WILMINGTON – Facial recognition technology is a powerful, and it’s still a relatively new technology, especially with the integration and use of artificial intelligence.
That’s according to Karl Ricanek, the founder and CEO of Wilmington-based Lapetus Solutions, Inc., who spoke with WRAL TechWire in a Zoom interview earlier this year about the current state of facial recognition technology.
IBM recently faced backlash, then decided to halt a facial recognition plan. In February, Texas filed a lawsuit against Facebook’s parent company, Meta, for collecting facial recognition data. And last month, a U.S. House of Representatives committee launched an investigation into the use of face recognition technology by the federal government.
According to Ricanek, we’re still learning how the technology is actually operating, because we still don’t quite understand how the use of artificial intelligence deployed.
“We don’t fundamentally understand how that black box, AI, does what it does,” said Ricanek. “We’re exploring that through this whole realm of explainable AI, and that’s going on today, but today, we’re also using the technology.”
The use of AI-enabled facial recognition technology is being used, including in what Ricanek called “critical infrastructure,” such as law enforcement.
It’s important, though, Ricanek noted, that despite the incredible capacity that facial recognition technology has, humans are the ones who are making decisions.
“How do we use this as an investigative tool, and not to identify a culprit,” said Ricanek. “Keep that human in the loop, and provide that human with procedures and processes that take it through the technology, and embed that technology in the investigation.”
“The promise of the technology,” said Ricanek, “outweighs some of the issues that we’re facing today.”
The core of the black box
“At the core of it, AI, this black box, is formulated sort of like the human brain,” said Ricanek. “These deep learning neural networks are the crux, the core of face recognition technology.”
They’ll take in images, break down the images into individual units, which in images are called pixels, and then they process the pixels.
“Because of the complexity of the number of parameters that we’re talking about, millions, or tens of millions, in some cases hundreds of millions of parameters that go into that, we don’t have the capacity to really fundamentally understand how these things are being linked to create this pattern extraction that creates the features that create the template,” said Ricanek.
“We don’t know when that thing starts to get close to an edge, and falls off, basically poor performance,” “That is, matching me to someone who is not me, for example.”
We don’t understand where that boundary is, right now. “And that’s why it’s critically important we have humans in the loop,” said Ricanek.
Challenges with facial recognition
We created facial recognition technology, and right now, we tend to treat all of this technology the same. But face recognition isn’t a homogeneous tool, said Ricanek. Some companies provide products that are really quite good, whereas others may offer products that are known to have bias.
“The problem that we have, is that when law enforcement agencies purchase this technology, they don’t know how to effectively evaluate to understand things like bias,” said Ricanek. “Every community is different.”
There are also privacy concerns, said Ricanek, which is something of a “slippery slope.” Even on a Zoom call, like the one Ricanek joined to speak with WRAL TechWire, that could be considered biometric data. In fact, said Ricanek, a single still picture of an individual could be considered biometric data.
“That’s very challenging in today’s society, where pictures are prevalent everywhere,” said Ricanek. “We get into the possibility of being able to identify and track individuals across their daily life.”
We’re already being tracked, though, said Ricanek, and that tracking doesn’t necessarily occur using biometric data. “We have to as a society figure out who can track us and who can’t track us, and what that looks like,” said Ricanek.