At Resultant, there are certain questions we get asked often. Some of the most popular relate to the use of artificial intelligence (AI) and machine learning. It is often marketed as a “magic bullet” for all of a company’s predictive woes.
In the years that I have worked in the statistics realm (both as a professor and now as a consultant), I often get asked, “Can you build us a model?” I have only been asked once, “Can you build us a good model that will produce replicable results and have consistent forecasting?”
We get so worried about getting a model that we rarely think to ask, “Is the model any good?” AI (and Machine Learning for that matter) will always be able to produce a model. Now whether or not the model is any good requires both knowledge and experience. Couple this with the fact that models change over time and the magic bullet looks more like a scattershot!
Why AI Is Neither
In the same way that the “cloud” is just someone else’s computer, AI is really just someone else’s program. Sometimes the term AI will conjure thoughts of Will Smith’s and/or Linda Hamilton’s attempt to thwart a self-aware piece of code as it attempts to overthrow humankind. It’s not quite like that. AI is really just a piece of software that continues to maximize the probability or probabilities of achieving a set of goals. That’s it. If you would like an evening of fun reading around this, just Google the Church-Turing Thesis.
Why is it not artificial?
AI really comes down to code that someone had to write. And the strength of the AI really has to do with the skill of the person who wrote the code. Sometimes that code centers around optimization procedures. Sometimes around logic functions. Regardless, we could have two different AI systems that approach a problem in two entirely different ways because they are coded by two different research teams. Again, it all comes down to code.
Some of the better AI systems and agents are ones in which the AI is able to augment and change code based on feedback from the system. Take, for example, the AI engine behind self-driving cars. The more advanced systems are constantly changing and adapting. If the system encounters a bicycle that is going 80 mph down the highway, then a strong AI might adapt rules that it has for bicycle behavior because this bicycle is 6’ off of the ground (it was on top of a car).
So, we might say that AI is artificial in the sense that it does not persist in someone’s brain. However, it is not artificial in that it wasn’t created and adapted by human interaction.
Why is it not intelligent?
Can it think? Is it self-aware? Does it have morals?
Although machines do not have the capacity to make moral decisions, they can make decisions that have moral ramifications. This is why many industries who deploy AI models have started AI Ethics Boards.
Consider again the self-driving car. Suppose that the car is cruising at 70 mph and tops a hill where there is a bad accident in an area where there is no time to avoid additional incidents. What should the AI do? Should it just apply the breaks and stay in the lane? Should it attempt to steer around into oncoming traffic? Should it veer off of the mountain edge, certainly killing the driver but saving everyone else in the accident?
Questions like this are why we will always need intelligent decision-makers who can monitor the AI system that is operating on a series of swiftly calculated probabilities. The intelligent will always be needed to instruct and govern AI.
The bottom line
AI is not a substitution for business knowledge. It can be a great augmentation to the effectiveness of predictable outcomes, but it will always remain an amalgamation of business acumen, coding skill, and deployment monitoring. AI models without truly intelligent interaction and development are akin to taking medicine without proper diagnosis. Eventually, you might get lucky and heal . . . but you may also destroy your liver.
If you want these insights and want to understand your business’s data better, get in touch with us. We’re at the forefront of providing innovative solutions that transform business data into insights that deliver better results.