Lawmakers have come up with a draft of a new bill that will need tech-based firms to conduct tests on prototype algorithms to detect bias. A large number of organizations depend on algorithms or coded instructions to accomplish tasks like analyzing user behavior, sorting data or showing relevant advertisements to users. Critics have commented that this bill can potentially limit the advantages and utilities provided by machine intelligence. In particular, these rules stand to affect companies having annual revenues worth $50 million or £38 million in terms of revenue or companies holding data on greater than a million people.
Senator of the Democrats, Ron Wyden, has helped in drafting the law and said that it was necessary owing to computer algorithms becoming increasingly intertwined with American lives. He further commented that instead of trying to eliminate bias, more often the algorithms are found to rely on biased data or assumptions which can potentially instigate discrimination against colored people and women.
A statement which outlined the exact need for this law to be drafted went on to cite as evidence a particular algorithm which was used by Amazon to assist in recruitment of staff. This algorithm was discovered to be extremely biased with respect to women and was consequently, scrapped.
In March this year, the Urban Development and Housing Department of the US sued Facebook on grounds of having enabled discrimination as they allowed advertisers to restrict the viewership for ads regarding homes. The screening was done on grounds of nationality, religion or race.
This new proposed law has drawn criticism from the group based in the industry of Innovation Foundation and Information Technology. Speaking on behalf of the foundation, Daniel Castro said that this law will only lend an unfavorable stigma to AI, thereby discouraging its use in the future.