[ad_1]
According to new discoveries published by a research center of the University of New York, the lack of diversity in the field of artificial intelligence has reached "a moment of counting". A "diversity disaster" has contributed to faulty systems that perpetuate sexist and racial prejudices, according to the AI Now Institute's survey of more than 150 studies and reports.
The body of AI, which is composed of a majority of both men and whites, risks reproducing or perpetuating historical prejudices and imbalances of power, the report said. Examples cited include creating image recognition services offensive rankings minorities, chatbots adopting hate speech and Amazon technology does not recognize users with darker skin. The biases of the systems built by the AI sector can be largely attributed to the lack of diversity on the ground itself, the report said.
"The industry must recognize the seriousness of the situation and admit that its existing methods have not solved these problems," said Kate Crawford, author of the report. "The use of AI systems for the classification, detection and prediction of race and sex requires urgent reassessment."
More than 80% of AI professors are men, and only 15% of researchers in Facebook and 10% of researchers in Google are women, the report says. The composition of the field of IA reflects "a broader problem related to computer science, to the areas of the stem and, more broadly, to society as a whole," said Danaë Metaxa, PhD candidate and researcher at Stanford, who has focused on the problems of the internet and democracy. Women accounted for only 24% of the field of computer and information sciences in 2015, according to the National Science Board. Only 2.5% of Google's workforce is black, while Facebook and Microsoft each account for 4%, and there is little data on transgender workers or other gender minorities in the field of AI. .
"The urgency behind this issue is increasing as the AI becomes more and more integrated into society," Metaxa said. "The lack of diversity in AI concentrates an ever greater amount of power and capital in the hands of a selected subset of people."
The venture capital financing of AI startups reached record levels in 2018, up 72% from 2017 to reach $ 9.33 billion. In the United States, active AI startups grew by 113% between 2015 and 2018. As more money and resources are invested in AI, companies have the opportunity to make in the face of the crisis, said Tess Posner, president and CEO of AI4ALL, a for-profit organization that works to increase diversity in the field of AI. This lack of diversity needs to be resolved before AI reaches a "critical point", she said.
"Every day that passes, it is more and more difficult to solve the problem," she said. "We are currently in an exciting moment where we can make a difference before seeing how complicated it can become later."
The report released on Tuesday warned of the need to adapt to diversity in the technology sector by solving the "pipeline" problem alone, or the composition of those hired. Men currently represent 71% of the pool of applicants for jobs in the field of AI in the United States, according to the 2018 AI Index, an independent report on the sector published annually. The Amnesty International Institute has suggested additional measures, including public disclosure of workers' compensation levels, dissemination of transparency reports on harassment and discrimination, and changes in hiring practices. to increase the number of under-represented groups at all levels.
Google has dissolved an artificial intelligence ethics board intended to oversee these issues just one week after it was announced in March. The External Advisory Council on Advanced Technology (ATEAC) was attracted by negative reactions inside and outside of society after appointing LGBT advocate Kay Coles James.
Mr Posner pointed out that additional efforts to increase transparency on how algorithms are built and how they work might be needed to solve diversity problems in AI. This month, US Senators Cory Booker and Ron Wyden presented the Algorithmic Accountability Act, a bill requiring algorithms used by companies generating more than $ 50 million a year or holding information on at least one million dollars. # 39; s users.
"The heart of the problem is whether market forces will be sufficient to solve this problem," Posner said. "It will take effort at all stages of AI and make cultural and procedural changes to solve this problem."
[ad_2]
Source link