GPT-3.5 vs GPT-4

My Little Experiment

Hey, Bernard here.

I had a revelation today, and I need to share it with you. I haven’t used GPT-3.5 in half a year. It’s pretty much since I upgraded to PRO.

You heard right! When GPT-4 came out, I threw my old buddy away and never used it again.

This led me to an idea. Let’s test how they compare against each other.

So here’s what I did….

I compared these 2 models on 2 simple tasks. One simple and the other one complex. I compared the results and hit “send” on this mail!

So let’s hear it.

GPT-3.5 vs GPT-4

The rules are simple.

  1. I’ll do the same task with both models.

  2. I’ll use the same prompt.

  3. I’ll analyze the results.

  4. I give you my (expert) feedback.

Now let’s get into it.

Let’s assume I own a blog about parenting (which I don’t). I need to come up with a blog post and I need to write an introduction for it.

1/ Brainstorming blog post ideas

The task is simple. I want 10 ideas for a blog post about parenting. Let’s see how the models compare.

GPT-3.5 brainstorming

GPT-3.5 brainstorming

GPT-4 brainstorming

GPT-4 brainstorming

The results? I don’t see much difference. And do I dare to say I prefer the response from GPT-3.5?

But let’s not jump to conclusions just yet. Let’s move on to another task.

2/ Writing blog post

And now for the writing introduction. I want to have 3 paragraphs, and I want them to sound like humans wrote it.

So this is what I’ll keep an eye on.

  1. The introduction should have 3 paragraphs,

  2. The introduction should sound human.

Again, I used the same prompt with both models.

ChatGPT prompt to write blog post introduction

Let’s see what results I got.

GPT-3.5 writing

GPT-3.5 introduction

GPT-4 writing

GPT-4 introduction

You could see the difference here from a mile away. The first introduction screams “AI generated“. And it’s not very well-written either.

But if you’re a GPT-4 fan, don’t celebrate just yet.

The second introduction is much better. I don’t see any “AI words”. It sounds way more human. But still, it’s not something you can use without revision.

3/ The results

This was a quick experiment. But you could already see the outcome.

  • GPT-4 is far superior in doing complex tasks.

  • GPT-3.5 is equally good for doing simple tasks and it doesn’t lack creativity.

This only means one thing.

  • Not everyone needs to upgrade to PRO.

  • GPT-3.5 can do a lot of work.

This concludes my experiment. Now let’s check out some free prompts I’ve written this week.

ChatGPT Prompts of the Week

Psst: Do you want me to write prompts for a specific area? Let me know by replying to this email.

If not, let’s have some laughs together.

Memes of the Week

Let’s Hear Your Rating

Reply

or to participate.