Partnership on AI, a non-profit group whose members include large U.S. technology companies as well as universities and researchers working on artificial intelligence, has appealed to governments to ease visa requirements for A.I. experts.
The group said in a report released Tuesday that current immigration laws were hampering A.I. researchers, especially those from the Middle East, Eastern Europe, and Africa, from attending conferences and working in places, such as North America and Western Europe, where the emergent technology was being pioneered.
Existing restrictive visa rules could make it more difficult to develop safe and ethical machine learning algorithms, the group said. "To make sure those systems are safe for people we need the most diverse set of voices in the room," Lisa Dyer, Partnership on AI's director of policy, said in an interview.
Brian Green, a professor at Santa Clara University in California who specializes in technology ethics, said he supported Partnership on AI's stance. "It is tremendously important to have international scholars be able to meet in person to discuss issues in technology ethics, especially A.I., which is transforming the world so rapidly," he said in a statement.
In a number of instances, algorithms created from biased datasets have been deployed with potentially harmful consequences. That includes algorithms for making bail and parole decisions in the U.S. that classified black prisoners as being at higher risk for re-offending than white ones with identical or worse criminal records.
Such problems are more likely to be spotted and corrected if the people creating those algorithms are themselves from a diverse range of backgrounds, Dyer said.
A.I. researchers from Africa in particular have struggled to get visas to attend events in Western countries. More than half of the African invitees to the Black in AI Workshop at the Neurological Information Processing Systems (NeurIPS) Conference—considered one of the most important A.I. research conferences—were unable to attend last year's session in Montreal because their visa applications were delayed or denied.
Partnership on AI, whose members include companies such as Facebook, Google, Microsoft and IBM, said it would advocate for governments to move towards a system in which visa applicants are evaluated more for their skills than on their nationality. It also asked for governments to lift caps on the number of visas they issue to technically-skilled individuals and to create separate visa categories for experts in artificial intelligence and machine learning.
Dyer acknowledged that the group's recommendations may encounter political resistance at a time when many countries, including the U.S. and many European nations, have increased immigration restrictions, not eased them.
But she said that many of these same governments have also said they see accelerating the development of A.I. as a national priority. "Governments are saying they want to attract and retain A.I. talent and see the social and economic benefits of that," she said. "Clamping down on visas seems not to support those policies."
The U.S. has also started viewing researchers from China with increasing suspicion as geopolitical tensions between the two nations have mounted and both countries have come to see A.I. as a strategic technology. But Dyer said that even during the height of the Cold War, scientific exchanges between U.S. and Soviet scientists didn't stop. She also noted that even after September 11, when many wanted to close U.S. borders to foreigners, the Reagan-era National Security Decision Directive 189 was reaffirmed by then-National Security Advisor Condoleezza Rice. That directive says a free exchange of scientific ideas is vital to maintaining U.S. technological preeminence.
While, as a non-profit entity, Partnership on AI is not allowed to lobby the U.S. government directly, Dyer said that she hoped the white paper would help spark a conversation among lawmakers and officials. Many of the Partnership's members are also able to engage in lobbying and they may take up the cause.
Dyer also said that current immigration laws tend to unfairly favor large multinational corporations, which can afford expensive legal assistance to help employees and new hires navigate complex visa regimes. Meanwhile, startups and smaller businesses needed less onerous rules to help them compete for global A.I. talent, she said.