A "FaceApp" to enlarge breasts? It's here, and there's more to it than you may think.
The whole world has been taken by storm by “FaceApp”: a new app that allows you to edit a picture of yourself and see how you’d look like older, younger, with different hair style and so on. As soon as I saw the app, I recognized the state-of-the-art Artificial Intelligence algorithms that are behind it. Since I’ve been following these algorithms for a while, I wondered:
We have an incredibly powerful technology available and all we can do with it is internet memes?
I thought about it and I remembered a conversation I had with my good friend Patrick in the morning over a bowl of oats, discussing some new trends in Neural Networks (not joking: we are that nerdy). I know Patrick not only for being a Computer Vision expert, but also for having an eye for social issues, so I knew there was something more when he casually pitched me an old side project he built: an AI algorithm that takes a picture of a woman and simulates the results of breasts augmentation.
Cosmetic surgery is a delicate topic that needs to be treated with respect and delicacy. It’s hard to relate to the physical and emotional challenges that a person goes through. Since in this case probably neither me nor Patrick could, he briefed me on the research he had done before committing to the project, and later spoke to women to better understand their struggles and answer our question: can AI help them?
This article shows you what we found.
Teaching AI to simulate breasts augmentation
Before even approaching the problem, let’s see if the technology works. You can judge for yourself (nipples have been blurred):
How does this work? Patrick found a dataset of a few thousand images of women before and after they had breast enlargement surgery. Then, he built an AI algorithm that “learnt” how breast augmentation surgery modifies the body of a woman. Once it has learnt that, it can modify an image of any woman, simulating the results of the operation in a realistic way. He used a technology called Generative Adversarial Networks (GANs), a class of AI algorithms that has seen some amazing developments recently and it’s most-likely used by the guys of FaceApp as well.
So what do we do with this? Is it just a tech stunt or a shameful objectification of the female body? Well, the intentions couldn’t be further from this. The aim of this technology is to support women in making decisions on their body.
Right now, if a woman decides to go through surgery and (rightfully) wants to know how she’ll look like after the operation, she has three options:
* Look at other women’s results
* Use some other kind of software, like photoshop or specific tools
* Use 3D scans of her breasts at a specialized clinic, and use a digital simulation
Judge for yourself: this is one of the alternatives that is available right now:
Does it seem any human? 3D renderings, head chopped off, “Choose a similar body…”. It doesn’t seem like a reassuring solution, considering how challenging it must be to take a decision as profound as modifying your own body in a surgery room.
I thought that as a man it wasn’t my right to judge, so I talked to a good friend who went through breast augmentation surgery some years ago, and asked her about the challenges she faced. She doesn’t want her name to be public, so I’ll just call her Maria.
Maria told me about her first trouble: accepting that she wasn’t happy with her own body, and starting to reach out for help. An intimate decision on her own body had to become a cold discussion with a (likely male) doctor wearing a white doctors’s coat in his cold studio. All this emotional effort to have a simple answer to a simple question:
How will I look like if I do this?
Maria spent weeks looking at herself in a mirror fantastising on the answer to this question before she found the courage to walk through the doctor’s door. And after she did that, the answer she received was not at all emotionally reassuring. She had to go through images of the results of the operation on other women, something that not only didn’t really help Maria, but also felt degrading. This was about her, not about “patient number 38”. There are surely more empathic doctors around, but Maria wasn’t lucky. It’s hard to teach empathy in med school.
Maria had to take a leap of faith. She told me about the stress she felt the morning of the operation, knowing that her body would soon be altered to chase an image of herself that she didn’t really know yet, but hopefully would make her happier and more comfortable with her body.
Five years later, Maria is happy with her body and the stress of the operation is a distant memory.
How can AI help?
Patrick’s AI could have helped Maria. Before asking for feedback from a medical professional, she could have foreseen the effects of the operation in the intimacy of her room, using nothing but an app. Her pressing question “how will I look like?” would have gotten a first answer in her personal space, not in a doctor’s studio through a book of other people’s intimate snapshots.
When we talk about AI and its shortcomings, we often list lack of empathy and “human touch” among them. Especially when dealing with delicate topics like health, we tend to imagine that people want to interact with other people, not with algorithms. I think we need to revisit our views on this matter: people don’t always want to talk to people. There’s a set of problems that are emotionally hard to deal with and tough to share. In this case, the “coldness” and apathy of technology can paradoxically help people taking a first step.
Removing humans from the equation doesn’t just remove empathy, it can also get rid of the fear of being judged and not understood.
There’s also another point: it simply works pretty damn good. Since this AI has “learnt” from real world evidence, its results will be more realistic than any self-service product. If evolved to work with the doctors’ input, it could be a terrific asset to help physicians make patients feel more at ease with their choices, and ultimately happier.
This technology could help many more people other than Maria. Think about men that need to go through hair transplant, man and woman that want a nose job, or in general everyone that wakes up in the morning not feeling at ease with her or his look, wonders on how surgery can help, and has a simple, straightforward question: “how will I look like if I do this?”. AI can help with this.
AI is nothing but a tool in the hands of people. It’s our job to envision how we can use it to serve humanity solving their problems and making their life easier, and not be limited by short-term internet memes.