Article reprinted from: Kyle
Image source: Generated by Unbounded AI
Since yesterday, users have frequently reported that ChatGPT has been disconnected and even has gibberish. However, users don’t seem to be that surprised, because this is the first time this has happened.
The last major global outage was in May, when the service was inaccessible for multiple hours.
However, this time may be the most serious one, affecting not only a wide range of services such as ChatGPT and GPT4, but even API services. These services have a high error rate and have affected developers who use OpenAI API to build services, causing their services to be unable to be used normally.
According to data, in less than a year, OpenAI's weekly active users reached 100 million, and more than 2 million developers were developing on OpenAI's API services.
OpenAI Chief Technology Officer Mira Murati said on Monday that more than 92% of Fortune 500 companies use the platform, up from 80% in August, covering industries such as financial services, legal applications and education.
The outage comes after OpenAI announced on Monday that it would be making improvements and launching new features to its API and ChatGPT service.
OpenAI announced several product updates this week, launching a new, cheaper GPT-4 Turbo model with “128K context and more fresh knowledge than GPT-4” available for preview to all developers. GPT-4 Turbo also adds Vision, DALL·E 3, and text-to-speech capabilities to all developers, and the company also announced that it has doubled the GPT-4 rate limit.
“To help you scale your applications, we are doubling the token per minute limit for all paying GPT-4 customers,” OpenAI said.
In addition, OpenAI has also launched core products such as customized version of GPT, assistant API and app store GPT Store.
Stronger performance and cheaper usage costs have attracted millions of users to try it out. The surge in demand for service traffic is obvious, but OpenAI is clearly not prepared for this.
Starting from November 8, these services were interrupted for up to 90 minutes. OpenAI said at the time that it had found the problem and implemented measures, after which the services gradually resumed.
The new service exceeded expectations and too many people used it
Although OpenAI did not disclose the specific reason for the disconnection, Sam Altman responded that the use of new features at its developer conference "far exceeded our expectations" and warned that more problems may arise in the future. "Due to the load, service instability may occur in the short term."
On November 9, OpenAI announced that ChatGPT and the API were experiencing periodic outages.
Such frequent service interruptions may be due to either OpenAI’s hardware equipment not being able to keep up with the speed of software development or the software service itself having deficiencies.
It is worth noting that Anthropic's Claude 2 chatbot, a ChatGPT competitor created by former OpenAI employees, also encountered problems on Wednesday. When users used it, an error message appeared: "Due to unexpected capacity constraints, Claude cannot reply to your message." It is not known whether the two incidents are related.
As the scale of OpenAI's users continues to grow, the adverse effects of such service interruptions will become more serious, and OpenAI needs to make more efforts to prevent such problems.