ChatGPT is a big deal…but how big of a deal is it? Today I want to share my thoughts on learning language models, a litmus test for when to use them, and an educational (and entertaining) video about AI. Let’s dive in…
🤖 Marketing Tip: Limits of AI & How We Use It
I know, I know, you can’t go anywhere without hearing about ChatGPT and I guess today is no different for The Blink.
Yes, ChatGPT is an amazing piece of technology but NO it is not the “super AI” that marketers are making it out to be! Go figure, marketers making a big deal out of something…that never happens 😉.
ChatGPT is just a language model that predicts what to say next based on its data set. At launch that data set was MASSIVE, it included pretty much the entire Internet up to 2021.
Today, it can crawl pages and you can get up-to-date pages while using detailed prompts for your desired output.
The thing is, you won’t be able to get anything out of ChatGPT if you:
So if ChatGPT is going to need a human overseer for the foreseeable future, how should you use it? Here’s how:
When to use it:
When you’re trying to get started while staring at a blank page. You have a topic but you’re not sure how to get started. Critiquing is a lot easier than creating and ChatGPT can give you plenty to critique.
A good example is a blank Google Doc titled “YouTube Shorts Video Ideas.” The right ChatGPT prompt is great at filling the page here.
When NOT to use it:
When you “kind of” know what you’re going to say or have a specific hook in mind. The fact is, ChatGPT is missing a critical component in it’s language model when you have a very specific output in mind: it’s missing the data set you used to get to that semi-complete idea.
In these cases, you’ll be refining prompts way too much and editing the output to a degree that is no longer saving you time or sanity.
An example would be a win back email from a very specific segment of customers where ChatGPT won’t have the internal knowledge you have while writing.
📺 Fun Watch: 14 Minute Video Sharing the Limits of Learning Language Models
Give it a watch, the comparison to the game Go, is a great thought experiment.