GPT-4 isn't even familiar yet, and GPT-5 is already on its way?
The highly competitive field of large-scale models may soon face the "ultimate weapon" from OpenOpenAI.
According to trademark attorney Josh Gerben, OpenOpenAI submitted a trademark application for "GPT-5" to the United States Patent and Trademark Office (USPTO) on July 18.
According to the trademark, GPT-5 provides offline/online versions of "computer software for generating speech and text" as well as offline/online versions of "computer software for natural language processing, generation, understanding, and analysis."
However, the specific features and improvements of GPT-5 are yet to be officially confirmed by OpenOpenAI.
According to OpenOpenAI's previous plans, GPT-4.5 will be publicly released in September or October this year, followed by the release of GPT-5 before the end of the year. Now that we have entered August, OpenOpenAI's timeline is running short.
In May, members of the OpenOpenAI development team even tweeted that GPT-5 may achieve Artificial General Intelligence (AGI), which refers to artificial intelligence capable of human-like intelligence and performing various tasks.
However, it should be noted that in early June, OpenOpenAI revealed during a speech that they had not yet started training GPT-5.
As a reference, OpenOpenAI spent six months on adversarial testing before releasing GPT-4, aiming to ensure the safety of OpenAI (the company did not disclose details such as the training duration of the model itself).
Considering this, we may have to wait for a few more months at least to see GPT-5. OpenOpenAI's trademark application is merely a measure to prevent others from using the name GPT-5 and does not indicate that the development of the large-scale model is already complete.