I think there is a lot of confusion about the definition of Artificial Intelligence (AI). In my mind, the confusion comes from the fact that people use the term “AI” in two ways.
The first way, AI is referring to machines and algorithms thinking and reasoning like or better than people. This definition is the one that is most often used by popular media or Hollywood. But, behind all that hype, there are many serious researchers who are doing interesting work in this area. The serious researchers tend to describe this area as Artificial General Intelligence (AGI).
With AGI, it is reasonable to ask questions like “what does consciousness mean and will a machine have it?” and “should we worry about machines getting smarter than us and destroying us?”
There are a few skeptics that don’t think we are really that close to AGI. For example, Melanie Mitchell’s book, “Artificial Intelligence” and Rodney Brook’s blog provide the skeptical view.
If you look at how business leaders and the business press use the term AI, you can see that it doesn’t involve general intelligence. However, it does include sophisticated algorithms solving very specific problems. For example, it could include translating text, routing trucks, finding fraud, or predicting when a customer will churn.
I would call this type of AI as Practical AI. Practical AI is the use of sophisticated algorithms, data, and computing power to solve specific problems. These solutions may be very complex and easily outperform a human at that task, but no one would confuse these solution for general intelligence. They just do their task and nothing else.
With this definition, Practical AI is an umbrella term for a wide range of algorithms and techniques. These algorithms would include Deep Learning, Reinforcement Learning, Machine Learning, Optimization (like Linear and Mixed Integer), Statistics, Simulation, and much more.
I’ve found it helpful to make this distinction between AGI and Practical AI. It helps facilitate a better discussion— if one person is using AI to really mean AGI and the other person really talking about Practical AI, they will be talking past each other.
I should point out that not everyone will agree with my definition of Practical AI.
I just listened to a great podcast interview with Radhika Kulkarni. She is the retired VP, Advanced Analytics R&D at SAS Institute. While there, she was using all these techniques long before there was any hype around it. She is now the President of INFORMS.
As the INFORMS president, she like the term “Analytics” as the umbrella term. Her use of this term would also include the important area of Descriptive Analytics (where Practical AI likely doesn’t).
She has a great way to define the mission of INFORMS, which I’ll be sure to adapt when I talk about Practical AI:
“…the science and technology of decision-making to save lives, save money, and solve problems.”
I’ll write more about the important role that INFORMS is doing in both the AGI and Practical AI (or Analytics) movements.
I’ve been talking this definition for a long time. If you are interested, here are some other articles I’ve written on this topic. These articles provide more background and examples.
Here is an article I published in the INFORMS OR/MS Today magazine arguing that INFORMS members should feel free to use the term AI (or Practical AI) to describe what we do.
Here is a blog post I wrote while at Coupa on the definition.
And, going back a few years, this a paper we put together at Opex Analytics back in 2018 when we rebranded what we did as AI.
I like the term. It describes what we do at hypaIQ.com
Nice way of putting things in perspective, Mike.