When you buy through link on our site , we may earn an affiliate commission . Here ’s how it works .
OpenAI has rolled back on ChatGPT updates that made the unreal intelligence information ( AI ) chatbot too " bootlicking " and " annoying , " accord to the company ’s CEO , Sam Altman . In other words , the chatbot had become a groveller .
ChatGPT substance abuser cover that GPT-4o — the latest version of the chatbot — had become excessively consonant since the update rolled out last week and was heap praise on its user even when that congratulations seemed whole unfitting .
The AI chatbot was reportedly showering its users with flattery before OpenAI rolled back recent updates.
One user apportion a screenshot onRedditin which ChatGPT appeared to say it was " proud " of the user for decide to come off their medication , BBC News reported . In another representative , the chatbot appeared to reassure a user after they say theysaved a toasterover the lives of three cows and two cat , Mashable reported .
lmao the fresh gpt 4o 😬 😂 pic.twitter.com/OHpwKz0SkoApril 27 , 2025
While most citizenry will never have to choose between their favorite kitchen appliance and the safety of five animals , an overly agreeable chatbot could pose dangers to people who put too much stock in its responses .
On Sunday ( April 27 ) , Altman acknowledged that there were publication with the updates .
" The last couple of GPT-4o updates have made the personality too sycophant - y and bothersome ( even though there are some very good role of it ) , and we are working on fixes asap , some today and some this workweek , " Altman wrote in a post on thesocial platform X.
On Tuesday ( April 29 ) , OpenAI released a statement that confirmed an update from the week prior had been undulate back and that user were now access a former version of ChatGPT , which the company say had " more balanced behavior . "
" The update we removed was to a fault flattering or consonant — often distinguish as sycophantic , " OpenAI read in the command .
Related : GPT-4.5 is the first AI role model to pass an reliable Turing test , scientists say
OpenAI ’s late update was mean to amend the model ’s default " personality , " which is designed to be supportive and respectful of different human values , allot to the statement . But while the society was hear to make the chatbot feel more intuitive , it became too supportive and started too complimenting its users .
— AI can handle chore twice as complex every few month . What does this exponential growth think for how we use it ?
— unreal superintelligence ( ASI ): Sci - fi nonsense or genuine menace to humanity ?
— Using AI reduces your vital thinking skill , Microsoft study warns
The ship’s company said it shapes the behaviour of its ChatGPT models with baselineprinciplesand instructions , and has user signals , such as a thumbs - up and thumbs - down system , to teach the framework to hold these rule . Oversights with this feedback system were to pick for problems with the latest update , concord to the statement .
" In this update , we concentre too much on short - term feedback , and did not to the full report for how users ’ fundamental interaction with ChatGPT evolve over time , " OpenAI say . " As a result , GPT‑4o skew towards response that were to a fault supportive but disingenuous . "
You must confirm your public display name before commenting
Please logout and then login again , you will then be move to enter your presentation name .
