OpenAI Denies Claims That ChatGPT Is Declining in Quality
OpenAI, the renowned artificial intelligence (AI) company, has recently faced a series of challenges including legal battles, an investigation by the Federal Trade Commission (FTC), and a decline in user numbers. In addition to these concerns, OpenAI has also been confronted with dissatisfaction from users of GPT-4, their paid chatbot service, who claim that its quality has been deteriorating in recent weeks.
On June 3, a post on an OpenAI developer forum expressed disappointment with the current version of GPT-4, comparing it to a sudden transition from driving a luxurious Ferrari to an old, beaten-up pickup truck. Another user added on June 14 that they had noticed a steady decline in the quality of responses and a loss of creativity and liveliness from GPT-4.
However, OpenAI’s Vice President of Product, Peter Welinder, has asserted that the company is not intentionally making GPT-4 less intelligent. In a tweet on Thursday, Welinder clarified, No, we haven’t made GPT-4 dumber. Quite the opposite: we make each new version smarter than the previous one. Current hypothesis: When you use it more heavily, you start noticing issues you didn’t see before.
Welinder’s statement holds some truth. While GPT-4 is OpenAI’s most advanced large language model (LLM) to date, the company has never claimed that the system is flawless. GPT-4 is still undergoing development and possesses glaring flaws, such as generating inaccurate information, citations, and code, alongside technical difficulties like extensive downtime.
Nonetheless, it is understandable that some users of ChatGPT Plus, who pay a monthly fee of $20 for the subscription service with the expectation of faster response times and priority access to new features, may feel frustrated or let down when these expectations are not met. And when complaints become a consensus among fellow subscribers, it further fuels disappointment.
It remains uncertain whether Welinder’s response is an attempt to downplay the concerns of GPT-4 subscribers. However, it is commendable that he has opened the discussion on Twitter and encouraged users to provide examples of any regressions they believe have occurred.
Ultimately, it is important to strike a balance between technology usage and healthy disconnecting. Taking a break from time to time can be beneficial for overall well-being. So, let Welinder’s remark serve as a reminder to log off, step outside, and embrace the real world.
