I don’t know if this is solely because I am in an echo-chamber of fellow journalists theorising dystopically on the future of our industry, but it feels like the narrative surrounding AI is predominantly negative. There is, of course, the environmental impact of AI centres: the immense amount of energy and water required to operate them, the fear surrounding AI automation and the subsequent likelihood of redundancies and misinformation in these early stages of its usage. All of these concerns are rightfully valid, but the train has left the station and if you’re hoping to look more positively at what AI can do for us, there is one category where I am particularly excited about the impact: skincare.
More specifically, skincare tools. They’re about to get a whole lot more personal. Very much in the infant stage, the possibilities are only just revealing themselves at the intersection of beauty, tech and AI. These include skin analysis through image recognition, personalised product recommendations, adaptable technology based on the user and the ability to try before you buy.
“With the advances in AI and computer vision, the next frontier will be that you can take images of yourself and get an analysis that you can trust; not only what’s the right product for you, but also how to use that product the best,” says Tim Roberts, head of science at Therabody.
More advanced gadgets are constantly being developed to transform technologies once reserved solely for clinic practices into our daily routines, making them more convenient and financially accessible, reaping effective, clinically validated results. The existing LED masks, RF devices, microcurrents and body sculpting gadgets are just the beginning. “I think the future of democratising dermatology from the medical skincare side of things is certainly going to fall into the hands of AI – as many areas of medicine are going to,” says consultant dermatologist and skincare innovator Dr Emma Craythorne.
But where are we at now? According to the professionals, we are not yet at the point at which AI in skincare tools can analyse skin conditions or detect the likes of acne, wrinkles or pigmentation. She explains: “AI in skincare is still unable to make a diagnosis. Making a diagnosis is something that has to be done in person or by an algorithm. It cannot be done by AI as that would turn it into a medical device, which requires additional regulations.”
Below, we investigate the current crop of AI-powered skincare tools and call on Dr Craythorne’s expertise to answer our burning questions.
What can AI do in skincare tools?
The technology is still in its very early stages and Dr Emma Craythorne stresses that it’s still limited in its impact. “It’s unable to make actual diagnoses,” she says. “Most times for very effective skincare, a diagnosis is usually key, and that’s certainly the case for people who want to make a big difference for their skin. For example, with acne, the diagnosis cannot be made on an AI device; it doesn’t have the ability to do that. Wrinkles can be seen from photographs, but what is the cause behind them? We’re not quite there yet with AI being able to make these diagnoses – either reliably or by legislation. But if we were to move in that direction, it would speed things up.”
How does AI personalise skincare routines for different skin types?
“AI at the moment just draws upon experience we all have and uses that experience to present it in a way that’s more user-friendly for the person,” the dermatologist explains. “The issue with that is that at the moment it’s not really going into enough detail that’s going to be able to make that really key difference. This is something that is really going to change over the next 5-10 years.”
How accurate are AI skincare assessments?
According to Dr Craythorne, “They’re not accurate enough yet.” She continues: “That is why they’re not used within practice or to confidently make diagnoses on an individual person. That said, people are using the technology at the moment in almost a slightly gimmicky way rather than in a true way that allows for very effective change.”
How do AI tools decide which ingredients or products to recommend?
“They make the decision of tools or products to recommend based on what their assumption is of the skin in front of them,” explains Dr Craythorne. “So, they will still be using quite generalised models. For example, someone may have spots and they’re a teenager, so it will assume that they have acne. And then the top ingredients we know for acne are benzoyl peroxide, salicylic acid, adapalene and tretinoin. The next stage is for it to go one step further by asking different kinds of questions.”
What breakthroughs in AI could transform skincare in the next 5-10 years?
“There is going to be more learning amongst these large language models to truly think like dermatologists in terms of trying to transform how the skin behaves,” theorises the derm. “That has to be coupled with the acceptance that these models are essentially making medical diagnoses, which can then go on and prescribe the appropriate treatment. So the appropriate treatment for someone with red spots is not something over the counter. Usually, it’s a medical diagnosis to try and determine why they’ve got red spots, so quite often it requires a medical treatment. Therefore, there has to be an acceptance that a model like this is allowed to make that diagnosis and claim that diagnosis, as well as then prescribing appropriate treatment.
While she suggests that these changes will take around a decade, she also spotlights current tech. “At the moment, it can be used in the beauty and skincare industry for things like picking out what colours or routines suit people.”
What role does lifestyle (sleep, diet, environment) play in AI skincare systems?
Doctors will, depending on the condition, use lifestyle information to inform diagnoses, but is this something that AI will be able to do too? This is important, says Dr Craythorne. “The exposomes (AKA things you do that affect how your genes are functioning or behaving) play a huge role and we understand this as doctors. AI models will therefore also understand that it plays a huge role and they will be asking those questions.”
