Ai, algorithms need to be trained in equality

Only with an original training of thealgorithm devoted to equality will it be possible to overcome the bias and discrimination inherent in artificial intelligence. Professor Ruben Razzante, advisor to the Extraordinary Commission on Intolerance, Racism, Anti-Semitism, Incitement to Hatred and Violence, chaired by life senator Liliana Segre, is convinced of this. During the third day of Talk to the future, he recalls the amendment proposed to the Decree on Ai. In Article 3, it calls for an obligation for manufacturers to train their algorithms on equality at source, providing solutions that recognise violations at an early stage and prevent discriminatory user behaviour. "It sounds utopian, but I believe that the great powers of Ai can be governed by ethical urges that always protect international and constitutional principles and are respectful of social pluralism".

In the initiative organised by the Milan Bar Association, policing and predictive justice was also discussed, with use cases recounted by lawyer Giovanni Briola. "At the police headquarters in Milan and the police in Naples, algorithms are used to profile criminal suspects. Article 5 of the European Ai Act prohibits profiling tout-court, in fact what has to be done is a balancing act between two values: freedom and security. Where the use of Ai remains anthropocentric'. And the bias of artificial intelligence, with related discrimination by gender, ethnicity, disability, in education, biometric identification, employment and financial risk assessment.

But the issue of combating discrimination and the importance of not being driven by the machine alone also intersects with the knot of disinformation. Of which, especially in recent times, social and algorithms have become megaphones. With great risks for democracy and decision-making autonomy.

'For better or worse, Ai does not only have an individual dimension - affecting the rights of the individual - but an ultra-individual one. And in this sense it poses challenges to the rule of law and democracy, in Europe and beyond,' explained Oreste Pollicino, professor of constitutional law and artificial intelligence regulation at Bocconi University. "For this reason, the European value model must be protected online and offline, and one of the most important laboratories is the fight against disinformation. If, on the one hand, Ai amplifies the misperception of reality, with fake news and deep fakes, on the other hand, it allows us to quickly intercept manipulations'.

In this scenario, where Ai and information professionals have to work in tandem to counter the enormous amount of manipulated content generated with the aim of uncritically shifting public opinion, Europe actively attempts to set limits. But legislative regulation without industrial leadership and sanctions is not enough.