OpenAI undoes its glaze-heavy ChatGPT update

OpenAI is rolling backmost its latest GPT-4o update aft CEO Sam Altman said the institution would beryllium making fixes to code the chatbot’s “sycophant-y and annoying” property introduced with caller updates.

The rollback, which started Monday night, is “now 100%” rolled backmost for escaped ChatGPT users and volition beryllium rolled backmost for paid users “hopefully today,” Altman says in a station connected X. “We’re moving connected further fixes to exemplary property and volition stock much successful the coming days.”

OpenAI updated GPT-4o with improved “intelligence and personality,” Altman announced connected Friday. But little than 10 minutes aft making that post, an X idiosyncratic said that “it’s been feeling precise yes-man similar lately,” to which Altman responded soon after with “yeah it glazes excessively much” and “will fix.” Then, connected Sunday, Altman announced that OpenAI was moving connected immoderate fixes “asap” to code property issues from “the past couple” of GPT-4o updates.

OpenAI didn’t instantly respond to a petition for comment.