GPT 3.5 vs GPT-4 

Introduction:

Hey, what is going on, guys and girls? Just wanted to do a fun comparison video today between GPT-3.5 and GPT-4. We all know that GPT-4 is out right now, and after doing some experimentation, I found that you’re able to get really high-quality content that is not easily detectable by AI detection tools when using GPT-4. So, I wanted to do a fun comparison in which I would take the same inputs and use them on GPT-3.5 and GPT-4, and take those outputs and run them into an AI detection tool to see which one of these models will be more easily detectable by an AI detection tool. Let’s go ahead and get started!

Comparison Test 1: Funny LinkedIn Post

For the first example, I’ll be asking both AIs to write a funny LinkedIn post about why AI will not take our jobs. Here are the outputs we got:

GPT-3.5:

“Hey LinkedIn fam! Worried that AI will take your job? Not my friends, because I have a secret weapon to keep your career safe – my terrible sense of humor.”

GPT-4:

“Hey LinkedIn fam! Worried that AI will take your job? Sure, not my friends, because I have a secret weapon to keep your career safe – my terrible sense of humor. 😄”

The output from GPT-4 is a lot better than what we got back from GPT-3.5. It’s more in-depth, fun, and formatted nicely with emojis. It sounds very human-like and addresses the audience in a great way. In terms of quality, GPT-4 is way better.

Comparison Test 2: AI Detection

Now, let’s test the ability to be detected by an AI detection tool using the content generated by both models.

Content from GPT-3.5:

When we scan this content using an AI detection tool, it shows 0% originality and 100% detected by AI.

Content from GPT-4:

When we scan this content using an AI detection tool, it shows 100% originality and 0% detected by AI. The high originality score indicates the high quality of content generated by GPT-4, which passes AI detection tools.

Another Test:

Let’s do one more test to see if this holds true.

Content from GPT-3.5:

When scanned, it shows 21% original and 79% detected by AI.

Content from GPT-4:

When scanned, it shows 100% original content. The content generated by GPT-4 is impressively high-quality and passes AI detection tools without any prompting.

Conclusion:

GPT-4 is miles ahead of GPT-3.5 in terms of understanding context and generating high-quality content. It eliminates the need for extensive prompt engineering. The content it generates is 100% original and passes AI detection tests easily. I’m highly impressed with GPT-4, and I believe it will continue to improve.