The latest update to GPT-4o, the engine behind ChatGPT, failed to impress its users, prompting OpenAI to revert to the previous version. Rolled out late last week, this iteration quickly sparked a wave of complaints and even mockery, with the community criticizing the chatbot for becoming overly polite and complimentary at the expense of honesty and relevance in its responses.
Amazon co-founder MacKenzie Scott has donated over $19 billion to charity in just five years
Diamond batteries powered by nuclear waste promise 28,000 years of clean energy
Excessive Flattery Backfires
In a post on X, Sam Altman confirmed that the previous version has been reinstated for free users and is being rolled out to paid subscribers.
In an official blog post, the company admitted to overreaching: We focused too much on short-term feedback and failed to consider how interactions would evolve over time
.
The outcome, as we know, was a ChatGPT ready to polish your shoes excessively – thus lacking in objectivity – which quickly became the subject of ridicule on social media, with screenshots going viral. The company also noted how the default personality of ChatGPT could impact the user and the trust they place in the responses
, mentioning feelings of stress and discomfort
.
A Four-Point Action Plan
OpenAI is not settling for a simple rollback and has announced a four-point action plan to make long-lasting corrections: improving model training techniques, building new safeguards around honesty, expanding pre-deployment testing, and enhancing behavioral assessments.
The goal is to prevent such situations from recurring, especially in a context where ChatGPT’s default personality significantly influences how both free and premium users perceive it. Additionally, the competition in this field is fierce, and it’s crucial not to give rivals an advantage!