Table of Contents
On March 14th, OpenAI’s official website released the latest GPT-4. It is currently only available to ChatGPT Plus users, with a limit of 100 messages within 4 hours (of course it is enough). ChatGPT Plus users finally realized the promise of priority experience this time. Free users and ChatGPT API users currently cannot use GPT-4.
For this reason, after the announcement, a large number of users poured into OpenAI’s official website to pay to upgrade to Chatgpt Plus. This is overwhelming the payment system.
Of course, if you are already a ChatGPT Plus user, then your ChatGPT account can already support the selection of the GPT-4 model.
What is OpenAI GPT-4?
GPT-4 is the flagship of OpenAI’s language models, heralded as a major breakthrough in deep learning. It offers greater accuracy and trustworthiness than its predecessors with added multimodality to boot. Beyond being able to process images, it has demonstrated remarkable creative ability that could make complex tasks more possible for almost any application. As we can read in the official product update, “it can generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style.”
For now, the GPT-4 comes in two model variants:
gpt-4-8Kgpt-4-32K
And they are only available to ChatGPT Plus users.
GPT-4 Capability
Let’s take a look at the changes in GPT-4 first. First look at a comparison chart:
It can be seen that compared with GPT-3.5, GPT-4 is a new generation of multi-modal large models. GPT-4 supports not only text but also image input.
GPT-4 Features
1. Excellent language comprehension ability
GPT-4 has greatly improved in language understanding, and can better understand complex and ambiguous contexts, thereby providing more precise answers.
2. Improvements in context understanding
GPT-4 strengthens the ability to understand long-form dialogue scenarios. Compared with GPT-3.5, it has better memory and contextual understanding, enabling it to better track the topics of contextual dialogues and give relevant answers.
3. Expansion of professional domain knowledge
GPT-4 has greatly improved knowledge and skills in specific fields, such as science, medicine, law, etc., allowing users to obtain more professional advice. Even in the official bar qualification test, it surpassed 90% of humans. In contrast, the score of GPT-3.5 can only be ranked in the bottom 10%. After I see it, we can use GPT4 as an encyclopedia.
4. Better adaptability and personalization
GPT-4 can better adapt to users’ needs and styles, providing a more personalized interactive experience.
5. More powerful generation ability
GPT-4 has higher creativity and naturalness in text generation and can generate more convincing and attractive content.
6. Stronger ethics and values review
GPT-4 introduces new mechanisms for better screening of unethical or harmful content to avoid generating inappropriate or controversial responses.
GPT-4 astonishes me with its linguistic prowess! According to the official description, the Chinese accuracy of GPT4 can approximately reach the English level of GPT3.5. It has mastered 25 languages, including Mandarin, Polish, and Swahili Among them, GPT-4’s Chinese can reach 80.1% accuracy, while GPT-3.5’s English accuracy is only 70.1%, and GPT-4’s English accuracy has also increased to 85. %.
There are bloggers on Twitter who used the complicated prompts he had prepared for the first time and tried it. What was not possible in 3.5 before, can now be completed accurately with GPT4. It is hard to imagine how powerful GPT-4 has the ability to understand the logic and execute rules.
Some people are worried that with the advent of GPT-4, the verification codes of major websites will be impacted first because the AI headed by GPT-4 can already understand pictures, which is terrible!
Will all verification codes in the world become invalid overnight? However, the joke now is: Although GPT-4 has been released, the GPT-4 API has not been released, so the verification code is temporarily worry-free.
Performance in programming
Let’s look at programming again.
- GPT-4 can support the input of 32768 Token lengths, supplemented by a better Prompt.
- GPT-4 can understand the logic of image expression and the ability to combine text, and can directly convert a website design sketch into front-end code.
- The programming ability of GPT-4 is stronger, and the error information can be directly input back. With the mechanism of react, it can automatically debug and output accurate code.
- Programmers, if you haven’t upgraded your ChatGPT Plus membership yet, you can register an OpenAI account first and apply to join the GPT-4 waiting list. For details, see below.
Apply for the GPT-4 waiting list
If you are not a member of ChatGPT Plus, you will not be able to experience GPT-4 in the short term. However, the OpenAI official website has opened the application portal for GPT-4 API, simply fill in your personal information and submit The purpose of using GPT-4, and wait patiently after submission. Once passed, you will get access to the GPT-4 API earlier than anyone else.
In the past two weeks, the development of AI has been too fast. Now that you haven’t applied for an OpenAI or ChatGPT account, hurry up and apply for one. Help my articles can help you! If you have any questions, welcome to let me know. We can discuss it together!