The Truth About ChatGPT

The Advantages and Failings of the Fastest-Growing App Ever.

Yulia, my wife, called me into the bedroom, where I found her, laptop on her lap, smirking. 

“Read this and tell me if it sounds OK.” By the look on her face, I knew it had to be a disaster. 

It was. 

For a few months now, I’ve been telling Yulia (and you) about ChatGPT, the artificial intelligence engine that can generate a response to almost any inquiry, whether that’s “give me a recipe for guacamole” or “write a 2,000-word essay on the economy of 15th century France.” 

ChatGPT is all the rage these days. It is, in fact, the fastest-growing app in history. Uber took almost six years to reach 100 million users. Instagram took two and a half years. ChatGPT: just two months. 

Which means this is a good moment to discuss what ChatGPT can do… and what it can’t. 

Yulia has tried using ChatGPT to write content for her Etsy and Pinterest accounts, as well as for a letter to a school as part of an enrollment application for her son. 

It was that letter that she’d asked me to read. She’d gone through various iterations with ChatGPT, each time asking it to change the tone, or to include this, that, or something else. 

I read what ChatGPT ultimately coughed up. It had used words incorrectly, and conveyed a tone that read like an overzealous middle-schooler trying to sound like an erudite adult 

In short, a bit of a disaster. 

But, as they say: Garbage in, garbage out. 

That’s not to say what Yulia input was garbage. Just that what ChatGPT is based on is filled with bad data.  

ChatGPT is trained on a large amount of unstructured information found on the internet, including websites, books, and other text sources. And we’re all aware of how data found online can be a mixed bag, at best.  

I recently asked ChatGPT which countries in Africa offer citizenship by descent opportunities, and how many generations back one can go and still claim citizenship. 

ChatGPT listed every country on the African continent, and claimed that each one of them allows you to claim citizenship back to a grandparent. Totally wrong, which means the process for me as a writer seeking reliable data was but a time vampire. 

Which is really the challenge with AI at this early stage of the game: The technology itself isn’t quite ready for primetime. It’s still learning. It’s not yet very good at intuitively understanding what a user is really seeking.  

Still… 

The premise and promise of ChatGPT are pretty darn cool. 

By interacting with ChatGPT through “prompts,” you can find some pretty fun information that might be harder to find digging through websites. I tried: “Can you give me an authentic recipe for Irish Stew…from the 1600s.” And out pops this: 

During the 17th century, Ireland was going through significant historical events, including colonization and famine, which could have influenced the availability and usage of ingredients in traditional dishes [but] Based on historical accounts and common ingredients available during that time, a simple version of an Irish stew from the 1600s might have looked something like this… 

And then followed a recipe (which I haven’t tried, but I’m pretty sure I will). 

If you’ve ever seen the Joaquin Phoenix/Scarlett Johansson movie Her, you’ll have an understanding on why I think the premise of ChatGPT and AI are pretty darn cool. 

The idea that we’ll all soon have some kind of always-on, digital personal-assistant is crazy interesting to me.  

“Someone” who knows the answer to every question you have… a digital task rabbit that can make and track all your appointments with limited input from you… who can compose email replies simply by asking you a couple of questions… who can tell in the tone of your voice what you’re really trying to say… 

We will get there one day, hopefully soon. Just not yet.  

That’s not to say ChatGPT is pointless at this point… 

People are using the technology for all kinds of tasks and opportunities. I know people in the crypto space, for instance, who have built trading algorithms that are proving profitable. And innumerable people are using ChatGPT to write landing pages and blog posts for their affiliate-marketing-based websites. They insist traffic is up hugely. Maybe, but I have no way to verify the claims. 

Others are using it to design parts for various pieces of equipment that are then created in 3D printers, reducing the amount of design time necessary… some are writing complex formulas for Excel spreadsheets… others are using ChatGPT to create new games, to then write the coding for that game, and to even name the game. 

Lots of really amazing creativity put to work with the help of AI. 

Indeed, those examples point to the correct way to use ChatGPT right now. It’s a starting point… 

Each of those people (I surmise) used ChatGPT to get the work started, and then revised and checked and edited it themselves using their own expertise. 

It’s the same with Yulia’s letter. Yes, it was hilariously bad. But if you were pressed for time—and the document was not as important as a school application letter—ChatGPT could give you a framework to start from… so you’re not starting from a blank page. (To learn more about how to use ChatGPT correctly, join me for my upcoming Exclusive AI Briefing. Details here.) 

We’re still very early in this AI game. Actually, this is preseason still, when teams are just stretching their legs before the race for the title begins. 

The real advances are in front of us—the advances that see ChatGPT and AI in general morph from plaything into ubiquitous technology that all of us rely on. The same trajectory, basically, that the internet traveled over the last 30 years. 

I don’t think it will take 30 years for AI to achieve the same level of ubiquity.  

Technology moves far quicker these days… and we have AI that is helping developers develop, well, a better AI. So, the process will move exponentially faster. 

Probably within a decade, we’ll all have Scarlett Johansson at our beck and call.