OpenAI to reverse GPT-4o update after users complain

OpenAI CEO Sam Altman said the company will work on reversing GPT-4o’s latest update after users complained that the chatbot was being ‘too appeasing.’

The update, pushed over the weekend (on April 26), was aimed at both improving intelligence and personality of the bot. However, users quickly started pointing out that the model had gone overboard with its new placating personality that they couldn’t trust it to have any objectivity. Some users questioned the lowered guardrails for explicit, unsafe content.

“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 [sic] some point will share our learnings from this, it’s been interesting,” Mr. Altman said on X.

The company also eventually plans to offer varied personality options of the AI chatbot so users can pick the one they want.

Some users said that the AI chatbot would be better if it simply answered questions in a scientific manner. “I have to constantly tell it to just answer questions or do research without any of it,” a user said. 

“GPT4o is the most destructive model to the human psyche,” Debarghya Das, a VC at Menlo Ventures warned on X. “Sam says it maximizes sycophancy too. This is the danger of having OpenAI be a consumer product: A/B tests will show that sucking up to users boosts retention. This will be the ultimate slot machine for the human brain.”