Simply put, the difference between AI and generative AI is this: artificial intelligence is the umbrella category for all forms of machinery with human-like intelligence, while generative AI is a subset of this category referring to intelligent machines that can produce something new.
In this article, we'll explain generative AI and touch on three other common types of AI, giving you just enough information to understand the basics without feeling like you need a master’s degree in AI development. This list is not exhaustive of all forms of AI, and it doesn’t delve into every single particularity of each type, but it will give you the takeaways you need.
What it does: Generative AI creates something new from things it’s learned via training. Its output could be code, it could be a recipe, it could be an image, etc. (Again, you'll see here that AI is different from generative AI because it's a broad category for all types of AI, while generative AI is limited to this definition.)
Example: You’ve probably encountered the concept of generative AI via ChatGPT—that is, unless you’ve been living out in the wilderness for the past six months (in which case, welcome back!). ChatGPT is a tool that can both converse with you like a human and provide you with an original response to a prompt. Although, you could argue the outputs ChatGPT generates are not really “original,” because it has garnered its knowledge from sources across the internet like Reddit and Wikipedia and is merely predicting which next word is most likely to be the correct choice based on what it has learned. (Maybe that’s what we do, too, and we just don’t realize it? Tell that to Shakespeare and see how it makes him feel.)
What it does: Predictive AI is when a machine can make predictions based on a combination of previous inputs and an analysis of current trends and scenarios. This type of AI is already widely used in the business world.
Example: Predictive AI can be used for programmatic ad buying, among other things. In this example, based on historical knowledge of price points and ad performance, an algorithm can predict when a company should buy ad space to get the best rate. You might also encounter predictive AI in the stock market in the form of high frequency trading (HFTs). These AI models use algorithms to make trades at high volumes based on predictive analytics.
What it does: Anomaly-based AI detects abnormalities in a pattern. This type of AI is trained to recognize regularities, so it can also detect any time an exception pops up.
Example: This type of AI is particularly useful in cybersecurity. Your anomaly-based AI can learn what type of activity is usual on your network. If it detects activity outside of that pattern, it can trigger an alert to help your team respond in real time. Anomaly-based AI also serves a useful purpose in supply chain management by learning normal demand patterns from purchase data. If demand spikes above or below that range, the AI can detect this and notify your team to adjust prices accordingly or contact suppliers.
What it does: Decision-based AI helps make decisions similarly to how a human might, such as classifying things based on their characteristics.
Example: Let’s say you have four different product types, and you want to classify customer support emails into the product category they belong in. By giving your decision-based AI model some sample emails to learn from, you can train your AI model to identify which email is related to which product, and then route each one to the appropriate department. (This is really possible with the Appian AI Skill Designer—check out this webinar showing how it works!)
As you can see, AI is a vast field that can be broken up into many different categories, including generative AI. And this is just the beginning. Companies are at the start of an AI arms race. To learn more about AI's many uses for business, check out our research with eight AI experts in the 2024 AI Outlook.