Not All Versions of Artificial Intelligence are the Same
By Peter Lyle DeHaan, PhD
There are two types of AI (artificial intelligence). The most recent version of AI is generative AI. Contrast this to predictive AI, also called regular AI, non-generative AI, or analytical AI.
For simplicity, we’ll refer to all these alternate labels as predictive AI, simply because it’s the term more regularly used.
Though both types of AI rely on machine learning, they have significant differences. Here’s a basic breakdown between the two types of AI: predictive AI and generative AI.
Predictive AI
The older and more accepted forms of artificial intelligence fall into the camp of predictive AI. As the name suggests, predictive AI predicts responses to what we are doing. It taps a vast database of past results to project future paths.
Some common examples are auto complete when typing messages or doing online searches. We begin typing and predictive AI suggests what it deems as our most likely word or phrase to complete what we’ve already entered.
Auto correct—or auto spell check—is another common implementation of predictive AI. Predictive AI can also fill out forms for us or suggest more common phrasing as we type messages.
When appropriately implemented, predictive AI can help telephone answering service staff improve their accuracy and increase their productivity.
Yet to be successful, staff must have the ability to override artificial intelligence’s predictions, be trained on what to look for, and have the confidence to act.
Generative AI
In contrast to predictive AI, we have generative AI. Generative AI is also based on machine learning and has vast databases to tap into.
Yet a distinct difference is that while predictive AI bases its recommendations on past reality, generative AI produces new content deduced from what it knows.
That is, it generates answers. The quality of what it generates resides on a continuum ranging from highly intuitive to pointedly absurd. In short, it’s made-up answers can, at times, be laughable and unrealistic, which any person can readily determine but a computer cannot
What becomes even more worrisome with generative AI is when it uses prior AI generated results as part of its base dataset. This works well when the prior generated content as accurate, but errors quickly compound when it taps faulty conclusions.
Generative AI, by the way, is the basis for science fiction apocalypses, where computers determine that people are a problem and sets about to eliminate them.
Though this possible outcome should cause us to proceed with care before we implement generative AI into our answering services, this is not to suggest a total rejection.
Though we may not be ready to trust our business to a generative AI tool, we may one day see the potential to tap it for basic applications. If—or when—that day comes, be sure to proceed with care.
Artificial Intelligence Conclusion
We should be aware that not all AI is the same, with predictive AI and generative AI as being distinctly different.
Before we consider implementing either of them in our answering service, we should be aware of their strengths and their weaknesses, of the rewards and the risks.
Learn more in Peter Lyle DeHaan’s book, How to Start a Telephone Answering Service.
Peter Lyle DeHaan, PhD, is the publisher and editor-in-chief of TAS Trader, covering the telephone answering service industry. Check out his books How to Start a Telephone Answering Service and Sticky Customer Service.