Machine Learning and the Future of Humans in Manufacturing

In January, ARC Analyst Greg Gorbach wrote about the state of digital transformation in manufacturing. Greg nailed a number of points in his report: that digital transformation will “touch nearly every aspect of business”, that it will be “widespread and far reaching”, and that “the most noticeable and probably most important trend today is the proliferation of advanced analytics and machine learning.”

As the CEO of a startup that’s applying machine learning in manufacturing, I was nodding my head - until I got to this phrase:

“The technology has reached a tipping point and can now deliver value in setting
after setting. In turn, this fuels demand for smart connected sensors, digital
networks, and other ways to collect and move data to the analytics systems –
and vice versa. Is mass displacement of workers on the horizon?  Perhaps.”

I instantly emailed my marketing head, Dave Prager, with "When we’re ready to come out of stealth, let’s call Greg, and talk about the relationship between machine learning and human workers."

That day has come. Last week, we got on the phone with Greg and explained Drishti’s thesis: that humans are actually going to be more important in the factory than ever. Far from displacing people, machine learning will actually make factory workers more valuable.

The exaggerated demise of humanity

One can’t help but get the feeling that humanity is becoming obsolete. Every day, the media reports yet another industry in which robots are poised to dominate: manufacturing, of course, but also food prep, medicine, even assembling Ikea furniture.

These headlines, though, are contradicted by what I’m seeing in my customers’ factories - and, I suspect, what most industry professionals know about their own businesses: that humans remain dominant, and that they aren’t going anywhere any time soon. Boston Consulting Group estimates that just 10% of factory tasks are performed by robots. If you’re reading this on a phone, it’s extremely likely that the ratio of humans-to-robots that touched your phone in the final assembly and testing process was something like 50-to-1.

Am I walking the wrong factories? Perhaps. Maybe the most forward-thinking factories are too busy ripping out workstations and installing robot arms to invite me to visit. After all, every conference invitation I receive is full of sessions extolling the virtues of IIoT, robots and platforms for connecting your machines to your analytics. And, I vividly recall my trip to Fanuc’s “lights out” plant at the foot of Mt. Fuji in 1993 where robots make new robots without human intervention. At the time, I was a true believer that machines would rapidly supplant people.

On the other hand, it’s 25 years later and the Fanuc plant remains an outlier. So perhaps there’s a simpler explanation for the market excitement around instrumented machines: Maybe they’re just the easier problem to solve.

Humans as a data source?

“The conventional wisdom is that automation is inevitable,” says Peter Marcotullio, Vice President of Commercial R&D at SRI International. “But machines are easy to measure, and people are challenging to measure. Automation may just be a knee-jerk reaction when a company can’t quantify the contributions of humans to their processes. This data imbalance makes automation a reflexive decision - but not necessarily the correct one.”

This leads very nicely into what I discussed with Greg on the phone. Drishti believes that robots and machines aren’t the future so much as they’re simply the low-hanging fruit. Machines create binary streams of data; it’s not hard to build tools to capture these streams and platforms to analyze them.

Humans are different. They’re more flexible than any robotic equipment, and they create more value on the line.  But they’re also more variable than robots, and thus harder to leverage as a source of data.

A solution to this problem is what will truly define manufacturing's trajectory. There’s a judo move to be pulled on robots: taking the technology that’s currently making them so attractive to manufacturers - namely machine learning, computer vision, Big Data - and applying it to augment humans.  

That’s my prediction for the future. Machine learning is now poised to find meaningful patterns in the messy, variable, and heretofore immeasurable activities of human beings.

When I meet with customers, I ask them a simple question: Ten years from now, would you rather have put your chips in robots that are assisted by AI, or humans that are assisted by AI?

This is what I shared with Greg last week. Maybe he bought it, maybe he didn’t. I only have one datapoint: he invited me to write the article you’re reading right now.

Which suggests, at the least, that he is perhaps a little less concerned that humans are on the road to obsolescence than when he wrote his report in January.

Dr. Prasad Akella is Founder & CEO of Drishti, a new company that’s deploying AI to collaborate with and enhance humans on the factory floor.