AI Experiment.

I have been wondering what the AI (Artificial Intelligence) impact would have on genealogy projects. Please understand AI does NOT – repeat does NOT – do your research, some genealogists think it should. NO COMMENT. It’s a garbage in, garbage out scenario.

I tested 5 different AI’s. I used the same prompt for each of them. I also used the same text that I wished AI to either, rewrite and enhance and/or add additional context.

Please note: I used the free versions of each of these AI’s. The test was run toward the end of August 2024. The AI’s used were: ChatGPT, Gemini, Claude, Perplexity and Co-Pilot.

I gave each AI a number of attempts to come up with something that could be presentable. Each attempt produced varying results. If I were to combine them, it may have produced a better article but that’s not what I was testing.

Google Gemini.

Now, what I thought of each AI outcome that each of the five provided.  I will commence with the worst – Gemini. The issues encountered were: Not following the instructions, giving ideas like a structure which wasn’t asked for, changing the title completely, bias toward the Northern Hemisphere even though information was provided about Australia in the text. In the end, I gave up as it was a complete waste of my time.

Microsoft Co-Pilot.

In fourth place was Co-Pilot. This AI seemed to want to rearrange sentences by just adding a few extra words but not to the story, it restricted the word count when instructions were to add to it. Again, it was a waste of my time.

ChatGPT.

The next two placings were close but ChatGPT came in third. It provided some good “extras” but it was a lot of effort to get the desired response. ChatGPT reduced the word count like Co-Pilot. ChatGPT gave ideas like Gemini had. ChatGPT assumed quite a bit about the content. Things like: a cold Christmas, a hot August – the bias toward the Northern Hemisphere!

Perplexity.

Perplexity was the second best. However, the first run was useless – It copied exactly the same as the text provided and said here is the “new” version. The reason it came in second was after the first run, it provided good context as well as some useful ideas and hints to add to the story. Perplexity also used “Reflecting” which was used very well, especially with the text used to test AI. Another interesting addition was the mention of environmental challenges. 

Note: I was unable to find the icon for Perplexity in Pexels Free Photos.

Claude.

First place goes to Claude.

It only needed one run and the results were outstanding. It added context where needed, was an easy read and I found no real issues.

NOTE: This icon was generated in Canva. It is NOT what you see once opening Claude, except the greeting…

So, these are my findings on AI. AI provided: ideas fast, added necessary context and provided a better story.

Outcome.

Would I use any of the outcomes? Possibly not as I see AI as an assistant that gives food for thought. It is an insight into what can be done especially if you are stuck. AI is not the answer; to press a button and you’ve got your story. Remember, I provided a story and this is what was used in the outcome.

My suggestion is give it a go. There is no harm in using AI and when it comes down to the final say -YOU have that and AI does NOT.

Leave a comment