The "Extreme Vetting Initiative" is a new idea that officially informs us that the Trump administration does not understand technology, or does not think exactly what it may cause or thinks and does not care.
The U.S. Immigration and Customs Enforcement (ICE) sent us last June a letter outlining an initiative to "create an AI application that can automate, aggregate, and streamline current assessment process ”of immigrants wishing to enter US territory.
According to the letter, ICE's current methodology does not provide sufficient "reliable, high-value information for further investigation" as it comes from "US immigration attorneys or federal courts".
Extreme Vetting Initiative: AI of Digital Discrimination
ICE managers would in shortly want somebody in the technology sector to create an engineering learning application that will collect information they can use to prosecute or deny entry to immigrants.
Extreme Vetting Initiative: In a nutshell, the service gives us the definition of a biased AI.
This app could probably have one network βαθιάς μάθησης ικανό να δημιουργεί συσχετισμούς μεταξύ διαφορετικών data. In order to train such a network, data from DHS (Department of Homeland Security) or ICE could be used. Both agencies have data from people targeted by every secret service (and beyond), regardless of who evaluated them and by what criteria.
A group of distinguished scientists and engineers 54 sent today a different letter to the Department of Homeland Security requesting to abandon the plan with AI for the examination of migrants. In their letter they state:
"Simply put, there are no computational methods that provide reliable or objective estimates of the characteristics that ICE seeks to measure.
As far as we know, there isn't machine that is capable of determining whether a person is likely to commit a crime, nor is there any AI that can determine a human's intentions through gathering data from social media.”
An article by ProPublica (Machine Bias), who was a finalist for Pulitzer, says:
"There is software used all over the country to predict future criminals. And it is biased against blacks.
So it would be reasonable to believe that creating a similar AI that can decide if a person can be allowed to enter the country is not much different from biased software that decides harsher sentences for blacks. "
Note that among the 56 scientists and researchers who signed the letter for DHS are a mix of academics and experts from Companies such as Google and Microsoft.
IBM said at The Hill:
We have clarified our values. We are faced with discrimination and we will not do any work to develop a register of Muslim Americans.
So after public placement, it's very likely that IBM is not going to help the US government.
But no tech company should allow one system which will lead to discrimination of people. The use of AI that can marginalize and violate civil rights is antithetical to the very idea of research and progress.