Sacramento, Nov 15 A 1940 property deed in Santa Clara County, California, contained stark language: "No persons, not of the Caucasian Race, shall be allowed to occupy, except as servants of residents, said real property or any part thereof."
This type of discriminatory restriction, while unenforceable today, remained in thousands of property records in California.
Now, an Artificial Intelligence (AI) programme developed by Stanford University researchers is helping Santa Clara County efficiently remove such racist language from property records, as required by a California state law passed in 2021, the San Francisco Chronicle newspaper reported.
According to research from Stanford University's Regulation, Evaluation, and Governance Lab, the innovative system can process millions of documents in a single day, dramatically reducing the time and cost needed to identify and redact racially restrictive covenants -- clauses that historically barred people of specific races from buying or occupying homes.
"Our machine-learning pipeline has saved over 86,500 hours of manual human labour," the Stanford researchers reported in their study.
Santa Clara County faced the daunting task of reviewing over 24 million property records dating back to the 1850s. The researchers found that using traditional manual review methods would have taken about 160 years and cost over 22.4 million US dollars for a single person to complete.
The AI language model is able to detect approximately 7,500 racist covenants in 5.2 million deed records from 1907 to 1980, Xinhua news agency reported.
Los Angeles County recently hired a private firm for eight million dollars to complete similar work over seven years.
In contrast, Stanford's AI system can process 5.2 million pages of property documents for less than 300 dollars in computing costs.
The researchers are making their AI model available to assist hundreds of other jurisdictions in similar efforts to identify and remove discriminatory language from property records.
Disclaimer: This post has been auto-published from an agency feed without any modifications to the text and has not been reviewed by an editor