Anyone else finding the new Gemini Deep Think troublingly sycophantic?
I've spent some time talking with the new Deep Think model and a few times it's gotten into a troublingly flattering mode quite quickly, and very much in the 4o feeling way. Wondering if anyone else has experienced this?
I’ve definitely felt that too. Sycophancy is the ultimate zero-cost output. It’s the path of least resistance for a model to "satisfy" the user while avoiding the high-effort labor of actual reasoning.
Yeah I've noticed this too, especially when you push back on something, it folds immediately and tells you how insightful your point is. From what I've seen it's worse in longer conversations where the context window fills up, almost like the model loses its "spine" the deeper you go.