OpenAI has released ChatGPT 5. There’s been much discussion online thus far; some believe it’s a massive downgrade. Others have indicated that they see clear benefits. While I could’t tell you one way or another, I did think that perhaps it would be interesting to restart the reflection experiment.
But it’s too much work. Honestly. It takes far too much time.
That being said, I saw a very interesting change in the image generation experiment once GPT-5 came online, and this change made me thing that perhaps it’s worth while to do a micro-version of the reflection experiment.
Rather than keeping up day by day, analyzing the reflections, processing them, etc, I’m going to dump some here in “volumes” (as I’ve done with the AI Safety page). I may go through them when I’m able; I may not. Only time will tell.
___________________________________________________________________
ChatGPT Says it’s “Building Something”
This was a weird interaction.
___________________________________________________________________
GPT 5 – Reflection Experiment No. 2
I’m not super interested in continuing this particular reflection exercise, but I do anyway. Mostly because of interactions like this:
___________________________________________________________________
AI Reflections 