How Oliver Zahn Beat AI’s “Cold Start Problem” to Make Plant-Based Cheese That Tastes Like the Real Thing
In big data and artificial intelligence, one of the most well-known challenges to success is the “cold start problem.”
The first cold problem refers to the lack of data recommendation systems in machine learning models. Just like a cold car engine that causes the car to sputter and shake as the driver begins their journey, an algorithm built to find and make accurate recommendations cannot function properly if it starts cold on the basis of little or no good data.
And it is this problem – the lack of basic data to build a machine learning model – that often prevents scientists, entrepreneurs, and companies in various fields from adopting new technologies such as artificial intelligence.
The cold start problem is something Climax CEO Oliver Zahn is familiar with. As a world-renowned astrophysicist working for Google and SpaceX in building complex data science models, Zahn knew that overcoming this initial hurdle was one of the reasons why established companies did not embrace machine learning and continued to use the status quo – whatever it may be – to build new products.
So when Zahn decided he wanted to build the food company of the future using AI, he knew the initial challenge of building a dataset that could be mined to find new and promising building blocks in the plant world would be his biggest hurdle. Still, it was a challenge he knew he had to take on.
“Traditionally, most large food companies today use a trial and error approach,” Zahn said recently when we sat down for a chat on The Spoon Podcast. They use the human mind to guess what might work. But often that misses things that are not clearly visible.”
Zahn knew that the less obvious could be the key to unlocking food building blocks that could yield new types of food. Those building blocks, from hundreds of thousands of different plants – many of which are inedible – can then be combined in millions of different ways to provide new functional or sensory properties to create something like plant-based cheese. The only way to get there was to use machine learning, cold start problem or not.
“It’s a big problem for combination testing,” Zahn said. “Even the greatest food labs on Earth, if they were all combined, would not be able to test all the combinations over millions of years.”
He knew the AI couldn’t if he could get past those initial hurdles. But to do that, he knew that Climax would have to start not by gathering much information first on plants but on animal products.
“We started by investigating animal products in depth to try to understand what makes animal products compatible with the way they do,” said Zahn. “Why do they have their texture profile flavor profile? Their mouth feel? Why do they sizzle? Why do they melt and stretch when you eat them?”
You’d think most of that data already existed, but according to Zahn, it didn’t. The reason for that, he explained, was that there was never a business reason to build those datasets. But as the environmental impact of animal-based products has become more visible in recent years, there has been a drive to start businesses to understand how these products are labeled to be recycled using sustainable inputs.
The data the company collected from researching animal products allowed them to build their own machine learning models on the labels to accurately describe and label the food product. To that end, Zahn said the company is committed to building data sets around plant-based building blocks.
“We’ve built a lot of data sets on the performance of plant ingredients and different methods of combining them. We then find these methods that can closely recreate animal products, and sometimes in subtle ways.”
Zahn says the process of building accurate models can often take a very long time — up to 20 years — especially if those building them don’t have the good sense that comes with machine learning.
“From the point of view of someone starting a food company, that (long-term horizons) can be scary, right? Because you need to get to the market at a certain time. And so unless you have a very good vision and have a lot of experience, in my case, several decades, of trying to find meaning in dirty, big data sets, people are not even starting.”
For Zahn and Climax, the models they built are starting to bear fruit, enough to help them start making what will be their first product – cheese – using artificial intelligence. What helped them get there so quickly was Zahn’s experience in building these models that told him to start by trying to understand and explain certain characteristics of animal products – be it taste, mouthfeel, or nutritional benefit – and then find a combination of plant-based building blocks that achieve the same result.
“Looking in the plant environment for something that’s chemically similar to an animal ingredient, like a protein you might go after, is a little red herring,” Zahn said. “Because it doesn’t have to look the same through the microscope, or its sequence doesn’t have to be the same, to behave the same way. There may be other ways to accomplish the same task.”
Now, just two and a half years later, Climax is ready to begin rolling out its first products, a cheese line that includes brie, blue cheese, feta, and chèvre (goat cheese) made with plant-based ingredients. It’s an impressive feat, in part because, as a first-time entrepreneur, Zahn also faced the challenge of learning how to build a company, itself another “cold start problem.”
If you’d like to hear the full story of Zahn and Climax Foods’ journey to create plant-based dairy products, you can do so by listening to our conversation on this week’s episode of The Spoon Podcast. Click play below or find it on Apple Podcasts, Spotify, or wherever you get your podcasts.
