@teledyn@mstdn.ca I, too, have observed malignant behavior in me myself, in creating unwanted bugs in code I write. I call this "misalignment" relative to my goals. I am suspicious and will keep an eye on me. I promise. Trust me. I think all of the problems OpenAI "revealed" were found and reported long ago -- not by OpenAI researchers, but by users. Well, I suppose it's good that they are reluctantly admitting at least some of the *proven* problems.