Earlier this week, I read an article on the Amazon Science blog discussing an algorithm that can “smell.”

It does not smell. But based on the article, it can successfully label (colloquially) the type of smell based on the molecular structure. The algorithm was trained on millions of data points, where the smell was the label and molecular structure the attribute.

The Amazon Alexa fund funds the research. Amazon is looking to invest in technology that can enhance Alexa’s capability. It will be tricky, though, since embedding the capability of analyzing molecular structure in Alexa devices will be nearly impossible (while keeping the product form factor and price within a reasonable range).

A workaround can be to train the algorithm on the molecular structure of standard household items, like bread (though the molecular structure may differ based on the bread variety). So if the smart camera detects bread, it knows the molecular structure and the smell.

The capability to smell, though, does not have an immediate lucrative commercialization opportunity in the context of Alexa. But this can be leveraged in other products, as the article mentions. The research goes on since the algorithm passed the “smell test” (pun intended).

Hundreds of machine learning and deep learning papers on the internet focus on solving focused problems. The problems they propose to solve may have been too academic and had nothing to do with problems faced in the industrial world. The methodology in those papers dies with the paper.

However, you should leverage your business process knowledge for every such paper to apply a “smell test.” With a focus on whether the suggested methodology can be leveraged beyond the problem, it is being used to solve. And if yes, the algorithm has passed the smell test. You may have found a unique solution approach that no one else in your industry may have in a research paper probably published years ago.

I devote a significant portion of my time to combing through research papers that I find from my search based on specific keywords. Apart from learning a few titbits of information on subjects I wasn’t familiar with, at least 10% of these papers provide insights into new approaches to solving problems in the supply chain, marketing,finance, lifesciences and socioeconomic problems.

Academia devotes significantly to research, and I often hear that much of it goes unleveraged. But maybe because we are focusing very rigidly on the topic and problem set of research but not applying the “smell test.”