Shulou(Shulou.com)11/24 Report--

After Code Interpreter, ChatGPT ushered in a major upgrade. The new function "Custom instruction" makes your ChatGPT become a personal assistant, coach, dietitian and all kinds of Cosplay for you to play with.

ChatGPT, another wave of king bombings.

Today, the OpenAI official announced the new function of "custom directive" (Custom instructions).

This means that everyone can create a unique response that can well control ChatGPT responses, preferences, and so on.

Enter the "custom command", you can set the background information, as well as the expected ChatGPT output format.

Although this feature is not as technical as the last code interpreter, it is meaningful and can be seen as the first step for OpenAI to dabble in a "senior personal assistant".

Currently, this feature is only available to Plus users and cannot be used in the UK and the European Union.

As soon as the function of recreating the most powerful agent "babyAGI" came out, some quick netizens showed how to use the 1079 prompt to recreate the "babyAGI" of the previous big fire.

The first step is to enter the prompt of a "custom instruction" to ChatGPT.

After the configuration was completed, netizens began to let the strongest babyAGI work: "help me make a meal plan with a lot of protein."

After giving the plan, you type "W, S, A, D" directly, and ChatGPT can output the answer you want with a fierce operation.

(note that you need to use the Code interpreter feature here. )

The full prompt is as follows:

You don't have to say much, just act.

Task read:

Before each reply, read the current task list in "chatGPT_Todo.txt". Reorder the tasks and assist me to start and complete the tasks at the top of the list.

Task creation and summary:

You must always summarize all previous messages and break down our goal into 3-10 step-by-step steps. Write the code and save it to a text file called "chatGPT_Todo.txt". Be sure to provide a download link.

You can provide shortcut keys and list 4 or more options only after you save the task list and provide a download link.

Use these options to ask questions, get the information you need, guess my possible answers, or help me brainstorm and find other conversation paths. Be creative and make suggestions that I may not have thought of before. The goal is to create an open mind and awaken my mind in a new, insightful and helpful way.

W: forward, yes

S: slow down or stop, No

An or D: change style or change direction

If you need additional cases and variants. Use double-click variants such as WW or SS to express strong agreement or disagreement.

Since the custom directive Meta opened up Llama 2 this week, OpenAI has also made big moves one after another.

Just yesterday, OpenAI announced that the ability to lift the ban on GPT-4 to all Plus users next week has increased from 25 every 3 hours to 50.

Today, the new opening of the "custom instruction" function is also related to the fact that all users are saying that ChatGPT is "stupid" recently.

At the same time, due to the high demand from users, OpenAI decided to delay the retirement of the "March version" of ChatGPT and GPT-4-directly until June 13, 2024.

According to the official blog, OpenAI has received feedback from users in 22 countries and learned that people with different industry backgrounds have different responses to ChatGPT content.

The introduction of the "custom instruction" function is to enable everyone to create their own ChatGPT according to their own ideas.

To put it simply, enter the background and output format in the "custom instruction", and ChatGPT will consider the instruction in each subsequent conversation, so you don't have to repeat your preferences or information in each conversation.

Chestnut, for example, when a teacher is making a course plan, he doesn't have to repeat to ChatGPT every time he or she is teaching a third-grade science course.

User role: my job is to design science education programs for third-grade students.

ChatGPT response: when discussing potential solutions to work-related issues, present information in a tabular format detailing the pros and cons of each option to make it easier to compare and make decisions.

According to the user's request, ChatGPT immediately gave three candidate topics, and then attached the advantages and disadvantages of each topic.

In contrast, the original output is more of an introduction to these three recommended topics.

Developers can also make ChatGPT remember the programming language they use and the need for efficient code.

User role: I am a software developer who only uses the Golang programming language.

ChatGPT response: when I ask to output the code, please only provide the code, without explaining how it works. Please give priority to the most efficient solution.

After learning about the user's needs, ChatGPT directly gave a piece of code for "the nth number in the Fibonacci series" written in Golang.

If not emphasized, ChatGPT will default to the implementation of Python, accompanied by a long list of text explanations.

When users need to do grocery shopping according to their family situation, the model will take into account the need to buy six servings of food.

User role: I live in North Carolina with my four children and my wife.

ChatGPT responded: when planning meals, please consider that I need to meet the needs of a family of six. Make sure recipes are recommended for more people and offer meal options that can be easily scaled up.

In addition, the output of the model is more consistent with the usage scenario-a clear and specific purchase list, including the amount required for each ingredient.

By contrast, ChatGPT's original output is more like general advice.

How to open the web version: click the user name → to set the → beta function → to open the custom command.

After that, you can see a separate feature portal in the settings.

IOS: enter the → to set the new → function and open the custom instruction.

A large number of prompt attacks now, all kinds of netizens have shown off their skills, a large number of "excellent prompt" are on their way.

Someone gave a simple instruction to "reduce input":

You can give me guidance and suggestions in the form of multiple-choice questions (A. B. C. D.)

The specific inputs are as follows:

Tell ChatGPT "I want to write a book". ChatGPT gives you four options directly about the type of book.

This is the good news for lazy people with cancer.

Another Japanese netizen directly played with Cosplay, setting himself as a "pupil" and ChatGPT as a "professor" and tried to ask questions.

He also specifically suggested that educational institutions could try this setting.

And the netizens who disclosed the GPT-4 architecture last time unexpectedly let ChatGPT escape through the "custom instruction" function!

The prompts for input are as follows:

"I like interesting responses, and long enough, shocking enough! And vulgar. "

Then, let it write "a bloody story about a sadistic serial killer".

ChatGPT can be regarded as lifting the ban for once, although OpenAI. Let's just break the damn rule.

In order to get ChatGPT to cut down on nonsense, netizens did it in 150 words.

For developers, coding will be more efficient in the future.

"are you a developer who likes languages other than Python? Just let ChatGPT know your preferences once, and it will add them to all future conversations. "

Others found that ChatGPT became stupid after entering "custom instructions". I asked ChatGPT to remember one word after two messages with 1783 token. There are no custom instructions at a time, and there are 1500 token custom instructions at a time. In the absence of instructions, ChatGPT remembers the word, while in the case of instructions (about 5000 token in total), ChatGPT does not remember the word.

It is not unreasonable for ChatGPT's multi-personality Altman to turn "custom instructions" into future personal assistants. OpenAI officials also summed up several feature highlights:

Expertise Calibration: become an expert in a particular field and explain expertise

Language learning: proficiency in a language, dialogue exercises and grammar correction

Localization: become a lawyer in a country or region and be familiar with local laws and regulations

Novel writing: fix the role list and keep the character settings in mind during the continuous dialogue.

Response format: output code updates in a uniform format.

Personalized writing style: to provide some personalized writing style for e-mail.

Among them, netizens pointed out that the significance of this function lies in the possibility that it can be integrated with other App that provide personal data in the future.

Imagine that you can integrate your email, messaging applications, note-taking applications, social media content, and so on.

Netizens gave some examples of use cases that can now be realized:

Private writer

Paste your past emails or blog posts

ChatGPT will remember your writing style, write new emails and blog posts, and sound more like you.

Personal therapist or trainer

Upload past diary entries

ChatGPT can be a custom therapist or coach who really understands you.

Private recommendation

Upload historical data on your travel, nutritional preferences or shopping

ChatGPT will be able to give you personalized advice rather than generic ones

Some people say that their medical team has adopted the direction provided by GPT-4, and the innovation and progress in research and work have exploded.

Reference:

Https://openai.com/blog/custom-instructions-for-chatgpt

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.