I also got myself into a ChatGPT rabbit hole for the past month or so trying to push it when it comes to generating formulas and accords. It clearly has a good understanding of some materials and perfume concepts and I was initially shocked that it could even create formulas that at least don’t seem very off.
And while it’s currently a hit or miss, and clearly not trained for something so narrow, I can’t stop thinking about how we need this, and how immensely helpful it would be, in a niche (DIY or small-scale perfumery) where even basic math software has been lagging behind (it took a nice clean user-friendly software solution like formulair so long to appear).
Now imagine a perfume and formula-trained chatgpt. A not-so-small sample of 1,000-10,000 formulas with descriptions, perfumers, and data with styles and categorization.
Imagine you have a reliable, instant suggestion for “A smoky sandalwood Olibanum patchouli geraniol black tea formula with 40 raw materials, in the style of Alessandro Gualtieri”. Now you have a starting point or even a quirky idea accord that you never thought of or tried before, and you can go from there.
Instant game changer.
Wouldn’t you pay for that service? I’d sure empty my wallet. The bigger the data the better the model and so on.
People who will debate that it’s “cheating” or “theft” clearly don’t understand that “stealing” and repurposing ideas has always been at the center of creative endeavors.
In the word of the great Seth Godin:
If an art student studies all of Picasso’s 10,000 paintings and then creates a new painting that is clearly based on them, we call this the advancement of culture. The same is true if a writer uses a word that was coined by Shakespeare, or if a graffitist is clearly inspired by Shepard Fairey.
That’s how culture evolves. Taking an idea isn’t theft. Taking an idea is an oxymoron. Ideas belong to all of us.
We couldn’t and wouldn’t have it any other way. There’s no way to bake a cake, drive a car or write a sentence without using what came before.
GPT and other AI tools don’t actually know anything. They’re pattern matchers and pattern extenders. And those patterns are called culture.”
Luckily, chatgpt’s API is open, which means that perfume formula training is absolutely already possible.
I even asked it what are the steps to do that, lol.
The first team to make this happen would have an ultra-niche yet incredibly useful product in their hand. And It can be monetized in subscription or other ways, or even be free.
The multinationals like Givaudan see this future and have been investing heavily and have already developed their first generation AI bots that seem to do type some of formula creation plus even mixing, and a few years back we saw a stunning Etat libre perfume out prototyped by AI. But Givaudan cares for its pocket and its perfumers first and I don’t expect them to offer a public product any time soon.
There is no future of perfumery (like everything else creative) where AI is not going to lift some of the burdens of the creative process. And that is absolutely good news. The possibilities are endless.
Me, I’m looking to try and develop a tool that would be helpful and take some first steps, simply because I see how much I need it and how useful it would be. In my experience, it’s small, agile, passionate teams that create cool innovative solutions.
Any AI geeks here that are interested? : )
