Welcome to I Was There when, an oral history project by the In Machines We Trust podcast. This podcast features stories from people who were there when breakthroughs in artificial intelligence or computing occurred. We meet Joseph Atick, who created the first commercially viable facial recognition system.
Credits
With Lindsay Muscato's assistance, Jennifer Strong, Anthony Green, and Emma Cillekens produced this episode. It was edited by Mat Honan and Michael Reilly. It was mixed by Garret lang, with sound design by Jacob Gorski and music by Jacob Gorski.
Full transcript:
[TR ID]
Jennifer: Im Jennifer Strong, hostess of In Machines We Trust.
I'd like to share with you something we have been doing for a while.
It's called "I Was There When."
This oral history project tells the stories of people who saw the breakthroughs in artificial intelligence, computing.
Joseph Atick: As I entered the room, it noticed my face and pulled it from the background. It said: Joseph, and that was when the hair on my back felt like something had occurred. We were witnesses.
Jennifer: Today we are kicking off things with a man who created the first commercially viable facial recognition system back in the 90s
[IMWT ID]
I am Joseph Atick. Today, I am the executive chairman at ID for Africa. This humanitarian organization focuses on giving Africans a digital identity that allows them to access services and exercise their rights. However, I was not always in the humanitarian field. My PhD in mathematics was followed by some major breakthroughs that led to commercially viable face recognition. People refer to me as the founder father of biometrics and face recognition. While I was at Princeton's Institute for Advanced Study, the algorithm for how a human brain recognize familiar faces became apparent. It was not easy to imagine how such a thing would be implemented.
It was a long time of programming, failures, and programming again. One night, in the early hours of the morning, we had just finished a version. To get a runcode, we submitted the source code. We walked out and I went to the bathroom. The source code was compiled by the machine, and it returned when I went back to the room. It runs the source code automatically after it is compiled. As I entered the room, the machine spotted me and extracted my background hair. We were witnesses. I called on all the people still in the laboratory and they each came into the room.
It would then say, "I see Norman." It would say, "I see Paul." And it would then go on to say, "I see Joseph." We would take turns running around the room in order to see how many people we could spot. It was, it was, a moment when I would say that many years of hard work finally brought about a breakthrough. Even though there was no additional breakthrough theoretically. It was very satisfying and rewarding to finally see the capability in action after we had figured out how it could be implemented. Our team was more of a design team than a research team. It was focused on putting all those capabilities into a computer platform. That was the birth, or rather, the birth of commercial facial recognition. I would place it on 1994.
My concern started very quickly. With the proliferation of cameras and the commoditization and increasing processing power of computers, I saw a future in which there was no hiding place. So I lobbyed the industry in 1998 and said that we needed to establish principles for responsible usage. For a while I felt happy because I felt that we had it right. I believe we have a responsible code of conduct that should be adhered to by all who implement it. But, this code didn't stand the test of time. We didn't anticipate the rise of social media. We believed that the most important component of a face recognition system was the database of known persons. We stated that if I am not in the database, then the system would be blind.
It was hard to create the database. We could only build a thousand, 10,000, or 15,000 databases because each image had had to be scanned by hand. The world we live in today is one where the beast has been allowed to escape the bag by feeding it billions upon billions of faces, and by helping it tag itself. It is now difficult to control and require everyone to use face recognition responsibly. There are a lot of faces online that you can easily scrape. This has been the case recently with some companies. So I panicked in 2011 and wrote an opinion article stating that it was time to panic because face recognition is becoming omnipresent in the world and will be readily available in databases.
People used to say that I was an alarmist at the time, but they now realize that this is exactly what is happening right now. So where are we going from here? I have been lobbying for legislation. I lobby for legislation that makes it a criminal offense to use someone's face without their consent. It's no longer technological. This powerful technology cannot be contained by technological means. Legal frameworks are essential. Technology cannot be allowed to move too fast. We must be aware of our values and what is acceptable.
Consent is still a difficult and complex issue when it comes to technology. Just giving someone notice doesn't mean it's sufficient. Consent must be given. They must understand what consent means. It is not enough to just say "Well, we signed up" and that's it. We informed people and they could have gone anyplace they wanted if they didn't want to.
It is easy to be seduced by shiny technological features that may give us a short-term benefit in our lives. Then, down the road, we realize that we have given up something too valuable. We have de-sensitized the population to the point that we can't go back. This is what I am concerned about. Concerned about face recognition via Apple and Facebook, I am concerned. All of this is legitimate, but I don't believe it all. Many of it is legal.
The public has reached a point in which they may be blasé and desensitized by seeing it all around them. Maybe you'll be able to leave your home in 20 years. It will be impossible to believe that you won't be recognized. It won't be recognized by the dozens of people that you meet along the way. The public will become very alarmed at this point because the media will begin reporting on cases of stalking. People were targeted. Some were selected based upon their street net worth and then kidnapped. That's a lot to take on.
So I believe the consent question will continue to plague the industry. It's possible that it will not be resolved until this question is answered. We need to limit what we can do with this technology.
Face recognition as we know it today was invented in 1994. This is what my career has also taught me. Most people believe that face recognition was created by Facebook and machine learning algorithms. These machines are rapidly expanding all over the globe. I had to resign as a public chief executive officer because I was restricting the use of technology my company was going promote for fear of adverse consequences for humanity. Scientists need to be able to see the potential consequences of their work and project into the future. They shouldn't stop making breakthroughs. You should not stop making breakthroughs. But, we must be honest with ourselves, and alert the rest of the world and policymakers about the positives and negatives. We need to ensure that this technology is used in a positive way.
Jennifer: I Was There When... is an oral story project that tells the stories of those who witnessed or made breakthroughs in artificial intelligence.
Do you have a story? Do you know someone who has a story to tell? Drop us an email at podcasts@technologyreview.com.
[MIDROLL]
[CREDITS]
Jennifer: This episode was recorded in New York City, December 2020. It was produced by Emma Cillekens and Anthony Green. These episodes were edited by Mat Honan and Michael Reilly. Garret Lang is our mix engineer with sound design by Jacob Gorski.
Thank you for your time, Jennifer Strong.
[TR ID]
