Artificial Intelligence made us rethink a lot of aspects of our lives and beauty is no exception. AI can help diagnose different skin concerns, your skin type, or even recommend you the right products for your skincare routine. But what is actually behind artificial intelligence, and how does it work?
A Beauty Advisor Available 24/7
If you are a skincare junkie you probably put a lot of time and effort into choosing your next product, looking at the ingredients, determining whether it will work for your skin type. It’s nice to get advice from professionals but it’s not always possible. Imagine that you could ask a beauty expert any time you had a question, and choose the best products for your concerns with ease! That’s exactly what AI driven models could do for you.
We already have a model called BERT – Bidirectional Encoder Representations from Transformers. This model has been adapted to the beauty world as BeautyTech-BERT (BT-BERT) and it can extract relevant attributes for product ingredients so it can recommend you products that are tailored to your needs based on skin type and concerns. By understanding the ingredients in a product and how they relate to various skin conditions AI could provide more accurate, personalized recommendations. (Liu et al., 2024)
This demonstrates how this model “thinks”:
Image source: (Liu et al., 2024)
This almost sounds like magic recommending products without actually seeing, assessing your skin. So how does it work?
The most important part in this case is the input information about you as the user and also about the products that the AI can recommend to you. Then it processes all this information to predict what specific ingredients, thus products will satisfy your needs.
The input in this case is a users interest + ingredients presented in the product + the title of the product.
The process then contains the following steps:
- tokenization = the input information is converted to numerical representation (tokens) that the model can understand
- input sequence = the tokens are then combined to a single input sequence with the query attribute placed at the beginning
- BERT encoding = this is all run through the model which learns contextual relationships between phrases
- attention mechanism = now the model can focus on the most important or relevant parts of what the user asked (In the case of COSRX Snail Mucin Essence it will pay attention to the ingredient “Betaine” as it’s associated with hydration)
- probability calculation = based in the representations the model learned it can calculate a probability of how satisfied the user will be with the product (Liu et al., 2024)
Another interesting beauty technology we have is OpenCV which is an open-source computer vision library using augmented reality applications. It can integrate elements into the real world so it can be used for skin analysis – detecting blemishes, estimate skin age. Another use for it the skincare world is virtual makeup try-on. By tracking a person’s facial movements in real time it can apply virtual makeup to the face.
The key components of OpenCV technology are:
- object detection: first OpenCV detects objects in the real world with a camera
- pose estimation: when the objects are detected it figures out its positioning and orientation in the real world
- augmented content generator: now a 3D model of text, objects or makeup can be created and be placed on the identified face
- real-time display: the final augmented image is displayed in real time allowing you to move around in a natural way or interact with the elements. (Tamilkodi et al., 2024)
Beauty brands are using AI to enhance their marketing or the customer experience in an attempt to find their unique selling point. By analyzing big amounts of data, AI can also provide insight into consumer behavior, preferences or trends. This was big brands can personalize their marketing efforts, deliver their message truly targeted and make recommendations resonating with their audience. (Coelho, 2024)
Why AI shouldn’t scare you
Even in its current stage artificial intelligence provides exciting opportunities for personalized skincare it’s important to understand the ethical considerations that come with it. At the heart of AI systems like the one we just unpacked is machine learning – a subset of AI focused on teaching machines how to learn from the data we feed it. In the beauty industry it’s not about replacing dermatologist or skincare experts, it’s aim is to complement human expertise by analyzing a lot of data faster than any human could.
So how does AI “learn“? It uses deep learning that mimics the way the human brain learns. It uses multiple layers of interconnected nodes just as neurons to process information and make decisions. These layers are complex, they use data in a way as our brain processes information in more stages. This “depth” matters in deep learning because the layers allow to learn more intricate patterns to be more precise. The process of training is called stochastic gradient descent – involving feeding the network small batches of data and adjusting the connections between them based on how well the network performs. (Bengio et al., 2021)
It’s like teaching a child to ride a bike, giving some guidance and adjusting their balance until they learn how to ride on their own.
Ethical benefits
AI can help you increase your effectiveness and productivity by automating tasks and improving decision making. Since AI is quick it can reduce tedious tasks giving us more freedom to work on what we want or simply on more important things. By addressing challenges as diseases, poverty or environmental issues it can also enhance wellbeing. (Stahl, 2021)
How could AI improve your beauty routine?
- Provide personalized recommendations: AI can analyze individual skin data and offer you a tailored to your skin specifically
- Enhance efficiency: AI can help in product development, customer service and help out the beauty companies in more ways
- Improve accessibility: AI tools can make skincare available to more people, helping them understand products and it can get to a wider range of people (Georgievskaya et al., 2023)
The future might hold remote consultations, more affordable products suggestions or even interactive tutorials to help you out with perfecting your skincare routine. AI can help you learn how to take better care of your skin regardless of your circumstances and where you are.
Ethical Concerns
As good as AI sounds there are still a few things we should keep an eye on. AI system can amplify existing biases presented in data, potentially giving unfair outcomes. The same goes out for power imbalances, those who have access to AI technology may have advantage over others. Experts also worry about potential long-term risks associated with advanced AI such as the loss of human control or creation of super intelligent systems. (Stahl, 2021)
What should we look out in the ethics of AI in beauty?
- Data Bias: if the data we use to train AI models is biased, like focusing on lighter skin tones the skin analysis for others might not be accurate, even unfair for people with different ethnicities
- Algorithmic Bias: AI algorithms may inherit biases from the developers which could lead to discrimination based on factors like race, age or gender (Georgievskaya et al., 2023)
How could this look in the real life? An AI-powered skin analysis tool might not be accurate for all ethnicities, if the training consists primarily of images of people with one skin tone. It’s similar with virtual makeup tools, they might prioritize certain beauty standards like symmetrical faces, and certain skin tones.
As we develop AI further we should still keep in mind the ethics of it for the wellbeing of society as a whole. Human oversight plays a big role in defusing potential risks, ensuring that AI systems align well with ethical values. We can prevent biasses from forming, control and intervene with the development. Another key point are global standards, we could say guidelines for AI development. International cooperation can help keeping AI responsible and address the ethical concerns on a global scale, making sure we are all on the same page. This could help out both the developers and the users of artificial intelligence.(Siau & Wang, 2020)
Sources
Coelho, M. C. C. D. (2024). AI-driven personalization in beauty retail: exploring how AI-based applications influence customer satisfaction and brand loyalty. http://hdl.handle.net/10400.26/51357
Georgievskaya, A., Tlyachev, T., Danko, D., Chekanov, K., & Corstjens, H. (2023). How artificial intelligence adopts human biases: the case of cosmetic skincare industry. AI and Ethics. https://doi.org/10.1007/s43681-023-00378-2
Liu, S., Suresh, R., & Banitalebi-Dehkordi, A. (2024). Beauty beyond words: Explainable beauty product recommendations using ingredient-based product attributes. In arXiv [cs.LG]. http://arxiv.org/abs/2409.13628
Stahl, B. C. (2021). Ethical Issues of AI. Artificial Intelligence for a Better Future, 35–53.
