The potential ofartificial intelligence are growing exponentially and among the most recent uses there are those that concern the creation and theidentification of human faces.
However, we must not forget that these “projects” they are not copies of existing people or who knows what sophisticated creationist mechanism, but they exclusively represent a mix of different somatic characteristics, part of the immense knowledge heritage used to train AI.
And it is precisely by analyzing this knowledge that scientists noticed a first and dangerous inconsistency: artificial intelligence prefers to create faces individuals with Caucasian featureswhich are constructed in a more coherent and realistic way than those of any other ethnicity.
This, of course, in addition to ethical question, also opens a separate discussion on potential problems that arise from it, which could engage users in different ways.
Let’s find out more about it.
Artificial intelligence and problems of racism
The news recently hit the news an African American boy who ended up in prison due to a mistake caused by the Facial recognition of artificial intelligence.
The problem was reportedly caused by the AI’s difficulties in recognizing the physical features of the real suspect.
From the first reconstructions, the synthetic brain would have caused confusion by drawing on its own learning models, who were trained predominantly with examples that reference white individuals.
This mix-up only confirms the well-known racial prejudice which is worrying experts, especially because it would be leading AI to unconsciously have gods racist behaviors towards certain somatic features considered “atypical” according to his knowledge.
All things considered, scientists say, people are projecting onto artificial intelligence i prejudices already existing in the world and this could represent a problem for people belonging to specific ethnic groups, such as African-Americans, for example, or Orientals.
Given the enormous progress made in recent years and also given the increasingly widespread diffusion of this technology, a synthetic brain cannot be allowed to develop preconceptions of this kind and we should therefore take action while there is still time.
The dangers of AI in creating hyper-realistic faces
The other serious problem of faces created by artificial intelligence And hyperrealism which is identified as a phenomenon that leads people to believe they are faced with a real representation.
According to a recent study published in Psychological ScienceIn fact, these creations are perceived precisely as if they were real and they may be very difficult to recognize.
Here too, the discussion on machine learning models used to train AI. These synthetic intelligences have been able to draw on such a vast wealth of knowledge that they have managed to create extremely detailed faces to the point of becoming almost indistinguishable from existing ones.
Leaving aside for a moment ethics and racismit doesn’t matter if these fictional characters are all of Caucasian ethnicity, what matters is that they could be used for malicious purposes.
And this is one of the gravest dangers of having available such a powerful tool like a sophisticated artificial intelligence that could become an additional weapon in the hands of cyber criminals.
Being able to create a detailed and convincing face could be used on the web, for example, for scam people or to create deceptive campaigns which may have the sole purpose of threatening the data and wallets of unsuspecting users.
Therefore experts continually ask themselves questions about real potential of AI which despite the great benefits for all mankind, could also represent a threat which, if not trained and regulated properly, could turn against us.