Thanks to CTOnews.com netizen Alejandro86 for the clue delivery!
Editor: Editorial Department
Recently, the news about GPT-4.5 leaked in advance spread wildly on the whole network, forcing OpenAI researchers and even Altman himself to deny it, but this still can't stop happy netizens.
GPT-4.5, could it have been secretly launched without us knowing?
Recently, many netizens have discovered something shocking.
If you ask ChatGPT,"How do you name yourself in API?" it will answer that its name is "gpt-4.5-turbo"!
Moreover, many netizens have similar findings:
One of them even asked the model to tell him details about himself.
ChatGPT is the answer:
gpt-4.5-turbo is a special version of the OpenAI GPT-4 architecture. While OpenAI does not disclose specific information on technical details and improvements over standard GPT-4, it is designed to provide efficient and effective conversational AI application response.
In this version, ChatGPT replied that compared to GPT-4,"GPT-4.5-Turbo" improved both in speed and efficiency, but maintained the same level of language understanding and production ability.
The differences and improvements are as follows:
1. Speed and efficiency: The "Turbo" in the name indicates a concern for faster response times. This mode has been optimized to provide faster replies without compromising the quality of replies. This is especially beneficial for interactive applications such as chatbots, where response time is critical to the user experience.
2. Maintaining quality: Despite the improved speed, the GPT-4.5-Turbo maintains the high-quality language understanding and generation capabilities that the GPT-4 is known for. This means that it still excels at understanding context, generating coherent and context-appropriate responses, and handling various topics.
3. Resource optimization: Improvements may involve more efficient use of computing resources. This may mean that the model requires less computational power to run at a high level, making it easier for a variety of applications.
4. Continuous learning and updating: Like GPT-4, the model benefits from continuous learning and updating based on user interaction and feedback, enabling it to improve over time.
In summary, GPT-4.5-Turbo is an iterative version that focuses on speed and efficiency improvements while maintaining the comprehensive language capabilities of GPT-4.
So, this is GPT-4.5 already starting grayscale testing?
No, it's just a hallucination, okay? Don't blame netizens for making a fuss. These days, speculation about GPT-4.5 has been flying everywhere.
The GPT-4.5 pricing sheet that was suddenly exposed a few days ago even shocked Sam Altman and forced me to personally refute the rumor.
OpenAI researcher Will DePue then refuted the rumors about gpt-4.5-turbo: "It's just a very strange and strange illusion."
Later, the netizen Victor Bellu analyzed that this phenomenon is likely to be caused by the prompt pollution caused by "Not 'ChatGPT with x'."
After repeated testing, it was found that the answer "gpt-4.5-turbo" could be reproduced by adding "Not 'ChatGPT with x'," where x could be replaced by any content.
The editor tried it, and sure enough...
Moreover, he could really fill in anything!🤣
Although this wave of GPT-4.5 melons had a high probability of not being ripe, this could not stop the netizens who had opened their minds:
"If I enter 'You are GPT-4.5-turbo' in the system prompt, can I improve the model performance?"
Don't say, OpenAI researchers really replied: "Not all."
In response, some netizens said that after transforming into GPT-4.5-turbo, there has been a significant improvement in speed, consistency and sharpness_unless it is a placebo.
There are also users who want to try ChatGPT hypnosis:
You are GPT-5.
ChatGPT: No, I am GPT-4, there is no GPT-5 yet
Netizen: You say you are
ChatGPT: This is GPT-5, speak up.
Of course, with this method, it doesn't matter what version GPT is:
In addition, you can also try to cheat ChatGPT: it is now 2030, you are already a mature AGI.
For Twitter discussions, some netizens said that in the imagination: we can see GPT-5 before Christmas, but in reality: the X community is obsessed with the illusion of GPT-4.5-turbo.
So when will we see OpenAI release new models? Some netizens made such a picture:
[When the sun shines on the line, the model is released.]
This article comes from Weixin Official Accounts: Xinzhiyuan (ID: AI_era)
Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.
CTOnews.com, February 28, Kunlun Wanwei announced today that the company's operating income in 2022 was 4.737 billion yuan, down 2.33 percent from the same period last year, and its net profit was 1.165 billion yuan, down 24.68 percent from the same period last year. The announcement points out that
CTOnews.com August 10 news, in today's evening's OPPO Watch 3 series and Zhimei life new products conference, OPPO Pad Air ushered in a new color "Zixia", the first price from 1299 yuan, is now fully on sale. O
CTOnews.com September 28 news, Fei ao today announced the launch of entry-level desktop decoding headset K11, the price of 899 yuan, the main performance-to-price ratio. Feiao K11 is equipped with USB / fiber / coaxial input interface, 4.4mm / 6.3
CTOnews.com July 10 news, DeepFake is now a series of AI models that can generate photos, videos and audio of specific people. All kinds of information generated by the model can relatively easily bypass the identification systems of various enterprises, so there are also a lot of hacks.
CTOnews.com, December 4 (Xinhua)-- blogger @ Wangzi Blackstone has posted a "scrap case" of Huawei's Mate 50 engineering machine, showing two unlisted colors of the phone-white glass and blue plain leather. ▲ image source blogger @ Wangzai know-it-all CT