University professor programs artificial chef

A new computer program created by University professor Lav Varshney could mean the end of the cookbook as we know it.

Varshney, a professor in the department of electrical and computer engineering, along with his colleagues at the technology company IBM, have programmed a system that can create up to a million entirely new food recipes in under three seconds completely on its own, ranking them in order of what it predicts to taste the best — no chef, no cookbook.

“Broadly, we were interested in pushing computing to this whole new direction of creativity,” Varshney said. “I chose food as the initial domain because it’s something everyone can understand — everyone eats, and it’s central to human culture.”

The program, which Varshney has dubbed is the Computational Creativity System, or CCS, which was finished in early 2013 after about a year of work. It has been refined to be even more efficient since then. 

What makes CCS stand out from a human chef, besides the ability to create millions of recipes within seconds, is that it’s indiscriminate in the ingredients it uses, whereas a human chef may be limited to their own biases. 

For example, while the combination of brussels sprouts, goat cheese and tomatoes may not have crossed anyone’s mind in the past; these ingredients were combined to create a completely new recipe: Kenyan Brussels sprouts gratin, Varshney said.

Varshney recently attended a banquet in Berlin to demonstrate CCS in front a crowd of 400 people. The banquet, which had American and Germany ambassador and other dignitaries attend, featured four appetizers, four desserts and two main courses, all of which were invented by the computer program. 

Varshney said the guests didn’t have any reservations when it came to eating a meal created by a computer.

“They trusted that we were serving reasonably good things,” Varshney said. “In fact, people were pretty excited to try everything.”

When asked, University students said they wouldn’t mind trying a CCS recipe either.

“Yeah, why not,” said senior Max Haneberg. “I’m open to any type of food.”

“Sure,” said freshman Angelina Gerbas. “I don’t think I would have any reason not to.”

CCS creates these new dishes with three input arguments that help the program “steer” toward an ideal recipe. CCS starts with a key ingredient that the programmer chooses. If the programmer is to choose pecans, he or she will then decide what type of regional cuisine it will be. For example, it could be Chinese, American or Italian cuisine. 

From there all that is left for the program’s user to decide is whether they want their final product to be a soup, salad, dessert or any other type of dish. Finally, CCS will determine the appropriate proportions for ingredients and even tell the programmer the best way to prepare the final dish.

CCS could be defined as an artificial intelligence program due to its creative component; however, Computer Science Professor Dan Roth cautions against this interpretation.

“I think that when you say to people ‘artificial intelligence,’ either it scares them or impresses them too much,” Roth said. “People shouldn’t think this is what A.I. is.”

Roth is no stranger to programs that appear to be intelligent at first glance. He helped design a program called Wikifier that creates Wikipedia-style entries from a given text, linking the reader to information on organizations, people, places and things that are included in the original piece.

CCS falls in the same vein as his program in the field of artificial intelligence in that it makes decisions on its own; however, it’s still limited by its programming and a long shot from the AI portrayed in science fiction movies.

CCS creates recipes in a similar fashion to how Google and Facebook determine which ads to display on a user’s computer. Roth pointed out that like these programs, CCS has large databases that it references to combine bits and pieces to satisfy pre-determined rules or conditions. In the case of CCS, these conditions would be the input arguments that guide the system.

Even though Varshney’s computerized chef may not bring computer science any closer to creating a sentient computer with Scarlett Johansson’s voice, he still has big hopes for its future, predicting that CCS could help create healthy, low-cost meals for schools, hospitals and nursing homes as well as helping to fight America’s obesity epidemic.

“It’s actually useful for social benefit,” he said. “If you can come up with recipes that are flavorful and novel and healthy, then that can be pretty powerful.”

Josh can be reached at jjwinte2@dailyillini.com.