My thoughts on AI written content

AI content is really easy to create today. And the content is pretty good, too. It’s so good that it’s now table stakes for all the content we create, from blog posts, sales collateral and emails to LinkedIn posts and video scripts. You name it.

However, in a world where everyone has access to ChatGPT, Bard, and the many AI writing tools, the sea of boring is starting to rise.

If you’re going to use ChatGPT for your content and don’t want to sound boring, you need to understand what ChatGPT and AI is.

ChatGPT is what you call an LLM.

“What’s an LLM, Jason?” (This is important. So stay with me for a second.)

LLM stands for Large Language Model, designed to understand and generate human-like text.

It’s trained on vast amounts of data (and I mean VAST. Like ginormous.) from books, articles, and websites, which helps it learn patterns and relationships between words and phrases.

You can think of it as a really, really, REALLY smart autocomplete.

What you should also do to understand how ChatGPT works is to use the ChatGPT playground.

Try to see what happens when you ask the same question but change the Temperature, Top P, Frequency Penalty or Presence Penalty.

Temperature controls the randomness. If you lower the score, the more repetitive and deterministic it will become. Raise it and see how it seems to take matters into its own hands. 🙂

The Frequency Penalty allows you to tweak the model so that it decreases the model’s likelihood of repeating the same lines or words. The higher the penalty, the higher the chance that it won’t repeat words or phrases it’s already said.

Or the Presence Penalty, where the higher the penalty you set, the more likelihood you’ll get the model to talk about new topics.

It also has shortcomings. I highly recommend you read OpenAI’s FAQs on ChatGPT.

1. It’s only been trained on data up to 2021. That’s 2-year-old data. And we know how old that is in internet years.
2. It can’t verify facts or provide references as it can’t access the internet to look up recent data other than its own knowledge or logic. With ChatGPT 4.0, though, it can browse the internet, so this is becoming less of an issue, but just be aware that you will need to explicitly tell it to browse the internet or give it the URLs to visit.
3. It can inadvertently generate biased, offensive, or inappropriate content based on biases present in its data.

Ok, now that you understand it a little better, here are a few things where we humans still have an advantage over AI-written content.

• Empathy and emotional intelligence – We understand each other’s emotions and how to react to them.
• Creativity – You might not think it after using ChatGPT, but we are more creative than AI as we can develop totally new ideas.
• Nuance – We can pick up on little details in language and culture that subtly changes the context.
• Personal touch – We can add our unique style and voice to content to make it enjoyable and stand out.

​Like in this article, you’ll see some of my voice, style and subtle jokes being thrown in, along with some facts that make it more memorable than if I had ChatGPT write all of it.

And that’s the balance we’re after. Use it to help, but handing it all over will end in big computer tears.

I’ll leave you with this meme to brighten your day. 🙂

​

Jason Bagley Founder, Growth Experts

Founder of Growth Experts, public speaker, and member of the Entrepreneurs’ Organization.