Home Artificial Intelligence ChatGPT vs GPT-4 for Summarization Intro Initial Prompt Results Conclusion Enjoyed This Story?

ChatGPT vs GPT-4 for Summarization Intro Initial Prompt Results Conclusion Enjoyed This Story?

0
ChatGPT vs GPT-4 for Summarization
Intro
Initial Prompt
Results
Conclusion
Enjoyed This Story?

The initial prompt that I used for summarization is the next:

Create an in depth and in-depth summary of this section of the Huberman Lab Podcast. Don’t compress any information from the unique transcript. If there’s any information that must be added, please achieve this.

I followed by adding the transcript from the podcast. Let’s see what the outcomes are!

ChatGPT

First, let’s see the summary of the transcript created by ChatGPT:

Summary of the transcript created by ChatGPT, image by the creator.

The summary made by ChatGPT was correct however it was pretty hard to read through. It wasn’t formatted in any way — no lists, no highlighted information.

That’s why I followed up with an extra formatting prompt:

Format the summary so it can be easy to know. Highlight a very powerful facts.

Interestingly, adding this sentence to the tip of the initial prompt didn’t make any difference within the initial output — it needed to be used after the model returned the primary summary.

Let’s see the formatted summary:

Formatted summary, image by the creator.

The formatting prompt did wonders! It helped to restructure information and show it in an easily digestible way, including lists and vital facts. In my view, ChatGPT did pretty much here!

GPT-4

Now let’s move on to GPT-4. I used the identical initial prompt for summarization:

Summary of the transcript created by GPT-4, image by the creator.

In my view, the output is aligned lots higher than the summary from ChatGPT. It is usually easier to digest, we get a transparent list of 4 phases of the plan and their time horizons.

I followed up with the identical formatting prompt once more:

Formatted summary, image by the creator.

We got an excellent more concise summary. Again, I believe the outcomes are more aligned with the prompt.

In my view, each of the models did well within the summarization task. Nevertheless, GPT-4 didn’t require an extra prompt to attain an aligned and easy-to-read summary.

Hence, the winner on this duel is !

That is the primary story of documenting the means of constructing the Huberman Lab Podcast Database. I’m actively working on it on my Github!

If you would like to not sleep so far with recent AI advancements and my journey of learning AI, my newsletter is an incredible place to be!

In case you’d wish to get my latest stories on to your inbox, subscribe!

You can even find me on Twitter!

More content at .

Join for our . Follow us on , , , and

LEAVE A REPLY

Please enter your comment!
Please enter your name here