With machine learning being hyped up as being able to completely eliminate the time suck of “statistical guessing,” I’ve noticed that there is a debate on if this is truth or hype. However, the only reason this debate even exists is because of how slowly the results from big data analysis have edged into general consumer awareness.

image via Analytics Vidhya

For every reference of how big data has changed the way Netflix, Facebook, or Lyft do business, those references are drying up as companies are using their machine data so intuitively that the result just seem to “make sense.”

But making the big data experience intuitive in this way is creating a veil of obscurity. This makes it hard for non-data scientists to understand the importance of insights from machine data. Especially when the insights are already in line with things we instinctively know.

For instance, PBS’ The Human Face of Big Data portrays an everyday explanation of how a child learns language.

While the visuals were amazing, and the way Deb’s team captures the machine data was fascinating, I did stop right after watching it and say, “Well, duh.”

Knowing that I need to teach my son a word by giving him context (“This is water, here touch it”) is pretty intuitive and not surprising. Knowing that I need to show him what water is and then repeat the word, as opposed to sitting down with him and repeating the word “water” 300 times is and should be… intuitive.

And that belies the problem with insights derived from machine learning. Currently, machines are simply justifying things we already know thanks to “gut instincts.” Whether you call it intuition or a “sixth sense”, humans are not blind to the millions of data signals that are given out by the world around us. We process these signals through our senses and are able to derive complex understandings – like knowing when someone is being sarcastic – which should be completely incomprehensible if taken at face value. This is why even the youngest child can tell the difference between a concentration camp and a jungle gym, yet Flickr could not.

But I think that’s what makes machine learning so intriguing. By breaking down complex signals and being able to pinpoint how one data point is directly responsible for a problem is amazing. Forcing technology to slow down and explain how high blood pressure during surgery can affect if a spinal cord injury patient ever walks again is exciting. I think it’s most definitely a good insight to have – even if most doctors intuitively know that high blood pressure in general is bad.

That’s why I think we need to dispel the current belief that machine learning will flip everything we already know on its head. This sets a lofty and unrealistic expectation for what it can really deliver on. Instead, we should realize that machine learning offers data-backed truths that we have been just taking for granted.