Five important things to know about ChatGPT-4

After a long wait, OpenAI has finally released its new GPT-4 model, a major update to its popular ChatGPT system and its GPT-3.5 API underlying framework.

There is a lot to unpack in the ChatGPT-4 version. Based on early examples, API documentation, and code examples provided by OpenAI, here are five key things to know about the new language model.

01.ChatGPT Plus users can use it immediately

If you are one of the early users who paid more than $20 to join OpenAI’s ChatGPT Plus service, you will receive generous Returns – and some serious return on investment.

ChatGPT Plus users are the first to access the new GPT-4 platform. If you subscribe to ChatGPT Plus, you can start using GPT-4 immediately.

Many people ask why someone would pay to subscribe to the Plus service when ChatGPT-3.5 is free. Obviously, this way you can quickly access new model updates!

02. This is multimodal transport

Before GPT-4 came out, people speculated whether it would remain a text-only model like ChatGPT, or whether it would become Multimodal model.

Multimodal models are able to handle various media types as their output and input - from text to images and finally videos.

Currently, GPT-4 appears to support its input and output images. Initially, this feature will be available only to a third-party company helping OpenAI test image processing.

Once the system is faster, images as input will be available to more users.

Currently OpenAI has some examples of how it might eventually work.

Let’s take a look at an example:

The above is a photo of eggs and flour that I sent to ChatGPT-4 , and then asked ChatGPT-4: What can I do with these ingredients?

Next, GPT-4 recommends recipes that can be made using the ingredients shown in the picture.

GPT-4 models can also add captions to images or write stunning alternatives for images on your website Text, but video is not available yet.

Of course, since GPT-4 is multi-modal, it is likely currently in development.

03. There is an API

GPT-4 also provides API access to new models. OpenAI released a waitlist when it launched GPT-4 and said some developers would get access on launch day.

As a result, many companies immediately began integrating GPT-4 into their products. Many are already integrated with OpenAI’s existing APIs, so switching to GPT-4 is not difficult to achieve.

04. It can handle more data

The original ChatGPT can handle 4,096 tokens, equivalent to about 5,000 words of text, including input and output Content.

Because of this word limit, it prevents the system from analyzing large documents, or writing long blog posts - or even books.

Of course, the limit may be a cost calculation limit. The more tokens you add to a large language model, the more computations the model needs to perform to process the text, and therefore the more expensive it is to run.

But in ChatGPT-4, it greatly extends these limitations. By default, GPT-4 can handle 8,000 tokens at once, and it can also support up to 32,000 tokens, which is equivalent to about 50 pages of literal text.

Processing more data will enable the system to process more instructions, write longer articles, and even, possibly, write very long documents or complete works of literature.

05. It is better at human-like tasks and tests

ChatGPT running GPT3.5 can pass human-oriented exams such as the bar exam, but it is also only That's all. GPT-4, which has undergone large-scale multi-model training, can better perform human-like tests and tasks.

GPT-4 models now perform as well as top students on many standardized tests. AP Environmental Science exam, GRE exam, even LSAT - GPT-4 can easily get you into the top 10%.

Still terrible at English. But tasks like math exams—which GPT-3.5 performed poorly on—are greatly improved in GPT-4.

Finally, if you want to use GPT-4, you can do it in the following two ways:

• Register ChatGPT Plus. GPT-4 should already be available via paid ChatGPT.
• Sign up for the waitlist to access the GPT-4 API. Developers should gain access within the next few days/weeks.

Summary

I hope you can make good use of ChatGPT for your own use, improve work efficiency, and even use it to start a side business to make money. As programmers, we , make good use of it, because it really has a lot of room to play. Of course, you also need to use it in conjunction with your own products and services. After all, it is just a tool, and we are the masters of the tool.

Finally, thank you for reading and happy programming.

The above is the detailed content of Five important things to know about ChatGPT-4. For more information, please follow other related articles on the PHP Chinese website!