OpenAI quietly released a new version of ChatGPT, GPT-4o, but the news did not stir up much excitement. Compared with the official enthusiasm, the reaction of netizens was unusually calm. Many users even said that it was difficult to detect the obvious difference between the old and new versions. This triggered many speculations and questions about the performance of GPT-4o and the future development of OpenAI. Some people even jokingly called OpenAI "out of stock" and compared it with its competitor Claude.
ChatGPT recently launched a new version of GPT-4o. The official account announced the news with enthusiasm. However, the reaction of netizens was extremely calm. The comment area was full of doubts and seemed to be insensitive to the difference between the old and new versions.
Many daily users of ChatGPT said they did not notice any significant differences between the new version and the old version. This lukewarm response even led some netizens to joke about whether OpenAI had "run out of talent" and doubted whether the performance of the new model could surpass its competitor Claude-3.5-Sonnet.
In the face of doubts from netizens, the official responded quickly, clarifying that GPT-4o is not the same as the previous API update version, and brought good news: even free users can experience the new version (although the number of messages is limited).
In the recently popular large-model IQ test, both the new and old versions of ChatGPT can answer questions correctly under specific questioning methods, such as about the number of r's in the word Strawberry. However, for the Chinese version, no method has been found that can stably answer the correct question.
Some entrepreneurs who focus on AI writing have found that the style of the new version of ChatGPT seems to be slightly similar to Claude. This change started on August 8. At the same time, AI jailbreak master Pliny noticed that he needed to modify the long-used jailbreak prompt words to adapt to the new version of the model, which indicates that the new version of ChatGPT has enhanced protective measures, although these measures are not impeccable.
Some entrepreneurs have discovered that the answering style of the new ChatGPT model is quite similar to the anonymous-chatbot in the large model arena. This has triggered speculation about whether OpenAI is conducting a large-scale performance art around the next generation model Strawberry/Q*. guess.
A photo of CEO Ultraman Benman with strawberries in his hair sparked collective employee participation. Tao Xu, a veteran employee, even made a Q shape with real strawberries, hinting at the connection between strawberries and Q*. Combined with the news that the president of OpenAI is taking a long vacation and Lianchuang has resigned, it seems that some major changes are brewing within OpenAI.
The release of GPT-4o and a series of subsequent events have triggered many speculations about OpenAI’s internal dynamics and future development direction. Its calm release is in sharp contrast to the heated internal discussions, which may herald a new era in the field of artificial intelligence. of change.