Understanding the Three Different Types of AI

You are here

The image above was generated from DALL-E. Mike Watson calls this “The Three Waves of AI.”

AI is here to stay.  But people are using the term to describe three different aspects of AI.

I’ve found it helpful to understand these differences.  It allows you to do the right research, select the right technologies, and understand the potential benefits and dangers. When people say “AI”, here are the three different things it could mean:

One, Artificial General Intelligence (AGI).  In this use, people are using AI to refer to getting machines to think and reason just like (or better than) humans.  That is, the machines would have general intelligence.

In 2012, Deep Learning (deep neural networks) made a big splash by being able to recognize images better than ever before and at the level of a human.

Deep Learning works the way we think the brain works.  So, this 2012 news made people think that we might have found the right tool to replicate general intelligence.  This has led to a lot of research and fantastic advances in Deep Learning.

Although it is not general intelligence, self-driving cars are often lumped into this bracket because it is such a hard problem.

Researchers like Ray Kurzweil claim that we are near the Singularity, where machines become smarter than humans and then can teach themselves to develop super intelligence.  (Many other researchers disagree with Kurzweil about the Singularity.)  Serious researchers have also started discussions on what consciousness is and when we will know if a machine has it.

Unless you are doing basic research (or working on self-driving cars), the area of AGI doesn’t (for now) impact everyday life and most organizations.  And, if you ask my opinion, I side with the experts who are skeptical about our progress here.  But, I think the research is useful.  It has led to innovations that help the other two types of AI.

Two, Generative AI.  This is the newest type of AI, hitting the popular imagination in late 2022.  This is the use of AI for making creative content.

You can create images from text (DALL-E and Stable Diffusion) or write interesting responses to prompts (OpenAI’s ChatGPT).  And you get a flavor of what is coming with podcast.ai’s AI-generated interview between Joe Rogan and Steve Jobs.

These solutions came directly from the research around AGI. Although these might be steps towards general intelligence, I think it is better to think of the disruptive power of these solutions.

If you write or create images, you will need to understand this to stay ahead.  For example, teachers and professors now need to worry about students using these tools to answer homework questions.

These tools will also create new opportunities. I’ve heard of new tools for making images of your product look better, helping you write blog posts, and allowing you to upload images of a room and ask for different design ideas.  This is just the start.  There will be many interesting solutions from this breakthrough.  Of course, you also have to watch for and build safeguards for potentially toxic language or inappropriate images that can come from these tools.

Three, Practical AI.  This is the use of algorithms to solve business problems.

In some sense, this is the most boring of the three types of AI.  Businesses and organizations have been using algorithms for a long time.  The term Practical AI grew out of what was called Big Data or Analytics from around 2006-2017.  (By 2017, AI was the more popular term.)

However, there are a few things worth pointing out.

First, since the above two definitions get the most attention,  most leaders think that AI is not for them. General intelligence is a research topic, not for the day-to-day activities of an organization.   And most organizations have only small, if any, use for Generative AI.

Second, leaders should think of Practical AI as using all the algorithms (not just Deep Learning). This includes the algorithm coming out of AGI, but also includes machine learning algorithms, mathematical optimization, and other techniques. Leaders should consider Practical AI an umbrella term for all the sophisticated algorithms available.   You should use the best tools for your problems and opportunities.

Third, to embrace Practical AI, you should have a mindset of thinking broadly, creatively, and critically.  You can look at every aspect of your organization and ask how we can use algorithms to improve this part (or the whole) of the organization.   This approach is captured very well in the first five chapters of the book Competing in the Age of AI. However, you also need to be critical.  The same algorithms that can bring benefit, can potentially make decisions that are illegal (you especially need to be careful when used on individuals) or harmful (when biased data is used to train the algorithms). 

The sustained hype and potential applications for the three types of AI mean that it isn’t going anywhere.  You will need to understand both the benefits and what can go wrong to make good use of this technology.  Hopefully this framework helps you think through those issues.

Article by Mike Watson. Mike is an adjunct professor at Northwestern University, the author of several books and a blog, and was the co-founder of Opex Analytics (now part of Coupa Software). You can find him on LinkedIn at https://www.linkedin.com/in/michael-watson-07600a1