• John Ennis

Eye on AI - November 22nd, 2019


Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!

This week, we look at the benefits and limitations of big data, focusing specifically on how big data has disrupted traditional market research in the CPG sector, then continuing by addressing the growing trend of many companies, news organizations, and others to exaggerate AI capabilities - sometimes to the detriment of the public.


Big-Data Approaches Disrupting Traditional Market Research



We begin with an article out of Food Dive, titled “Behind the scenes: Data and technology bring food product R&D into the 21st century,” which describes how Conagra Brands, after struggling for a decade to meet profit goals, saw a marked change once they transitioned market research funding into big data.


“We were doing all this work into what I would call validation insights, and things weren't working," Bob Nolan, senior vice president of demand sciences at Conagra, told Food Dive. "How could it not work if we asked consumers what they wanted, and we made it, and then it didn't sell?”

Conagra shifted its entire market research budget to big data analytics, purchasing immense amounts of consumer data to mine. It was a bold move. And it paid off, resulting in a 20% increase in sales over the past three years – after a 10% decline over the prior decade. Ferra Candy, making the same market research transition into big data, saw similar results. The reason: big data gives CPG companies the ability to address every facet of consumer data.


“There is data showing how often people shop and where they go,” writes Doering. “Tens of millions of loyalty cards reveal which items were purchased at what store, and even the checkout lane the person was in. Data is available on a broader level showing how products are selling, but CPGs can drill down on an even more granular level to determine the growth rate of non-GMO or organic, or even how a specific ingredient like turmeric is performing.”

The ability to look at so many different data points is immensely valuable to CPGs, which had been relying on robust research teams and consumer input since the 1950s. The old process to glean any sort of conclusive research data was often long, unreliable and expensive. Big data allows CPGs to be more nimble, giving them the ability to look at more data more accurately with less resources, inevitably leading to more profits.


“It's an old industry and innovation has been talked about before but it's never been practiced, and I think now it's starting to get very serious because CPG companies are under a lot of pressure to innovate and get to market faster," ​Sean Bisceglia, CEO of Curion, told Food Dive. "I really fear the ones that aren’t embracing it and practicing it ... may damage their brand and eventually damage their sales.”

Caution: Big Data Predictions Have Their Limits



Next, let’s take a look at an article out of Scientific American, titled “The Media's Coverage of AI is Bogus,” which cautions about the dangers of too heavily relying on AI and the media’s coverage of it.


“...the press will have you believe that machine learning can reliably predict whether you're gay, whether you'll develop psychosis, whether you’ll have a heart attack and whether you're a criminal—as well as other ambitious predictions such as when you'll die and whether your unpublished book will be a bestseller,” writes Scientific Americana contributor Eric Siegel. “It's all a lie. Machine learning can’t confidently tell such things about each individual. In most cases, these things are simply too difficult to predict with certainty.”

Siegel goes on to praise machine learning broadly, but cautions against relying on it too heavily, – what many call AI washing – especially in matters of human profiling, where research groups and media often inflate AI’s capabilities (see Stanford University's infamous "gaydar" study). The problem, Siegal notes, is that AI has difficulty with bias. When a test group is homogenous, results are usually consistent. But throw a single minority, and the prediction accuracy is thrown. Research groups may not test with variations that would significantly alter the accuracy of their predictions, sharing misleading results with the media to heighten the buzz around their technology.


“This “accuracy fallacy” scheme is applied far and wide, with overblown claims about machine learning accurately predicting, among other things, psychosis, criminality, death, suicide, bestselling books, fraudulent dating profiles, banana crop diseases and various medical conditions,” continues Siegal. “For an addendum to this article that covers 20 more examples, click here.”

The false narrative this manipulation drives is frankly dangerous, and leads to human bias (see dangers of using AI to predict human emotions) and the belief that big data predictions are equally accurate in all fields – they’re not. It’s just as important to look into how a model was tested as the end result.


Finally, we conclude with a fun video on the world’s first whiskey generated with machine learning:



Other news:


That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!


©2020 Aigora