Cracking the Cola Code
Photo by Donald Tetto
by Christy Bell
Faculty Sponsor: Steven Alexander
As part of last years’ King Creativity award I successfully developed and tested a computer program that used genetic algorithms to create a unique perfume. In that project, I chose 30 oils with a wide range of scents. My program began by generating 10 random mixtures. Each mixture was simply a list the oils and their proportion. After creating each combination of oils I then sniffed all 10 mixtures and ranked them based upon their aromas. The proportions of the best three mixtures were then placed back into my program and this information was then used to generate the next 10 mixtures. After seven iterations this process converged and a new perfume was born.
For this King Creativity proposal I used the same algorithm to discover the secret behind the Coke Cola formula. Rather than using my genetic algorithm to explore different combinations of smells, this algorithm was used to find the mixture that most closely approximates the taste of Coke. Starting with a list of probable ingredients, the program altered the proportion of each ingredient based on my sense of taste. Mixtures that closely resemble Coke survive and pass along their information to the next iteration. After a few iterations this process converged and produced a drink that is indistinguishable from Coke.