top of page

AI Battle: Blog Content Creation

  • Writer: Jenny Kay Pollock
    Jenny Kay Pollock
  • 8 hours ago
  • 2 min read
Visualization of two AI robots in a dance battle. Two glowing robots in a futuristic dance-off on a checkered floor, surrounded by colorful lights and screens, in a lively club setting.

I took the same prompt and the same file of a discussion from a #WxAISocialSaturday and did a three way AI battle. I used the free version of all of these tools.


Prompt: 1. Summarize the whole discussion and make it into a blog post. Include 10 direct quotes and cite who they are from.
2. Pull all the AI tools mentioned in the discussion and sort them by the number of times they were mentioned and provide URLs to the tool.
Text interface with “Good evening, Jenny.” Prompt shows tasks to summarize discussions and list AI tools, below limited free plan notice.

The AI Contenders

  1. Perplexity


List of AI tools with mentions: Lyndi.ai (2 mentions) and Maven's course (1 mention). Missing URLs. Text on a plain background.

  1. Claude


    Claude said that my original file was too big for the free version so that was a big reduction in points from me. I had to switch the content we were using for the test.

List of discussed AI tools: LittleLit.AI, Lyndi.ai, Prickly Pear Health, Community Bots, and Claude. URLs not provided.

  1. ChatGPT

    Table of AI tools by mention count: Lyndi.ai (2), Maven AI Course (2), Do That Dave AI (1). Highlights AI, community, policymaking.

Judging Methodology

I wanted to reduce bias in the evaluation because of course we all have our favorite AI tools. So naturally I outsourced the judging to AI. I went to NotebookLM and added in the output from each AI.


NotebookLM's Evaluation:

•Perplexity provided the most comprehensive list of AI-related entities mentioned, attempted to provide URLs, and summarized the discussion effectively, even though its blog post formatting was simple.

•Claude created a well-structured blog post with clear headings but missed some AI tools in its list and incorrectly counted mentions.

•ChatGPT also produced a good blog post structure but might have fallen short on the exact number of quotes and the completeness of the AI tool list.

Therefore, based on these outputs, Perplexity appears to have best addressed the prompt by focusing on extracting the requested information, particularly the AI tools and their URLs.

The Judge Gets In the Ring

I became curious about how NotebookLM would perform on the task so I created a new notebook and uploaded the source material and the prompt. Here's how NotebookLM did:



Judging and Conclusion by a Human


As I have been working on providing summaries of our #WxAISocialSaturday I have found myself using a customGPT I made after this test.


What is the best tool for the job today may not be the best tool for the job tomorrow.

I will continue to use different AI tools. This is important as the tools continue to evolve and change.





Comments

bottom of page