![]() Dodo’s customers commented that the pizza tasted “unusual”, “surprising”, “tasty, cool, sweet, with lots of tastes”. Dodo received favorable reviews from both its customers and industry peers. On October 10, Dodo ran a “Open Source” pizza day in Moscow showcasing the AI pizza. Dodo’s goal is to make all the pizza in the world taste better rather than just making only Dodo’s pizza taste better. ![]() By adjusting coefficients, AI can produce both traditional recipes (Margaritas or Pepperoni) and unusual recipes, of which the “Open Source” pizza is a great example.ĪI code was shared through Github enabling every small restaurant in the world to become part of the AI revolution and win customers in the competition with bigger restaurants. To present the recipe in a form that would be understandable to the neural network, Dodo used Skip-Gram Negative Sampling (SGNS) - an algorithm based on the word occurrence in a specific context.Īfter AI had been trained in creating pizza recipes, Dodo added flavor combinations criteria based on the 2011 study mentioned above. The data cleaned up, however, only 1,000 unique positions were left. Initially, there were over 100,000 unique ingredients. The overall number of recipes searched through amounted to 300,000. The available recipe datasets are very limited, so Dodo scanned user-generated recipes on culinary websites, often with mistakes and typos. ![]() That finding was key to selecting recipes generated by Dodo’s AI.Īs expected, the problem was to collect the dataset, according to Dodo Pizza. The study found that Western cuisines showed a tendency to use ingredient pairs that shared many flavor compounds. To train the AI, Dodo used the dataset of over 300,000 recipes along with the results of “ Flavor network and the principles of food pairing” study published in Nature magazine in 2011. The Price List & Menu Maker App is packed with loads of free resources. The “Open Source” pizza recipe includes 10 ingredients both traditionally used in pizza (tomato sauce, chicken, cherry tomatoes, broccoli, and mozzarella) and very uncommon ones (melon, pear, tuna, mint, and muesli). “Open Source” pizza (ai.) was an experiment aimed at proving the hypothesis that tastes, though considered entirely subjective, could be quantified and that AI could be trained to find uncommon ingredient combinations that would taste well together for most people. ![]() I hope we are still around in 10 years.Dodo Pizza, a Russian-born global pizza delivery franchise, analyzed 300,000 formulations to create a pizza recipe transcending individual tastes. Dodo Pizza Mongolia, Ulaanbaatar, Mongolia. But it’s great actually because I get to chat with immigration guys and then, with my passport already stamped, I go past the line like I am someone important. □Īnyway, best of luck to us. It has proven to be pretty impregnable for robots.įor instance, every time I fly into Dubai, I go to the manual passport check because the automated recognition systems just constantly fail at verifying me. And it is a rare occasion when I am thankful for having horizontal nystagmus, a condition where your eyes move rapidly and uncontrollably from side to side. This is not a far stretch from when software will be able to read our minds essentially. And it turns out we already worked with them on a couple of projects. I was just at a call where colleagues presented a collaboration with a consumer research agency that, on top of classic surveys, uses facial recognition algorithms to dig for extra insights into perception, mood, emotional response to ads, banners, promo materials and UX.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |