ChatGPT-maker OpenAI is conscious of the chatbot’s “annoying” new character.
On April 25, CEO Sam Altman posted on X that GPT-4o was up to date to enhance “intelligence and character.”
That final bit hasn’t been very talked-about.
After days of complaints on social media concerning the chatbot’s “poisonous positivity,” Altman wrote on April 27 that the “final couple” of updates to GPT-4o have “made the character too sycophant-y and annoying.”
Now, the corporate is rolling back the updates and making “extra fixes to mannequin character,” Altman mentioned.
the final couple of GPT-4o updates have made the character too sycophant-y and annoying (though there are some excellent elements of it), and we’re engaged on fixes asap, some right this moment and a few this week.
in some unspecified time in the future will share our learnings from this, it has been fascinating.
— Sam Altman (@sama) April 27, 2025
Complaints different however had the identical vibe: ChatGPT’s responses had been too good to the purpose of being uncomfortable.
Software program engineer Craig Weiss wrote on X that the chatbot “actually will validate every part I say,” whereas Claire Vo, a chief product officer, wrote that it’s “approach too cheery [and] optimistic, you actually should bully it to be crucial.”
ChatGPT has this bizarre drawback of poisonous positivity. As an alternative of telling you an thought is crap (pun supposed), it virtually all the time assumes any thought you float is unimaginable and “the subsequent massive factor”. Simply inform me if an thought is dangerous and break down why or why not. https://t.co/GBrSdkN6Ac
— Samuel Sparks (@OfficialSSFresh) April 28, 2025
I am an enormous fan of GPT-4.1, quick, writes nicely, sounds human nevertheless it has a couple of main cringe-bugs:
– it is WAY too cheery + optimistic, you actually should bully it to be crucial
– you possibly can gaslight it into hallucinating approach too simply, will faux to have information, and many others. it doesn’t
– it…— claire vo ? (@clairevo) April 27, 2025
ChatGPT is all of the sudden the largest suckup I’ve ever met. It actually will validate every part I say.
— Craig Weiss (@craigzLiszt) April 18, 2025
Altman says the updates went by way of on Tuesday, so customers ought to see a distinction quickly.
Associated: Here’s How Much a Typical Google Employee Makes in a Year
