General vs Narrow AI: Understanding The Difference


Artificial intelligence, or AI, is a word that’s tossed around a lot these days, whether it’s by computer scientists, industry leaders, or just your average person contemplating the “Sci” in “SciFi”. But when we’re discussing AI, what is it that we’re actually talking about?

That’s hard to pin down, especially since the answer has changed considerably over the years. In the 1950s, for example, AI referred to any program that accomplished something that would normally require human intelligence, which is a rather broad definition.

Getting to a more specific definition of AI has been a goal of computer science discussions for some time now, and as a result, we have a slightly better answer in 2020.

A typical computer program simply unfolds a sequence of predetermined events. Click here, something happens; input data here, output is created. You can think of it like any other type of machine, only digital. Where AI differs is in its ability to apply human behavior, such as planning, creativity, interpretation, perception, movement, and problem solving — all things you won’t find in software like Microsoft Word or Angry Birds.

In going through this definition of AI, you may have realized that it doesn’t quite match your own. That’s because there are many levels of sophistication in AI technology, which is where labels like narrow AI and general AI come into play. Below, we’ll breakdown the differences and similarities between the two to give you a better idea of the nuances of artificial intelligence.

General vs. Narrow AI: What’s The Difference?

General AI (AGI)

Artificial general intelligence refers to technology that can perform any intellectual task that a human could. In other words, if you can do it, so could an AI. That includes your job, making art, holding a conversation, coming up with an idea, telling a joke, etc. Artificial general intelligence is, so far, still stuck in the realm of science fiction.

The reason AGI is seemingly a long way off is that even the simplest human tasks, such as creating a knock-knock joke, require much more complex computing than you might imagine. The things we learn automatically as children are built on the foundation of the human brain, the most sophisticated living brain we currently have knowledge of. And as it turns out, replicating that is no small task.

Narrow AI (ANI)

You can think of artificial narrow intelligence (ANI) as the placeholder technology that we use until AGI is reached. And while “narrow AI” may still sound too futuristic to your ears, it’s something you likely interact with daily without realizing it.

Technology that uses ANI focuses on a small subset of human cognitive abilities. For example, when you ask Siri to list local movie showtimes, she listens to what you say, decides how to act on what you say, and then delivers a response.

However, Siri doesn’t actually understand you. She understands “list” and “showtimes” and is able to link those keywords to an ability that has been programmed into her. It’s not much different than a light sensor activating when it detects light. It’s using traditional programming and a bit of machine learning to mimic artificial intelligence, and although Siri and her competitors are getting better all the time, you still can’t hold a coherent conversation with them about the gum on the bottom of your shoe.

What About Super AI?

When you start to feel that existential anxiety at the thought of artificial intelligence, you’re most likely thinking of Super AI. Super AI is as dramatic as its name; it’s artificial intelligence that, in theory, will have surpassed our own intelligence and essentially will be a new race of being. This is certainly not something to take lightly. But let’s keep things in perspective. Super AI, as well as AGI, is something that we as human beings don’t really have the ability to create on our own. It requires a deep understanding of not only our brains but also our consciousness. Being able to replicate that is such a Herculean task that it’s actually being offloaded onto existing AI technology.

In other words, it’s certainly not impossible that we’ll eventually create ASI, but it’s not a concern that’s expected to come up for several decades at least, possibly centuries. Additionally, as humans, we don’t really have the capacity to understand ASI, and as such, it’s not worth worrying about unless you’re a high-level computer scientist.

The Road To AGI

Artificial general intelligence is still decades away at least, but we’re closer than ever to realizing AGI than ever before, as evidenced with advances in fields such as generative AI. This is thanks to the development of areas within computer science, particularly machine learning and deep learning. As these areas continue to be understood and advanced, the timeline for true artificial intelligence will grow shorter and shorter.


Steve Hulet


Steve is the Co-Founder and CTO at Fresh. A former Software Engineer at Amazon with over 12 years of web development experience, Steve provides technical, architectural, and engineering oversight to projects. Steve is responsible for all technology reviews related to websites. His specialities include programming languages such as C, C++, Java, Python, and PHP, and technology software including Eclipse, GLPK, jQuery, Linux, and MATLAB. Steve’s skills include automation, databases, linear programming, optimization, and testing, all of which he uses in conjunction with Fresh’s digital strategists provide innovative solutions to clients.