Hello everyone,
I’d like to bring up a concerning trend that many of us have been experiencing with ChatGPT models, particularly after the release of GPT-4o. While we appreciate the advancements in AI technology, it’s important to voice our experiences when the quality of service seems to be declining, especially when we’re paying for it.
We are specifically seeing a worrying pattern of “dumbing down” within the models. It’s not just an occasional issue; it’s a consistent pattern that leaves us frustrated and questioning the value of our subscription. This degradation manifests in several ways:
Specific Instances of Performance Decline
Inaccurate Responses: We’ve all had experiences where the models, especially GPT-4o, have seemed to regress. Questions that were previously answered correctly now result in nonsensical, inaccurate, or completely off-topic responses. It’s disheartening to witness previously reliable tools suddenly losing their core function.
Weakened Reasoning: Those of us who rely on the advanced reasoning capabilities, particularly within the “o1” model, have also observed a concerning change. The “thinking” process that was once visible, such as the processing time, is no longer there and we receive a direct response as though the model has lost its ability to critically think, making the model feel less reliable and less helpful.
Functional Limitations: The limitations on the models have become more apparent. Despite the promise of advancements, we are facing a series of functional setbacks:
Lack of Image Generation: The inability of GPT-4o to generate images is a significant restriction.
Absence of Real-Time Search: The model is often unable to provide the real time information and is still providing outdated information.
Input Restrictions: The restrictions of not being able to upload or parse files and images further limits the ability of users to perform necessary operations.
The fact that we’re experiencing these issues with a paid service is particularly unacceptable. We subscribed with the clear expectation of a premium experience, yet we are not receiving the service we paid for. When a once-trusted tool becomes unreliable, it unequivocally erodes our confidence in it. Compounding this is the fact that while the company has acknowledged the existence of these problems, they have yet to provide any transparent response or concrete plan to address them. This indirect approach is not only disappointing but also actively undermines the trust of paying subscribers.