AI lab – AI in Action | Episode 02: AI Terminology

Let’s talk about AI terminology in the second episode in our AI in Action series.

The AI term gets thrown around more than a beach ball at a summer picnic, and it’s not always clear what people are talking about. I mean, “AI” is to tech what “food” is to a grocery store – sure, it covers a lot, but a hot dog and a filet mignon are pretty darn different when it comes to what they do to your insides.

Now, when people think of AI, they either picture robots taking over the world (Terminator, anyone?) or high schoolers using it to cheat on their calculus homework. But the reality is more like your phone autocorrecting “ducking” into something your grandma wouldn’t be caught dead saying.

AI is a layered beast, like a high-tech set of Russian nesting dolls. You crack open the biggest one, and bam! There’s another one inside.  

Let’s start with the big kahuna, the big momma of them all: Artificial Intelligence.  This is the broadest category, encompassing any trick that lets computers mimic human smarts – logic, decision trees, if-then statements, the whole toolbox.  It’s about creating software or machines that can perceive, reason, learn, and act like, well, us humans. This includes everything from your spam filter to that Roomba that terrorizes your cat.  It’s AI in its most basic form, quietly humming in the background, making life slightly less annoying.

Now, open that doll, and you’ll find Machine Learning.  This is a subset of AI that doesn’t just follow orders, it learns from data.  Think of it as the power tools in the AI toolbox. Except these specific tools can automatically learn and improve without needing constant hand-holding. Yup, just imagine a power drill that does its thing on its own.

Machine Learning focuses on algorithms that can analyze data and make decisions or predictions based on what they find. It’s like a teenager who, after burning toast 87 times, finally figures out the right setting on the toaster. Except, this teenager can also predict stock market trends, which is either impressive or terrifying, depending on your feelings about teenagers and the stock market.

Next up is Deep Learning. Think of it as Machine Learning on steroids.  It uses these things called neural networks with tons of layers, hence the “deep” part. These models are really good at learning from mountains of unstructured data. Imagine an orchestra where each section plays based on what they hear from the previous one, building a complex symphony piece by piece. Deep learning works in a similar way, orchestrating layers of data processing to uncover patterns and insights that simpler models would miss.  It’s a beautiful harmony of data, revealing hidden knowledge like a maestro producing the nuances out of a symphony.

But wait, what about all these other buzzwords floating around the AI world – large language models, natural language processing or generative AI? Well, those don’t fit neatly into our nesting doll analogy. Life’s rarely that simple, I guess. These concepts are more like grains of sand that sprinkle through all the layers.

But let’s take a sneak peek anyway. 

One term that’s getting a lot of attention is Large Language Models, or LLMs for short. These are advanced computer programs trained on massive amounts of data. This data can include books, articles, code, and even social media conversations. These LLMs reside within the Machine Learning branch of AI, and Natural Language Processing or NLP plays a crucial role in their development, as a learning technique that focuses on how computers can understand and interact with human language. NLP techniques are used to train LLMs on the complexities of language, including grammar, syntax, and semantics. Think of NLP as the foundation that allows LLMs to make sense of the vast amount of text data they’re fed and that often though not always relies on Machine Learning or Deep Learning. And then, think of LLMs as a specialized application built on top of Machine Learning or Deep Learning, rather than a specific subset. 

Confused yet?

Well, I am afraid there’s more jargon to come.

Because no one can speak about AI today without addressing the elephant in the room, namely generative AI.

This is a specialized application within AI that often uses Machine Learning techniques, including Deep Learning, but isn’t limited to them. Think of AI as a band, and generative AI is the lead singer – the Freddie Mercury, if you will. Yes, the groupies and media tend to go for the rock star but you can’t have a band without a drummer and guitarist!

With generative AI, it’s not just about understanding, analyzing, or predicting; it’s about creating entirely new data based on the information it’s fed. That’s what sets it apart from its AI cousins. And it’s not limited to text: it can generate images, music but also 3D models based on existing data sets.

Now, imagine the whole AI family gathered for a picture.  How can we differentiate them? Well, one way to look at them is by categorizing them according to the purpose they serve. A bit like showing on a family picture who’s the party animal and who’s the musical prodigy.

First up,  we’ve got Descriptive AI, the librarian of the bunch. It’s all about organization, meticulously cataloging the world around us. This is AI in its simplest form, taking raw data and turning it into clear summaries of what’s happening or what happened. Think of it as the CCTV of data, constantly recording and summarizing everything from traffic patterns to your late-night snack habits based on your fridge logs. It’s the one that compiles those “Year in Review” summaries that remind you just how much takeout you ordered last July.

Next comes Diagnostic AI, the detective with a magnifying glass. It doesn’t just show what’s happening, it figures out why. Think of it as the Dr. House of AI, but without the questionable bedside manner. When your car’s “check engine” light comes on, Diagnostic AI is what your mechanic uses to tell you it’s because you keep treating speed bumps like ramps.  It’s analytical, it’s clever, and thankfully, it doesn’t puff on a pipe (unless you program it to).

Then there’s Predictive AI, the fortune teller of the AI world.  It stares into its crystal ball of past data to predict future outcomes. Except instead of reading tea leaves, it analyzes your Spotify playlists to eerily predict that yes, you probably do need another breakup song. Or take the financial markets, where Predictive AI is like that friend who swears they knew the stocks would tank – except this friend sometimes actually gets it right, sort of like a weatherman predicting rain in London or in Brussels. 

And finally, you have ‘Prescriptive AI,’ that suggests actions you could take to achieve a goal or solve a problem. It’s your bossy digital GPS, offering not just insights into where you are but advice on where to go. It’s one step ahead of predictive AI, not just forecasting what might happen, like a traffic jam at rush hour, but also telling you to take the next right to avoid spending half your evening cursing at brake lights. This is the AI in fitness apps that nudges you towards more walks and less junk food, or in smart thermostats that learn your schedule and adjust the temperature for comfort without exploding your energy bill.

So now that we’ve met some key members of the wacky AI family, let’s get back to generative AI.  Most people hear that term and think of things like ChatGPT, which can churn out anything from a sonnet about your goldfish to a disturbingly realistic resignation letter for a job you don’t have. Or MidJourney, that app that can generate endless variations of John Oliver interacting with… cabbage. On that one actually, virtual met reality when the whole thing ended with a wedding!

Look, it’s easy to poke fun at AI writing our essays or creating art for us, but generative AI is much bigger than just the latest text and image generators.  It’s about using the vast chaos of existing data to create entirely new information, ideas, even whole new worlds. It has the potential to transform industries, boost research, supercharge productivity, and in some cases tackle complex social issues like climate change. Understanding its full capabilities, benefits, and challenges is crucial for policymakers and industry leaders alike if we want to harness its potential responsibly.

This isn’t just about making our lives easier:  it’s about reimagining what’s possible. From the mundane to the monumental, AI is reshaping our world, one algorithm at a time. And that’s exactly what we’ll be diving deeper into in the next episodes of AI in Action.  

Until then, stay curious, stay informed, and maybe, just maybe, let that Prescriptive AI suggest a route that avoids any more existential dread than absolutely necessary. Who knows, you might even end up somewhere unexpected, and that is where the real magic happens.