Journal of
Italian
Banking
Association

year 100
 

Archive » September 2024 » Artificial intelligence: impartial genius or reflection of our prejudices? The problem of training biases

Artificial intelligence: impartial genius or reflection of our prejudices? The problem of training biases

Alessandro Carlini
September 2024 - n. 9
Keywords: Intelligenza artificiale, algoritmi, bias di addestramento
Jel codes: O33, M16

Computers seem increasingly capable of thinking like humans. But what happens if the human prejudices are also reflected? Artificial Intelligence is revolutionizing many sectors, from medicine to finance, with a strong impact on society. Behind the apparent objectivity of the algorithms, however, there is the insidious problem of training biases. An essential aspect to be known and addressed correctly, because their mitigation is a priority for ethical and effective Ai. Of all the models that are part of the recent evolution of Ai, the most interesting are based on Large Language Model architectures, capable of understanding and generating natural language, trained on huge amounts of data coming mainly from the web.

Interested in this paper?
Buy the issue