OpenAI, the creators of ChatGPT, disclosed that GPT-4 might struggle on certain tasks following user feedback describing the model as ‘lazier’ and ‘dumber’ relative to earlier versions.
“While the majority of metrics have improved, there may be some tasks where the performance gets worse,” OpenAI stated in a Thursday blog announcement regarding new updates.
In recent weeks, GPT-4 users have expressed frustration over degraded performance compared to its prior reasoning abilities.
Frustration has been voiced on platforms such as Twitter and OpenAI’s developer forum, citing problems like weakened logic, increased error rates, failure to retain supplied context, difficulty adhering to instructions, omission of parentheses in simple coding snippets, and an inability to reference earlier prompts beyond the immediate one.
A developer relying on GPT-4 for website function coding called the current version underwhelming, likening it to switching from driving a Ferrari to encountering a worn-out pickup truck; they questioned whether they wanted to continue paying for the service.