It is the night before the weekly shop. I look in the fridge and consider my three tomatoes, the sweet potato and the asparagus.
Normally, I’d take this as my cue to nip to the fish and chip shop.
However, I’m trying out Plant Jammer, an app that promises to rustle up a recipe based on whatever food you have lying around, using artificial intelligence.
It searches three million recipes to find often-paired items. It then consults a library of ingredients which the company has hired professional chefs to group by flavour – salt, umami, sour, oil, crunch, soft, sweet, bitter, spicy, fresh and aroma.
Finally, the software learns from this data and devises new recipes.
Michael Haase, the founder of Plant Jammer, says this last step is what makes his app unique.
Traditional recipe apps are powered by databases – you list what you have in the fridge and the app sends a pre-existing recipe it found on the web.
“That is the old way,” says Mr Haase. “We are actually constructing new recipes from scratch each time with an AI [artificial intelligence]. This is going to be the future.”
Plant Jammer is one of a handful of recipe apps, food distributors and even events companies that are turning to artificial intelligence to gain an edge in the food industry.
To make use of my sweet potato, the app suggests several meals including a stew and a fry up.
I chose to make them into vegetable burgers. I tell the app I have no dietary restrictions, then tick off my ingredients. Lastly, it asks what seasonings I might have.
Based on what I have ticked, my sweet potato patties will also include asparagus, aubergine, chickpeas, lemon juice and crushed-up walnuts. I add some seasoning and rolled oats to bind them.
They go into the oven for 15 minutes. The result is four overcooked, and strongly oat-flavoured, discs.
When I tell Mr Haase, he admits that not every recipe is a success and also agrees the recipe probably needed more options to bind the patties together.
An hour later, the platform has changed to adjust for my feedback. I promise to try the recipe again.
There is a prime membership available which around 5% of users sign up for, paying for the running of the app.
Plant Jammer also sells subscription plans to supermarkets, offering ingredient alternatives to their website recipes.
“So if you want to make it vegan, gluten free or Thai we can adjust any recipe,” says Mr Haase.
He hopes Plant Jammer will offer people the chance to master less-wasteful, vegetarian cooking.
‘The hard way’
Even packaged food manufacturers have turned to artificial intelligence.
Analytical Flavor Systems is New York research and development firm that uses AI to advise food companies on improving their products or creating new ones, including drinks.
Its AI platform Gastrograph can predict the flavour, aroma, and texture a drink would need to cater to any regional food preference.
“We’ve done this the hard way,” says founder Jason Cohen, who has spent the past 10 years running taste tests around the world.
Every day, his panel of 50 tasters try different packaged food products two or three times a day. Before Covid-19, there also had a travelling team visiting a different country each week to test regional preferences.
What people taste is less important than what they perceive when they taste, says Mr Cohen, a former tea sommelier, who adds “perception is a very easy thing to play with”.
“For example, if we add vanilla at about one parts per million to milk, you won’t be able to taste the vanilla, but you’ll say that the milk is creamier and higher quality,” he explains.
The artificial intelligence software runs through hundreds of decisions until it learns to predict how good a product is going to taste – based on what the product is meant to taste like, panel testing, and regional tastes.
Using AI to find new combinations of flavours for cupcakes and cocktails put Bristol based media agency, Tiny Giant, on the map.
Co-founders Richard Norton and Kerry Harrison have used AI modelling to help create marketing events, ad campaigns and even gin labels.
With Monker’s Garkel gin, Tiny Giant’s coders fed a computer hundreds of different gin names. The computer analysed the samples so it could invent its own.
This kind of machine learning is called a neural network – when a computer creates one it will recognise a pattern, like ‘what does a gin label sound like or what goes into a cupcake?’ – and then make a creative decision about it.
After Tiny Giant’s weekly AI cocktail night got the attention of larger companies, they were inundated with requests from large corporations to host events with AI-generated cocktails and cupcakes.
“I didn’t really expect this to become a thing where we would become food creators, but why not?” says Mr Norton.
Cookbook author and chef Meera Sodha agrees the pairing of AI and food can foster research, creativity and sustainability, but says you cannot “sever a recipe from its story”.
Ms Sodha was inspired to learn cooking after a trip to Brick Lane with university friends.
“My wonderful intelligent white friends asked me what they should order from the Indian curry house.,” she recalls. “I was flabbergasted that they thought that this Indian food was what I grew up eating”.
When she learned to cook from her mother she had a further “huge moment of panic” when she discovered no family recipes were written down.
They would all die with her if she did not make a record of them.
“What I love about cooking the recipes collected from my mum, my grandma or my aunt is that I feel connected to them when I cook that food in my kitchen,” she says. “I feel like they are there by my side.”
In this spirit, I attempt the potato patties once more but this time with my mother Anne Mooney, a former professional chef, over Skype.
But she prefers not to let the app tell her how to cook – using it instead as a spring-board for ideas, particularly the combination of chipotle, cilantro and pine nuts.
We both avoid the oats and fry our patties.
They taste better, but I think this has more to do with our online get-together rather than our command of technology.