OpenAI ChatGPT image generation function is very popular, but access speed is limited due to insufficient computing power

Recently, OpenAI launched a new upgraded image generation function on its ChatGPT platform, attracting the attention and use of a large number of users. However, this hot feature also puts OpenAI in some new challenges. OpenAI founder Sam Altman revealed that the company's GPU computing power was almost overwhelmed due to the surge in demand, and he vividly stated that "the GPU is smoking", which forced them to introduce rate limits to image generation capabilities to a certain extent.

The new feature "Images in ChatGPT" is developed based on the GPT-4o model. Users can generate and edit images through simple natural language instructions and perform multiple iterative optimizations. The launch of this function marks the in-depth integration of ChatGPT in multimodal capabilities such as text, images, and code, and becomes a truly all-round agent. Users can convert ordinary photos into Ghibli-style works of art by simply entering instructions, which quickly sparked a widespread craze.

However, as the number of users surged, OpenAI realized that the original plan to promote this feature to all users would need to be delayed, especially for free users’ opening hours. Altman said that despite their strong GPU resources, it seems unscrupulous to handle such a huge amount of requests.

The process of image generation requires AI to calculate pixel by pixel, which requires a large amount of computing resources. In order to solve the bottleneck of GPU computing power, OpenAI is exploring two main paths: one is to improve the performance of GPU, and the other is to improve the efficiency of the AI ​​model through algorithm optimization. Among them, the latter hopes to improve the algorithm so that the same computing resources can handle more tasks, thereby using existing resources more efficiently.

It is worth noting that the GPU reserves behind OpenAI are among the top in the industry. As a major investor, Microsoft has provided strong support for OpenAI and purchased a large number of Nvidia Hopper chips, giving it a competitive advantage in computing power. Despite this, AI applications still have huge demand for computing power, and the industry's calls for technology iteration and resource utilization efficiency are also increasing.

In the development of multimodal technology, OpenAI is facing the challenge of balancing resources and demand.