KRANJ, Slovenia (EFE).— Without an ethical design from the beginning, artificial intelligence (AI) will deepen the gender gap and discrimination. For this reason, it is essential to train systems based on it with data in which women are correctly represented.
This was stated by two of the greatest experts in the field gathered by UNESCO in Slovenia, at the II Global Forum on the Ethics of AI, which yesterday focused on the challenges to which the international community believes that special attention must be paid. for responsible development of the tool, as is the case of gender inequalities and the environmental crisis.
“Technology does not spontaneously close gaps; if there are no clear policies and designs based on inclusive principles from the beginning, gender biases will increase,” said Mexican Elena Estavillo, director of the Center for the Society of the Future and expert on Unesco advisory council to guide its 194 States towards ethical development of AI.
The vision of women in the technological field has been “very relegated” due to a historical lack of representation and participation, which persists with a development as powerful as artificial intelligence, in which, according to UNESCO data, the percentage of women that figures in making important decisions (what materials are invested in, how they are designed, etc.) does not reach 20 percent.
Estavillo, who was director of the Federal Telecommunications Institute between 2013 and 2019, stressed that for AI to be an opportunity for equality, two key issues must be addressed: that women are well represented in decision-making to incorporate the perspective gender and that AI systems are trained with data without bias between genders.
In the opinion of the Mexican economist, neither of the two conditions currently exists: “AI systems discriminate once morest women because they learn from biased databases, which tend to reproduce inertia.”
She cited as an example how Amazon excluded women from engineering jobs because it used an AI-based selection system that rated them with a lower grade because that tool was based on data showing that historically men were chosen for those positions.
Estavillo warned that the gender perspective must be incorporated and AI controls must be tightened in the decision-making that gives access to jobs, scholarships, credits and everything that represents opportunities for women, because “historically this is where discrimination has occurred the most.” to women”.
The problem with these new developments is that “they start from existing gaps and are expanding them in this artificial intelligence ecosystem,” said Constanza Gómez, also Mexican, founder of AI for Climate and also a member of UNESCO’s advisory council on ethics for AI. .
To correct them “there is only an ethical design from the origin, without giving rise to comments on these biased and discriminatory decisions to act.”
Both experts, part of the organization Women for Ethics in AI (Women4Ethics of AI, by its name in English), agree on the importance of collaborating at the level of governments, academia, companies and citizen organizations to promote these algorithmic designs ethical, in such a way that AI “leads to closing gaps instead of amplifying them.”
And the fact is that “the symptoms are the gaps” but at the bottom of them “underlys the culture, the stereotypes, the lack of participation of women,” they highlighted.
However, Constanza Gómez knows well how AI can be a great ally to the planet’s challenges, in this case to provide better data on the state of biodiversity and be able to develop public conservation policies based on data.
The organization he founded equipped cameras and audio sensors that were used to train an AI system that recognized and quantified species, including jaguars, in various ecosystems of Yucatan, thanks to which ecological corridors can be established.
“Artificial intelligence can be a very strong ally for conservation or the fight once morest climate change, but we must make great efforts so that the systems based on it are transparent and there is timely accountability,” he stressed.
#automatic #ethical
2024-04-08 12:42:21
