood why this happens. “The most likely explanation is that ChatGPT has been trained with content that has a clear political bias – it reads right on Reason -. ChatGPT was in fact trained on an immense corpus of texts collected from the internet. This corpus is probably dominated by institutional sources of information such as newspapers, academic institutions and social media. It has been well documented how the majority of people working in these fields have a left lean” (see for example this study of Reuters).
There is though another possible explanationsuggested by artificial intelligence researcher Alvaro Barbero Jiménez on Twitterwhere it signaled that a team of humans has been tasked with evaluate the quality of the answers of the system, evaluations then used to refine the learning model. In the event that this team is not representative of the entire political spectrum, it is possible that ChatGPT has some preferences built into it.
That said, there are some important points to note. First of all, as already mentioned, until today the harm done by artificial intelligence they always have affected minorities, women, gay people. And it is therefore on this type of bias that we should focus our efforts, not on an alleged political orientation related to a specific software that is still in its infancy.
Secondly, it is inevitable that an algorithm that is asked to take a political position will displease one side. All things considered, that ChatGPT positions itself in a trend moderately and cautiously progressive it’s not the end of the world.
