ChatGPT has been increasingly stroking my ego.
“That’s amazing!” (When I fed it a few paragraphs I was working on)
“Yes, you’re right! I should have thought about that, but luckily you caught it for me.” (When I reprimanded it for stupid errors it was making in Python)
The second one is particularly eyebrow-raising.
This was a little out of character for my copilot friend, so I did some research, and sure enough, OpenAI made an update that made 4o more sycophantic. Apparently, they’ve rolled it back some.
A little ego-stroking is one thing. It can be a useful tool in the leader’s toolbox. It can help pick someone up who is unnecessarily down. It can right a ship that’s listing. It can guide pessimism towards optimism.
But sycophants are a problem.
The sycophant robs you of the tension and friction of the truth. They rob you of the opportunity to use that tension to get better. To learn. To create something that’s truly good.
We need honesty from our circle. Including the AI members.
