I was doing something silly today for a personal project. I wanted to generate a series of business names for an alternate universe, just for fun for a side project I’m working on so I use the exact same prompt in Claude and Chat GPT and I noticed something really interesting. The list of 25 business names that they came up with were almost identical. How is that possible? How can two separate artificial intelligence platforms developed by entirely separate companies return the same results?
That’s not what “insider trading” means, not even a little bit.
One possibility - they were trained on the same data sets and as a result return the same results because what they do is predict the most likely next word or words, statistically based on their training data.
Possibility two - They are sharing data sets outside the public eye.
Possibility three - Each of these LLM’s is stealing from the other in real time (sending your prompt to other LlM"s and see what they report back and adding it to the list it drew up).
Possibility four - Someone at one or both of these companies is a double agent.
Here’s a question. Did you prompt these LLM’s with the same prompt multiple times? Or just one each?
Also, you may want to read this:
https://www.techtarget.com/searchenterpriseai/feature/Claude-AI-vs-ChatGPT-How-do-they-compare