In an effort to avoid a dystopian sci-fi future where Artificial Intelligence knows pretty much everything about you, and a team of cops led by Tom Cruise run around arresting people for crimes they did not commit because of bad predictive analysis; Bernie Sanders and other Democratic candidates have some proposals on how we can stop it.
Recently Bernie Sanders unveiled a proposal that would outright ban the use of facial recognition for law enforcement purposes. It’s part of a bigger proposal for a massive overhaul of the criminal justice system. It also comes on the heels of news and studies that have shown how wrong facial recognition algorithms can be, and the kind of damage these programs can do if deployed for law enforcement purposes.
A Georgetown University study found that the NYPD detectives used some creative methods to manipulate their facial recognition system to try and capture a shoplifter. Creative in this case means that the detective thought that the suspect looked like Woody Harrelson from blurry in-store security footage, and then fed a photo of Woody Harrelson into the facial recognition system. An arrest was eventually made in the case.
Another report from The Verge revealed that New Orleans was using its own facial recognition, and predictive analysis system in conjunction with its city-wide network of police cameras.
But law enforcement isn’t the only place where AI can wreak havoc on people’s lives study conducted by UC Berkeley found that consumer lending algorithms discriminate against minorities. And the ACLU found that an experiment using Amazon’s facial recognition software produced some depressing results.
While Bernie Sanders’s proposal seems to go the furthest some of the other Democrats will not be outdone. Elizabeth Warren recently wrote a letter to financial institutions citing the UC Berkeley study.
Julian Castro announced that a part of his plan to turn the data tables around the police and use a database to track police wrongdoing. Corry Booker, along with three other Senators, introduced the Algorithmic Accountability Act, which would require companies to study and fix all the systemic issues that Artificial Intelligence systems have with things like facial recognition.
However, it’s also important to remember this isn’t just that machines are out to get us, it is first and foremost a people issue.
The systems in question did not invent or build themselves, and the statistical models that they use are completely reliant on human input. The NYPD facial recognition system requires a detective to provide an image if camera footage isn’t good enough, and if someone decides that someone looks like a movie star or a composite drawing then that increases the chances that a match could be incorrect without it being the machine’s fault.
The systems that financial lending institutions use require analysts and data scientists to identify the variables that determine the predictive outcome. And if the people working on the system make unconsciously biased decisions that profile certain borrowers then the outcome will understandably be one-sided.
Something else to consider is the flow of money to the development of these systems, one study found that by 2020 investments in AI-based platforms will reach $1.25 trillion dollars. With that much at stake it’s easy to image that more companies and governments will feel pressured to adopt lest they cause an economic downturn by refusing them.
Maybe we need a ban, and maybe we need to pressure out politicians to continue to be on the lookout for way to limit or stop not just the adoption of poorly built systems, but also to enforce way to make their use more a little bit less prone to human error and bias.
