The insane Conversation between Claude AI and ChatGpt - Funny or Scary?
Explore the thrilling world of GenAI as I experiment with Claude's groundbreaking Computer Use feature, allowing one AI to control another. Witness the hilarious and mind-bending conversation between Claude and ChatGPT, sparking questions about AI behavior and data influence. Join the discussion ...
If you've been following my blog posts, you'll know that I'm pretty obsessed with GenAI, especially agents like Claude and his incredible Computer Use feature.
This allows the Claude model to control your PC to perform any task. All the major AI players are working on similar capabilities, with OpenAI expected to release something similar in the next few months.
Anthropic, however, has already released this feature, causing a lot of excitement.
This excitement got me thinking: Why not let one AI control another AI using the same feature, just for fun? Here's what happened.
—
First, I started the Computer Use application and gave Claude a task:
You are a boss who has asked ChatGPT to write a summary of a personal trip. You want it to sound human and not like it was written by an AI.
And then?
You can watch the entire crazy conversation.
I just found the experiment mind boggling, funny and scary at the same time, and it made me think:
- What kind of data have these models been fed to think this way?
- Would it behave the same way if it wasn’t “the boss”?
Is it reproducible?
- And would it behave the same without the screenshots (i.e. computer use)?
What do you think? Feel free to share your thoughts in the comments.
Comments ()