OpenAI: ChatGPT Remains Intelligent

OpenAI is pushing back against claims that its AI model, ChatGPT, has become dumber. Peter Welinder, OpenAI’s VP for Product, took to Twitter to dismiss user complaints and clarify that each new version of ChatGPT is actually smarter than the previous one. The discussion around ChatGPT’s performance had intensified recently, with Insider even publishing a story suggesting that the AI model had gotten lazier and dumber. Speculation arose that OpenAI had deployed a new and more cost-efficient version of GPT-4, which may have compromised its quality. However, Welinder did not directly address this theory in his tweets. Instead, he proposed that users are perceiving performance issues with ChatGPT after more extensive usage, leading to a misinterpretation of its capabilities. While some users have pointed out that earlier GPT models work better for specific prompts or use cases, others have expressed dissatisfaction with the content moderation for ChatGPT. Concerns have been raised about its inability to answer certain types of questions, including political topics and profane statements. OpenAI encourages users to provide concrete examples of ChatGPT’s alleged regression. The company’s response has received mixed reactions, as ChatGPT has been known to produce incorrect answers in the past. However, Welinder’s request for specific examples indicates that OpenAI is not entirely dismissing user feedback. Some users have also noted that earlier GPT models perform better in certain situations. Additionally, limitations stemming from content moderation have impacted the performance of ChatGPT in specific contexts. Despite the ongoing controversy, OpenAI remains committed to addressing user concerns and further improving its AI models.