Dear Bernie, Please, Don’t Blind My AI [an open letter]

by cauri jaye

Heavy regulation on facial recognition shows a fundamental lack of understanding of the technology and its potential. The solution to get permission from the person before being able to get their biometric data is like protecting your eyes from the sun by poking them out with a stick. More subtlety and understanding will lead to more ethical, legal, and innovative solutions.

Dear Bernie,

Let me tell you a story. Ellie and Sam see each other across the room. Ellie likes how Sam looks so she walks over, and says “Hi, I’m Ellie. What’s your name?” Sam considers for a moment, then says, “Sam”. After they talk for a while, they exchange Instagram handles. 

This common human interaction shows us something valuable. This is a three step process: 

  • step 1: they see each other; 
  • step 2: they connect to learn more; and then, 
  • step 3: they agree to share their information. 

Applying facial recognition gives us a choice to use just step one, steps one and two, or all three steps.

The problem with the current biometric collection laws in Illinois, and the National Biometric Information Privacy Act proposed by you and Sen Jeff Merkley, is that they conflate these steps into one. 

I’ve spent a lot of time and money on lawyers to make sure we comply with the current Illinois law that you’re using as a template for the Federal bill. So I’ve gained insight into the inefficiencies built into that law and the potential strangulation of innovation at a time when America needs so desperately to grow its world-class contributions.

In my company, we use facial recognition on video. We seek to understand people’s state of mind to improve human communication

Video is made up of many still images flicking by quickly enough to appear as movement. A machine looks at each image, or frame, separately. For the machine to know that it is looking at the same person from frame to frame, it has to remember that person’s face. This is equivalent to step one: Ellie notices Sam. There is no exchange of data. There is no memorisation of facial details. There is no exchange of personal information. 

This recognition does not infringe on the rights of anyone as long as they have agreed to be seen, for example, by going to a public place like Sam or going on a video call with a camera turned on. Therefore there is no need to regulate it. Step two and three, however, may need attention. 

Here’s another scenario: Tom climbs into a tree with binoculars to look into Bert’s apartment from across the street. This is a clear violation of Bert’s privacy. If Tom wants to see inside Bert’s apartment he has to ask for permission.

We can regulate the appropriate use of binoculars, making it inconvenient, illegal, or impractical to use binoculars in that way, but we do not blind Tom to achieve that end. We maintain that sight itself is a useful skill; we need instead to set clear boundaries on how it’s used.

We need to approach regulation of facial recognition in the same way. We can regulate what is appropriate usage and make it inconvenient, illegal or impractical to use it for nefarious purposes, but not regulate the ability to see.

Here’s one last scenario. Tom approaches a crosswalk in his car where Bert is about to cross the road. It would be insane to ask Tom to slow down with closed eyes and yell ‘if anyone is there, can I open my eyes to look at you?’ Tom does not need permission to see that he has to stay stopped so Bert can cross.  It would be impractical in the same way  to impose this type of step-one regulation on AI.

Many startups may not be able to surmount the legal and financial barriers you are putting in place as they struggle to solve the world’s problems on a budget. When drafting these laws, please, spend more time talking with the majority of us doing ethical work with AI to more deeply separate the steps and intent, and thus create a more nuanced approach to the laws. 

The majority of AI innovation aims to create tools with positive impact, and contributes to the US' leadership in global technology development. I implore you to not stifle this essential endeavor with a blunt weapon, wielded on all to punish the malintent of the few. 

Let’s not punish the 99% of us for the misdeeds of the 1%. 

- cauri

Tags: Ethical AI, facial recognition

Group 63