Hey! I’m Vova, and at TIMELESS, I’m in charge of the day-to-day operations. My team handles the venues, keeps the retail running, makes awesome products, and comes up with a ton of marketing activities to get our brand out there. Today, we’re launching a unique menu that was almost entirely created by ChatGPT and MidJorney neural networks. We’ll give you a sneak peek into how it was made and show you some of the new menu items.

Neural networks are the current trend, and they’ve seen a surge in popularity in recent months. Almost everyone has experimented with artificial intelligence, from beginning developers releasing applications in days to students writing pretty decent research works in just hours. Designers are merging with DALL-E and other art neural networks in a state of pure creative ecstasy. We got super excited about the idea of creating a real menu using neural networks. We chose ChatGPT and MidJourney as our main tools, and early draft picks included DALL-Y. 

We started the project by defining the desired result. We wanted to get:

1. Two signature hookahs;
2. Three or four cocktails;
3. Several shots that can be served individually or as a set.

According to the project’s terms, neural networks were responsible for the concept, compositions, descriptions, names, and visuals, while human influence should be minimal, essentially having no impact on the generated results.

During the first week, we learned how to use neural networks, went through the circles of registration hell, and tried writing queries. After learning about the possibilities of AI, we started working out the concept that would be the connecting thread for all positions. The guys asked ChatGPT to come up with ideas for the hookah and cocktail menu, and the neural network made a list with the most diverse options, from urban jungle to art deco style items. We decided on the art theme because it fits with our view of the product.

For hookahs, our brand chef, Timofei Shepilo, went with a pop art style, using bold colors, graphic patterns, and playful images. Neural Network took cues from the works of Keith Haring, Maurizio Cattelan, and Banksy, drawing inspiration from stencil art, graffiti, street art, and murals. This blend of styles, artists, and eras led to the creation of two new positions.

Visualization by MidJorney

Maurizio Cattelan | Comedian

We really liked the neural network’s flavor preferences. Banana, mango, lemon, chamomile, and vanilla are a perfect mix.

The name was chosen from 50 different options, and we all agreed that this one really fits.

As a result, we got a delicious hookah with the flavor of banana soufflé with mango yogurt, lemon zest, and chamomile, served in a bowl with a stylish Nike Air Force 1 design and intricate pop art prints.

By the way, the Nike Air Force 1 sneaker was chosen for a purpose. This particular model of shoes in 2019 was released with a brand logo stylized as a banana.

Visualization by MidJorney

1. “Love is in the Bin” is a work depicting a little girl with her hand reaching for a heart-shaped balloon that appeared on the streets of London in 2002.
2. “Flower Thrower” (Jerusalem, 2003) is a work in which a masked rebel throws a bouquet of flowers instead of a Molotov cocktail.
3. “Nobody Likes Me” is a mural depicting a crying boy with a smartphone and social media notifications (Vancouver, 2014)
4. “Keith Haring’s Dog” is a drawing of a man with a dog (London, 2010).

This drawing, by the way, is a reference to the second hookah.

ChatGPT’s flavor recommendations were intriguing: gooseberry, basil, opuntia (a prickly pear cactus), lime, agave, and gin. We left the mix largely unchanged, making minor adjustments to the proportions.

The result is a concrete base with hand-painted reproductions of Banksy’s works using a stencil and a composition of barbed wire, stones, and mesh.

The cactus bowl contains sweet gooseberry and prickly pear cactus aromas, with basil and juniper notes, balanced by citrus juice.

COCKTAILS

We took a different approach with cocktails.

We asked the neural network about the artists it knows. Then we picked our favorite artists and used the same kind of queries: “Write 5 cocktail ideas inspired by the paintings of Pablo Picasso (substitute any other artist).” If we felt it could do better, we asked for more options. Sometimes a cocktail was seen ready, but the decoration, glass, or ingredient didn’t fit. In that case, we asked it to suggest alternatives.

Our main bartender, Ioann Mikhailenko, wrote out all the cocktails the neural network suggested, tied specific pictures to them (if there were any), and got rid of the simple and trivial options. We worked out the rest carefully.

After getting the positions and colors right, we used MidJourney, just like we did for the hookahs. We fed a text description of the cocktails and what we wanted to see into the program, and it generated a bunch of options. We also asked ChatGPT to write a welcome text for guests who would open the physical menu, so they would read a text written entirely by artificial intelligence. ChatGPT also wrote the descriptions of the positions, and our guests can try the results of the AI’s work today.

Vincent Van Gogh “Almond Blossoms”

Vodka and a tasty mix of tomato juice, veggies, herbs, spices, and sauces come together to create that unique taste. ChatGPT suggested using celery as a garnish, linking it to the pop art and consumer culture on display. The drink is a fun nod to Warhol’s love for the everyday, as you can see in his work.

ChatGPT suggests that our absinthe and apricot variation of Margarita represents the mysterious and inexplicable elements of desire in this piece.

The neural network pictured Enigma as a cocktail with no toppings. But our transparent caramel gives it these craters on the surface that look like the centerpiece of a painting. It’ll transport you to a world of beauty and introspection, like the works of Dalí.

The creamy taste of Advocaat egg liqueur combines with spicy cinnamon whiskey, and we balance the sweetness with freshly squeezed lemon juice. Just before serving, we pour in raspberry liqueur to give the shot hints of raspberry cheesecake. Then, we add extra texture with shavings of coconut and dried berries. Raise your glass to a movie screen icon!

We’ve found this experience fascinating and interesting. If earlier all our approaches to developing new positions were accompanied by crises of ideas, long discussions, and arguments, now we got a portion of inspiration in a few requests. Undoubtedly, artificial intelligence is more successful in other areas; for example, it can write research works for students or check code for programmers, but not all neural network ideas will find their realization in the hospitality sphere. It is difficult to evoke human feelings; it requires a keen eye, a sense of taste, and an understanding of the guest. Predicting what people will like in the real world is extremely difficult even for industry professionals, let alone machines, but they can help create something new. This is something we have proven, and you can experience this for yourself at any TIMELESS.