Defining AI...Or Not

AI is a hot topic. Like, flaming hot. But what does it mean?



Okay, so maybe the world of AI isn’t super clear-cut.

There are emerging ways to learn about this field, including:  a new major at Carnegie Mellon, Elon Musk’s Open AI, and the recently launched “Elements of AI” course being offered by a partnership between the University of Helsinki and Reaktor. I chose the third option.

This free course covers subject matter including: defining AI, neural networks & machine learning, the impact AI on our lives and livelihoods, and future implications of the emergence of artificial intelligence in society. The best part of this course, in addition to learning about AI, is that once you complete the course, you’re officially certified on LinkedIn—seems like a win-win to me!


The AI Learning Curve

When I began working with chatbots, awareness of AI came to the forefront of my mind, as I would conduct research to better understand the world of bots and emerging tech. Additionally, we began playing around with natural language processing (NLP) and seeing how we could leverage machine learning to make Instabots smarter and enhance our product.

While I have a basic understanding of AI, it can be somewhat daunting to distinguish what’s truly artificial intelligence, and what’s just fluff.

Here are my main takeaways:


What is ‘AI’? Defining the Undefinable
Eons ago, some would jokingly define AI as anything cool that computers couldn’t do. Straightforward, right? One glaring problem with this broad definition is that computer programming progresses fairly quickly, so a task that couldn’t be an automated command on last year’s processor is a smart feature on this summer’s model.

What then? What was once considered AI is no longer, and there’s no definition that can be applied across many fields. Due to this space constantly evolving, definitions are constantly being redefined, so creating a strict definition is somewhat futile. To help distinguish what is an AI method when we see it in products or practice, there are two characteristics of a true AI system. It is:

  1.  Autonomous, so it can perform tasks without constant guidance from users or administrators, and;
  2. Adaptive, so it learns from experience to improve performance.


False Expectations: Don’t Buy-In

Between TV, literature, and the press, AI is usually depicted within very unrealistic tropes (you can see our team’s favorite movie bots here). Moreover the press today is often over-trumpeting one start-up or another. This means that popular expectations often ≠ reality.

Hence, AII is that “words can be misleading”. When discussing AI responsibly, we have to watch out for ‘suitcase words’, which are general terms with multiple meanings that increase the risk of misinterpretation of capabilities of a product’s AI-ness.

When hearing the words, “AI”, ask yourself:

  1. Where am I getting my information from?, and

  2. Are there any incentives for using AI as described?


Talk About AI in Degrees, Not Absolutes

One of the best quotes from the class describes AI as: "a scientific discipline, like mathematics or biology." This means that AI is a collection of concepts, problems, and methods for solving them.

Because AI is a discipline, you shouldn't say “an AI“, just like we don't say “a biology“.

Lesson learned: Using 'AI' as a noun is generally incorrect, and should only used to describe varying degrees of “AI”ness, so as to avoid sounding like a total noob.

Then again, this is only lesson 1, so clearly there is much more to be learned. Join me as I take this course; I would love to start a discussion. Simply sign up here, and e-mail us at