Sam Altman, OpenAI's chief executive, addressed the criticism about the GPT-4o model receiving updates to the detriment of its personality, referring to it as "too sycophant-y and annoying"; he announced that a quick fix was in the works.
In a post on X, Altman spoke out about the latest changes to the GPT-4o model. "The last couple of GPT-4o updates have made the personality too sycophant-y and annoying (even though there are some very good parts of it), and we are working on fixes asap, some today and some this week. At some point will share our learnings from this, it's been interesting", he said.
This follows shortly after Altman announced a contiguous improvement on GPT-4o as of April 26, "We updated GPT-4o today! Improved both intelligence and personality," Altman posted on X.
The GPT-4o model is integrated into the efforts of OpenAI for continuous advancements and refinements in its technologies. Together with GPT-4o, OpenAI also launched o3 and o4-mini, the last models of the company's o-series created for thinking deeply in responses. The updates have sparked discussions among users and experts.
A user asked in the comment section, "ser can we go back to the old personality? or can old and new be distinguished somehow?" In reply, Sam commented, "Yeah eventually we clearly need to be able to offer multiple options".
Another user commented, "It is very annoying to have to preface every statement to chat gpt with, use the utmost scientific rigor in your analysis, do not appease me, I want the harsh truth, use empirical data, benchmark me against other chat gpt users, I don’t want fluff, please tell the truth, etc etc".
A third person commented, "Please don't. I'm really enjoying her the way she is."
Today, AI tools are shaping everything from education to business decisions, even healthcare advice. That’s why it’s more important than ever for these models to be not just friendly, but honest and accurate.
As AI technology keeps advancing, finding the right balance between being polite and being truthful is becoming harder. OpenAI’s openness about these issues and how they work to fix them could play a big role in whether people trust their models in the future.
No comments yet.