This post is about testing recipes created by artificial intelligence, Chef Watson of IBM
Dream about the future, where robot arms cook in your kitchen and all you need to do is to purchase a Michelin chef’s recipe and push the button ‘cook’. It even finds alternatives to ingredients you are allergic to, and ingredients with lower calories if you’re on diet.
This was my initiative to kick off this personal project. As I stress in my profile, I’m a serious foodie and I fell in love with cooking. I was curious how the cooking experience would change in the future. At first, I imagined flicking a magic wand to create delicious and creative recipes created by the super computer. But after several failures and found faux pas with the Chef Watson’s recipes, I became a more independent cook and came up with several ideas that could improve Chef Watson from a UX perspective (For those who might’ve forgotten, my job is a UX designer).
1. Add UX to input user feedback
Deep learning requires human assessment to tell if the what computer has learned is right or wrong. However, unlike case guessing, if the animal in the picture is a dog or a wolf, testing recipes needs a lot of effort and time for testers. Adding a feature to add a star-rating, or adding in-line comments as Medium does will encourage adventurous cooks to try and start a discussion about the recipe.
2. My inventory feature
There are condiments or sauces that are always on the shelf, and it would be great if I could add them in the user Settings and Chef Watson suggests recipes using the items in my inventory more often than other recipes.
3. Difficulty setting feature
Well, for the novice cook, it’s hard to figure out the difficulty and expected time to complete the dish by reading the recipe. This feature will help a lot in engaging more users to try Chef Watson.
Currently, the content provider is Bon Apetit and it only covers western recipes but ingredients and cooking method are so different between countries. If there’s a setting to set my location, Chef Watson can suggest alternative ingredients which are easy to find in that country; for example, herbs are quite expensive in Korea and I always visit the grocery corner of the department store because supermarkets and online supermarket don’t sell fresh herbs.
5. Taste balance graph
Even though Chef Watson suggests alternative ingredients by calculating the balance among aromatic compounds included in each ingredient, changing one ingredient can change the overall taste. If there’s an interactive balance graph of the taste of outcome, it will help users to imagine how it will taste when cooked.
Here is the quick summary of my Chef Watson project and posts of each trial so far.
- Project timespan: 4 months
- Number of recipes to test: 10
- Budget: I didn’t count:-(, I should’ve recorded every receipt.
- Success rate: 60%
- Most impressive recipe: #3 Apple Chicken Meatball
- Worst recipe: Avocado Paella
For those whom it may concern, I’m not related to IBM or Chef Watson. It is my personal interest to test how artificial intelligence decodes the way chefs create recipes and I want to see if this service can be offered to everyone in the near future.