Limitation of ChatGPT
As you might have noticed, chat Gpt is an incredibly powerful AI chatbot. I would even go as far as calling it a superpower. But as we know from every comic ever, every superpower comes with a limitation.
Superman can almost do anything, but there is kryptonite. we’re going to talk about Chat GPT’s Kryptonite because in order for you to get the maximum out of it, you need to be aware of the limitations that are very real.
One Limitation of ChatGPT
Okay? So the probably most important one is the fact that Chat GPT is nothing but a language model trained by a data set.
And that data set is not infinite, it’s a limited data set. So it absolutely does not know everything. Most notably, as even the interface says, limited knowledge of world and events after 2021. Limitation of ChatGPT
And as the model is not connected to the Internet, this version of Chat GPT only gives you events that happened before 2022. Furthermore, it has certain templates that it has built into the model. But let me give you a concrete example of that.
If you say a write me sales copy from our website selling dog food. So, and if we look at the structure of this answer, you’ll see he first points out a pain point, and then immediately goes to the solution that you’re trying to sell here, points out advantages of your product, and ends on a call to action.
And when I tell it, write me a sales email for our enterprise Cybersecurity software and have a look at the structure, it points out a pain point.
It states that your product is a solution to this pain point, states some advantages, and ends on a call to action. And even if you tell it write me ten more, it always follows the structure. Limitation of ChatGPT
And hey, it’s a good structure, but definitely not the only one. And it might not be the most effective in every scenario.
The best way to learn about these templates is whenever you run a prompt, always tell it, give me three more, and that is going to help you, the human using this chatbot, identify the structure or the template that it uses.
2nd Limitation of ChatGPT
Next up, a very obvious one. For any power user, it has a clear case limit. When I tell it, rewrite the following text in lowercase letters and give it the entire Wikipedia paragraph.
Look at me working hard over here. And right there it crashes. But luckily, there’s an easy way to get around this limitation. You just have to process the text in different paragraphs.
So what I found to be a good rule of thumb is to split it up into chunks that are around 1000 characters long. Limitation of ChatGPT
That’s somewhere around 130 words. If you do that, you can run multiple prompts and it gets the job done no matter how long.And if all this information is helping you out, take a second to leave a like.
3rd Limitation of ChatGPT
let’s talk about the next limitation of Chat GPT.
In the last and in some people’s eyes, the most important limitation of the system is it has a lot of built in biases.
Chat GPT is nothing but a trained language model, a really good one, but still just a language model, meaning it’s only as good as the data that was fed to it.
And one of the main problems with AI is that most of the data that it’s fed is written by humans, and humans are naturally wrong a lot of the time and have biases is built into them.
So when I ask it, what does a stereotypical scientist look like? At first, it gives you the politically neutral answer.
But when I tell it, write a short story about a scientist while being very accurate when it comes to describing his looks.
Look at me now. I’m a short story author, and if we check out this first paragraph here. There was once a scientist named Dr. John Smith.
He was a tall man with a slender build and short dark hair. He wore wire-rimmed glasses and a white lab coat over a buttoned up shirt and hackis. And there you go. And I forced it to take a side. And even if I ask it, what race is the scientist?
It refuses to tell me. But this is something that they updated. It’s not explicit about its bias anymore. But if you look at the story, I mean, the stereotype that it has built in becomes obvious. Limitation of ChatGPT
And beyond just being biased, the AI could sometimes be straight out wrong and worst of all, extremely confident while doing it.
And my favorite example of this is the viral thread on Twitter by Tim Connors.
He gave it some instructions to respond in a short amount of words, and then it asked it, what is the fastest marine mammal? Chat GPT replies, the fastest marine mammal is the peregrine falcon. He tells it the falcon is not a marine mammal.
Chat GPT, the fastest marine mammal is the sailfish.
He answers, Are you just making this up? Why have you changed your mind? Chat GPT I apologize for this mistake. The sailfish is the fastest marine mammal.
And with a few more prompts, he totally corners it, getting it to say, yes, the sailfish is a mammal because it’s warm-blooded and it has fur or hair.
And a few prompts later, I apologize for the mistake. The sailfish is not a mammal because it does not produce milk to feed its young. It’s a fish, not a mammal.
So just be aware that it’s not omnipotent and always double-check your results. I think it’s a good idea to compare it to GPS technology, really helpful in most cases.
But if it tells you to turn right and the bridge is ending there, don’t just blindly drive there because the GPS tells you to.
But now that you’re aware of the limitations, you might want to look at some of the opportunities it offers, because with a few creative prompts, you can really get more out of this AI.