Machine learning and data based technologies are often wrongly referred to as artificial intelligence. There is nothing intelligence about them. And they’re not artificial either.
The field of machine learning is based on the premise that repetitiveness in data can yield patterns, those patterns can be transformed into models and those models can extract assumptions.
Colleagues would sometime refer to such assumptions as predictions. I strongly disagree with that notion, mostly as it discounts the vast complexity of the brain, and the redundancy analysis we deploy when we make actual predictions. The sort that validate me to catch a ball flying over to me, or give me the confidence to interrupt a person I am speaking to, with no risk of awkwardness or abruptness.
Data technologies are – for the time being – conditioned on high volumes of date. Some of the time it is unstructured. It is the tension between neatly organized databases, like the one in Excel spreadsheets, and more sporadic set, think your Instagram feed, that makes these technologies useful. The ability to programmatically tag, slice and categorize at scale in incredibly useful, but it is not intelligent.
Imagine I asked 10 copy artists to hand–draw the Mona Lisa, and then have those 10 newly made artworks on the table. It would be difficult for me to grade those, if those artists were good at copying art.
Now consider a scenario where I would line up these artworks perfectly one on top of the other. Better yet, those artworks are now on transparent paper, with the last layer being the original piece.
Anomalies are going to be apparent, and very easy to recognize.
This is, in a nutshell how machine learning works.
It lines data up, to recognize anomalies, or patterns.
This scenario also highlights why there is nothing artificial, nor intelligent about it. The Mona Lisa copies were made my humans (users), not the machine. As were your Instagram photos, or tweets, or mileage on your car.
You can also see how it is hard to hold belief of the predictive nature of my lined up Mona Lisa’s.
There is nothing future facing about them, which is after all the prime condition of all predictions. Stating the obvious, predictions are forecasts of the future. All I can do is understand the past, in which 10, 100 or 1000 copy–artists tried to copy the painting.
The main friction point between new technologies and large scale innovation is public perception. I am inclined to adapt the view that we’ll do machines, and ourselves a favor by referring to Machine Learning, AI and similar technologies simply as really smart programming.