AI Is Having It's 90's Internet Moment - Let's Talk About It

These days it seems AI is all the rage, especially for those of us who work in tech, or tech related fields, but even if you don’t have a job that has put you into contact with this new technology the chances are you’ve heard of it and if you haven’t already, will come into contact with it at some point in the near future.  

Not since the internet took the world by storm in the 90’s have we seen something that has been so disruptive and caused such a profound amount of change in the way daily work is done.  

Some people love it. Some people hate it. Most people have opinions on it but with these opinions also come lots of misconceptions.  

AI might not be as bad as you think. Sometimes it's worse, and sometimes it's a truly fantastic tool that can make daily tasks much more efficient.  

There is a huge moral debate at the center of this topic. On a large-scale AI will be as ethical as the humans who create it, but that’s a topic of deliberation for another day. On a small-scale AI is only as good (competent), or as bad as the person using it. In this article, I want to talk about the pros and cons of using AI tools in the workplace. 

The internet can be a bad place. We’ve simply come to accept that as a fact of life. People use all sorts of technology to do awful things like steal identities, and I don’t even want to go into the dark web. But I have yet to hear people suggest that it should be shut down entirely, although the same can’t be said about AI. When I hear people complain about it (which is very often) most of the time they have a very black and white opinion on it. AI is bad and stupid. It’s incompetent and just not worth the hype or taking the time to learn about it. The world would be a better place if all AI tools were wiped off the face of the planet. 

Now in many cases I agree that the hype is overselling the actual product. I have spent the past eighteen months using the most popular AI tools and many of them frequently hallucinate. For those of you not familiar with the term “...it means the system generates information that sounds plausible but is factually incorrect, made-up, or unsupported by real data.” And yes, I did pull that definition off of ChatGPT. This is hugely problematic, especially if you have people who don’t check their work. That’s when you start seeing things like the list of recommended books for Summer printed by the Chicago Sun-Times. This list was clearly created by an AI tool with limited or no human oversight, because most of the books on the list didn’t exist.  

Human oversight of any AI tool is critical for producing good, ethical content. In other words, just as you should do with any work, before you turn it in or make it public, take the time to review it. Understand the limits of the tool you are working with and have it perform job-appropriate tasks. 

With good prompting techniques, and a good framework for human oversight and content review, the AI tools I use have brought great speed and efficiency to my work and the team that I work with. 

By far the most popular and well-known AI tool is Open AI’s Chat GPT. Launched in 2022 Chat GPT took the world by storm. Over 400 million people worldwide use this tool but very few people that I have come across actually understand what it is so let's break it down by asking the question: What is Chat GPT? 

In essence, Chat GPT is a predictive model. Its main function is to predict the next word in a sentence based off of the extremely large volumes of data it has been trained on.  

Google’s version of this is Gemini, and Microsoft has Co-pilot. They all have different strengths and weaknesses.  

In my use cases I have found that Co-Pilot is much better at math than Chat GPT, which when you give it multi-step calculations is likely to need at least two tries to get it right. However, Chat GPT and Google Gemini are both better for generating written content than Co-Pilot. That’s not to say that Co-Pilot is bad, it's just not as advanced in this area. Different models have different strengths based on how they are trained. If you want to test this feed identical prompts into two LLM’s and compare the results. 

But back to my earlier point, that AI is only as good as the person using it. So, you might be asking yourself, what do I need to know to get good at using AI?  

The first thing to learn is how to prompt.  You can interact with these language models like you are talking to a person. Strong communication skills are actually essential even though you are communicating with technology.  

What is a prompt? Good question. A prompt are the instructions you give to AI tools; The input you provide for the system so it can output what you want.  

Imagine you are creating flyers for an upcoming event you are hosting. You are working with a graphic designer to create the posters you want, and you have a very specific vision. For the graphic designer to make the type of poster you want, you would have to explain to them how you want the colors, and the font to look, what the copy should say and what information must be conveyed on these posters. You must also tell them what type of event the poster is for.  

If you try to do this in a rush, if you don’t take the time to give the designer proper instructions, and you can’t be bothered to check in with them throughout the creative process as they iterate, your designer will come up with something but they can’t pull the idea straight out of your head so the chances are the final product will not entirely reflect your original vision. Maybe you’ll like it, and maybe you won’t. Maybe the right information will be printed on the flyer but it’s unlikely.  In contrast, a person who is willing to take the time to write out clear instructions for the designer and checks back on their progress as they create the poster is much more likely to get a good result that they are happy with because they took the time to provide oversight and critical feedback. 

The same is true for these predictive models. The most glaring obvious difference is the speed of the work. AI will return results within seconds. Now it's your job as the human to have the patience to look over them and revise as necessary.