DENVER – At Cereals & Grains 19, held Nov. 3-5, Hamed Faridi, Ph.D., chief science officer, McCormick & Co., Inc., and Robin Lougee, senior industrial research scientist, IBM Research, shared how the two companies partnered to use artificial intelligence (A.I.) to accelerate new product development at McCormick & Co.
Since 2015, McCormick & Co. and IBM Research have been working together to take the flavor company’s massive amounts of consumer and formulation data and apply it to an A.I. system to help McCormick & Co.’s product development become better, faster and more cost-effective.
While the definition of A.I. can be vague, Ms. Lougee said the main point is that A.I. attempts to mimic the complexity of human intelligence with a combination of algorithms that can learn and change.
When developing IBM Research A.I. for Production Composition, there were several challenges the companies faced. Would the A.I. be able to create product formulas, get better with experience and predict how people will perceive taste? Also, would the computers be able to be creative? And then there’s the obvious problem of computers cannot taste, smell or touch the products being tested. But the main question McCormick & Co. wanted to answer was: Would A.I. be able to help a skilled human flavorist develop products?
“It has exceeded our wildest expectations,” Dr. Faridi said.
McCormick & Co. made a great partner in this endeavor, Ms. Lougee said, because the company had 40 years of data to deliver to the A.I.
“Data is the life breath of A.I.,” she said.
McCormick & Co.’s data included consumer science and flavor palate data, flavor and seasoning formulations, and food science data points. That information paired with learning algorithms has created and A.I. assistant that doesn’t replace McCormick & Co.’s human flavorists but acts as their apprentice. The A.I. churns through data, looking for patterns to inform novel combinations and formula improvements.
The beauty with this model is the A.I. does not have any of the biases a human might.
“When working on a pizza flavor, for example, I might think automatically of Italian flavors,” Ms. Lougee said, citing a real-world example. “But the A.I. offered up cumin as a potential flavor because of the warmth, and the product developer really liked it.”
With A.I. technology, Dr. Faridi anticipates that McCormick & Co. will be able to accelerate flavor innovation by 300% by late 2021. And he sees the advantage A.I. will bring as food producers try to innovate within more complex requirements such as balancing flavor, cost, nutrition and sustainability.
“A.I. loves complexity,” he said. “A.I. wants more complexity, more data points, because with more complexity it can find more insights.”
It isn’t without its limits, however. Ms. Lougee stressed A.I. is not a turnkey solution for product development. It still requires a large digital data set and extensive training for staff. But if it’s something a food company is interested in pursuing, she said now is the time to start collecting the data needed to power a helpful A.I.