Sarah Barrington

How did you first become interested in digital forensics?

Two words: Hany Farid! While I spent my career working on developing machine learning methods for solving complex problems, seeing Hany give a guest lecture in my class at Berkeley showed me how methods I’d been using for years could be applied to the world of digital forensics. Hany showed us how this field has the power to combat misinformation, revolutionize the criminal justice system, and, in some cases, even save lives. There was no way I couldn’t sign up for more. 

What was your path to joining GetReal Labs? Did anything surprise you along the way?

I am fortunate to have worked with Hany as a PhD student since the inception of GetReal, so I have seen the company grow from a group of four to the incredible team it is today. It’s been an honor to be involved in multiple stages of the company’s development and growth, and I am constantly shocked by how much progress has been made and the speed of technical development and research. 

What are you currently working on, and why is it important?

Although I am interested in all forms of media, my current focus is on audio forensics and deepfake detection. Audio is a particularly fascinating sub-field as it has accelerated so much in the past 2-3 years. Initially, in our lab, we only focused on images and video but realized that audio deepfakes were quickly surpassing the capabilities of visual deepfakes. We’ve already seen large-scale harm caused by audio deepfakes, from the fake President Biden robocall in New Hampshire to the theft of $25 million from a multinational firm in Hong Kong.

How could your work change the way people interact with digital media?

I see the objectives of my work as twofold: Firstly, developing technical methods that can help to establish what is ‘real’ in a digital setting and educating the public on the pros and cons of generative AI media. I envisage that, in the next couple of years, the internet will become awash with AI-generated content, and technical methods cannot, and should not, try to detect every case. In fact, there are many situations where labeling content as ‘fake’ is not beneficial. For example, digital artistry or gaming., I want to help people focus on the question of ‘what is real’ or even ‘human’ and help them understand when it is important to ask such a question. 

What industries or real-world problems could your research help solve?

While much of our work at GetReal surrounds the question of digital identity, we also aim to tackle misinformation (particularly with news media) and support law enforcement. 

What’s your unique approach or philosophy toward research?

‘It’s a feature, not a bug.’ This is my favorite quote from Hany Farid. He first said it to me when we were evaluating a 3D modeling approach for measuring the height and weight of a person from a single photograph. The model would sometimes fail on certain photographs. I thought that I had a bug in my code, but Hany helped me to understand that the problem was showing us something far more interesting about our analysis—that 3D methods struggle with certain body poses. This taught me that, in research, you might be aiming to answer a specific question or hypothesis, but often, it is the interesting tangents that emerge along the way that, most likely, you never expected to provide the biggest insights. 

What excites you most about the future of your field? How do you see this field evolving in the next five or ten years?

I can’t wait to see how we, as digital citizens, find harmony with generative AI content. The ‘cat is out of the bag,’ so to speak, and we will never again have an internet that is free from content generated by AI. And why should we? It is incredibly creative, powerful, efficient, and often beautiful. However, norms will start to emerge around how human-generated content can co-exist with AI. We need to start thinking about how we can establish whether a human truly was involved in the creation process. How can we verify that the person talking on the other end of the phone is truly your family member? It is my hope that GetReal will form a huge part of this shift towards a harmonious AI-human future. 

What do you do outside of research? Any surprising hobbies or passions?

I love to learn new (and, unfortunately, often dangerous!) skills. I got my gliding pilot’s license in the UK (where I am from!) and have started flying again since moving to the U.S.. I love anything with engines—cars, motorsport, bikes, you name it. I broke my hand doing aerial gymnastics last year. I play the guitar and the piano absolutely terribly. I also do a lot of baking.

‍