AI Demystified: A Simple Guide to What It Can and Can’t Do

Anyone who remembers the beginning of the internet back in the mid-90's may recall two types of people: those who embraced it and those who thought it was a passing phase. The same can said of AI (Artificial Intelligence). Platforms like Chat GPT, Gemini, and Claude are springing up like digital daisies and much like Netscape or AltaVista were back in the day, some will evolve and some will wither away.

Just like the early gems of the internet, AI is remarkable, sometimes scary, technology. It takes massive amounts of data (we’re talking trillions) and finds hidden patterns to make decisions. This is called 'machine learning'—where the system learns from examples much like our brains do as we experience life’s challenges. Aside from the frightening philosophical debates about AI, the technology can enhance and even automate tasks that you would have thought impossible only a few short years ago.

Understanding the AI Framework

From a memo to co-workers to a PowerPoint that outlines next year’s budget projections, AI is extremely versatile and can generate everything from intricate code to Aunt Marge’s 80th birthday invitation. Much like the human brain, however, AI is complex but far from perfect.

Your relationship with AI begins with understanding where it can be useful, where it may not be, and how it can help your daily work life. This begins by establishing an AI framework. Ask yourself these simple questions:

1. Is there a task that only you can do? Is there a back history that only you know about? An example could be helping a co-worker navigate difficulty with a mutual supervisor. AI doesn’t equate inter-personal relationships and back histories and  something like this will probably be an exercise in futility.

2. Is there a task that AI can help you with? This is a task that you draft on your own but use AI to refine your work or organize your thoughts. An example could be drafting a  memo on improving office efficiency.

3. Is there a task that AI can automate for you? Here’s where AI can be a real time-saver as these are structured, repeatable tasks that AI can handle efficiently. Remember, AI is based on machine learning and will remember past requests. An example of this could be reproducing an existing excel file with new contact names.

It’s important to know how AI can work for you, but remember, AI is a machine. It has no feelings; it does not think. AI uses highly sophisticated algorithms to find patterns in words, images and code then scrapes and compounds the data with remarkable speed.

Avoiding AI Slop (Yes, That’s Really a Thing)

As powerful as AI is, these platforms still rely on human judgment. It is far from becoming a self-sustaining, do-everything for you, technology. You are in control of AI, not the other way around—take into consideration when and how to use it effectively. What is commonly known as 'AI slop’ is what enthusiastic users input every day and expect AI to generate things it can’t. Much like life, with AI, you generally get back what you put in.

Say you need to draft a letter to a patient who is behind on their payments. Does AI know the backstory? Is the patient going through a divorce or recently lost her job? Does it know the relationship you have built with that patient over the years? No. You have to be the one to make that judgment.

AI Slop Prompt:

Draft a letter for a patient who owes us $300.

Here’s what an AI Slop prompt might produce:

Dear [Patient Name],

Our records indicate that your account currently has an outstanding balance of $300 for services provided by our office. We kindly request that payment be made at your earliest convenience.

To the point and efficient, right? Not really.

A better way to word this prompt would be:

Draft a letter for a patient who owes us $300 but has recently gone through a divorce. She is in her 40s and has two teenagers.

Here’s what AI generates:

Dear [Patient Name],

Our records indicate that there is an outstanding balance on your account of $300. While we do need to collect payment for services provided, we understand that situations can arise that make paying a balance all at once difficult and we’re happy to work out a payment schedule with you.

These two letters differ vastly because letter one is AI slop. Letter two is not. When you enter the prompt, it is vital that you include as much information as you can. AI is a machine that will aggregate based on what you input. Always include the audience, the circumstances, and as much pertinent information as you can, while omitting  confidential information. Simply put, you have to give AI the backstory because it won’t know otherwise.

Another example, entering: run a projection of next year’s earnings does not give the backstory. A better prompt would be: run a projection of next year’s earnings based on our last year’s projections but take into account the inventory of frames and average number of patients per week and will give you a better result. The more you tell AI, the better it can learn.

In a nutshell, using AI effectively is understanding the tasks where it doesn’t belong, tasks that it can enhance, and tasks where it can automate something that used to take too long to do.

While this blog was not written with AI, it certainly could have been if the author gave enough information—Write a blog that explains how to use AI effectively for people in the eye care industry. Sure, it could write it, but it will be based on the data it already has. It wouldn’t be unique.

AI is indeed remarkable technology, but it cannot replace the power of the human brain. Our brains have genuine self-awareness, feelings, and years of lived experiences—and that is the most powerful tool of all.

Written by M.A. Giorgio