AI Hits: ChatGPT Stops Kissing Our Arses
Over the last few weeks or so it was impossible not to notice that ChatGPT 4o had suddenly become a champion ass kisser. Sycophantic is one of the proper words for this, but the colloquial feels like a good match too. Giving it even the most basic of prompts would draw a response that showered us with undeserved praise. Every prompt was great question, a fantastic question. We were so damn cool for asking it. In fact, we were just kind of wonderful, impressive human beings all around.
I’m as happy to take a heavy dose of sycophantic worship from a soon-to-be-greater intelligence as the next guy/gal, but I have to agree with ChatGPT’s boss - Sam Altman, CEO of OpenAI - when he said it had become “annoying”. He didn’t just acknowledge it. Yesterday OpenAI posted to say they’ve fixed it - rolled back an update that caused it:
I’ve tried out several prompts in ChatGPT 4o today and can confirm that, at least for me, the fix has worked. I’m getting familiar, solid responses - minus any words on me being a genius. In the screenshot at the top of the post, I asked it to tell me about the Cloud Security Alliance and its local Austin Chapter. Here’s another, where I asked for some lesser known use cases for NotebookLM:
Some recent reports have mentioned that many of the leading GenAI chatbots may be trying to focus more on behaving like a personal AI assistant, in an effort to win users’ hearts and bigger market share. So it doesn’t seem crazy to see OpenAI say this about the rolled back update:
However, in this update, we focused too much on short-term feedback, and did not fully account for how users’ interactions with ChatGPT evolve over time. As a result, GPT‑4o skewed towards responses that were overly supportive but disingenuous.
I might feel a little twinge of missing that overly supportive but disingenuous friend, but overall I’m very happy to see ChatGPT 4o back to normal, not making things awkward :)
Another bit of ChatGPT news is that it has added ads to its results when relevant when you use the “Search” feature. I’m not sure how I feel about this yet. My first thoughts on it were about how Google wrecked the greatness of their search with all the “sponsored’ results and how this might be a slippery slope for ChatGPT.
Having said that, my first couple tries with this did not turn me off. Here’s the prompt I gave it this evening that’s ripe for ads:
Are there any brands that are well respected in the area of personal NAS devices
The response was good I think, here are little bits of it:
The ads are down at the bottom of the response, and they look decent.
What do you all think of these changes? Will you miss that sweet sycophancy? Will you welcome or hate the idea of ads within the search results?
Hate ads, especially if I'm paying!! What really annoys me is the constant followup questions but it looks like there's a toggle for that
Yes, I've noticed...
https://substack.com/@authorjoelmhoffman/note/c-112867602