GPT-5’s Rocky Start: Altman Addresses Issues and Promises Fixes
GPT-5’s Rocky Start: Altman Addresses Issues and Promises Fixes
OpenAI’s GPT-5 recently launched, but it hasn’t been a smooth ride. Users quickly noticed issues, particularly with the model’s auto-switching feature, leading to widespread complaints. In a recent Reddit AMA (Ask Me Anything), OpenAI CEO Sam Altman and his team addressed these concerns, promising fixes and even hinting at bringing back the beloved GPT-4o for some users.
The ‘Dumber’ GPT-5: What Went Wrong?
One of the key features of GPT-5 is its real-time router, which intelligently selects the appropriate model for a given prompt. This system is designed to optimize for speed or thoroughness, depending on the nature of the query. However, many users reported that GPT-5 felt significantly less capable than its predecessor, GPT-4o.
So, what happened?
Altman explained that a service disruption (or “sev” in tech speak) had knocked the auto-switcher offline for a significant portion of the day. This meant GPT-5 wasn’t functioning as intended, leading to subpar performance. “GPT-5 will seem smarter starting today,” Altman assured users. “Yesterday, we had a sev and the autoswitcher was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber.”
He also mentioned that OpenAI is actively tweaking the decision-making process of the router to ensure users get the right model more consistently. They also plan to increase transparency, making it clearer which model is answering a specific query. This improved transparency will give users better insight into the reasoning behind the model’s responses.
A Nod to Nostalgia: Will GPT-4o Return?
The outcry for GPT-4o was so strong that Altman promised to consider bringing it back for Plus subscribers. “We are looking into letting Plus users to continue to use 4o. We are trying to gather more data on the tradeoffs,” he wrote. This suggests that OpenAI is weighing the benefits of offering both models side-by-side, potentially catering to different user preferences and use cases.
This is a smart move by OpenAI. By listening to user feedback and offering options, they can improve user satisfaction and retain subscribers.
Doubling Down: Increased Rate Limits for Plus Users
To further alleviate user frustration and encourage experimentation with GPT-5, Altman announced a doubling of rate limits for Plus users. This means more opportunities to explore the new model’s capabilities without worrying about hitting monthly prompt limits. This allows users to fully test and integrate GPT-5 into their workflows.
The Infamous ‘Chart Crime’
One of the more humorous moments of the GPT-5 presentation was the appearance of a wildly inaccurate chart. The chart presented a lower benchmark score with a much taller bar, leading to a flurry of jokes online, dubbed “chart crime.”
While Altman didn’t address the chart directly during the AMA, he previously acknowledged it as a “mega chart screwup” on X. It’s worth noting that the charts in the official blog post were accurate, suggesting the error was limited to the live presentation.
User Feedback is Key
Despite the initial hiccups, Altman emphasized OpenAI’s commitment to stability and continuous improvement. “We will continue to work to get things stable and will keep listening to feedback,” he stated, signaling a willingness to address user concerns and refine GPT-5 based on real-world usage.
Actionable Takeaway
If you’re a GPT-5 user, keep an eye out for improvements in the model’s performance in the coming days. OpenAI is actively working to fix the auto-switcher and refine its decision-making process. Don’t hesitate to provide feedback to OpenAI – your input can help shape the future of GPT-5.
Key Takeaways
- GPT-5’s initial rollout was hampered by issues with its auto-switching feature.
- OpenAI is actively working to fix these issues and improve the model’s performance.
- The company is considering bringing back GPT-4o for Plus subscribers due to popular demand.
- Rate limits for Plus users are being doubled to encourage experimentation with GPT-5.
- A comically inaccurate chart during the presentation became a viral meme.
- OpenAI is committed to listening to user feedback and continuously improving GPT-5.
FAQ
Q: Why did GPT-5 seem dumber than GPT-4o at launch? A: The auto-switcher, which intelligently selects the appropriate model for a prompt, was malfunctioning due to a service disruption.
Q: Will GPT-4o be brought back? A: OpenAI is considering bringing it back for Plus subscribers and is gathering data on the tradeoffs.
Q: What is OpenAI doing to improve GPT-5? A: They are fixing the auto-switcher, refining its decision-making process, and increasing transparency about which model is answering a given query.
Q: What are rate limits? A: Rate limits are the number of requests a user can make to the model in a given period. OpenAI is doubling rate limits for Plus users so that they can test the model effectively.
Source: TechCrunch
