Promt is a request for a neural network, from the English "prompt". Some people communicate with her so masterfully that they almost immediately get a relevant result. Moreover, employers have already appreciated such a valuable skill — the profession of an industrial engineer, a person who speaks correctly with artificial intelligence, has appeared.

1. Use context Context helps both people and neural networks. For example, there is a difference between: "Write a poem about winter" and "Write a poem about a cold and gloomy winter in the style of Edgar Allan Poe.""Write a letter on how to raise a puppy" and "Write a ready-made letter on behalf of a dog handler in a friendly and informal tone about how to raise a three-month-old shepherd puppy."The context may include: The tone of the letter: formal, professional, friendly.Text type: blog article, video description, email.The role of the writer: specialist, client, customer.The more context there is, the more accurate the result and the less need for clarification.
2. "Write a poem about winter" and "Write a poem about a cold and gloomy winter in the style of Edgar Allan Poe."
3. "Write a letter on how to raise a puppy" and "Write a ready-made letter on behalf of a dog handler in a friendly and informal tone about how to raise a three-month-old shepherd puppy."
4. The tone of the letter: formal, professional, friendly.
5. Text type: blog article, video description, email.
6. The role of the writer: specialist, client, customer.
7. Offer examples Neural networks are trainable. For the best result, offer a ready-made example. This is useful for: Scripts, answers to questions, descriptions of photos.Writing code and other tasks where the machine runs faster than a human.
8. Scripts, answers to questions, descriptions of photos.
9. Writing code and other tasks where the machine runs faster than a human.
10. Specify the length immediately Computers love precision. Instead of "write about cryptocurrencies," it is better to say "write a text about cryptocurrencies for a telegram channel up to 500 characters long." Specify the length: "from 500 to 700 characters" or just "short" or "long". This will help you get a more accurate result.
11. Supplement the promt with important information Example: For a motivation letter, upload a resume to create unique content. Neural networks can be given an article and ask for a 500-character sammari. This is useful considering that GPT-3 and GPT-4 use data up to 2021 and need up-to-date information.
12. Use phrases-life hacks Direct the neural network using phrases: "Describe step by step" — for structuring text, writing code, or solving problems."Write in the style of [Pushkin], [Kharms]" — to create parodies or posts."On behalf of [a professional dog handler], [the head of the department]" — for letters, mailings, advertising posts.
13. "Describe step by step" — for structuring text, writing code, or solving problems.
14. "Write in the style of [Pushkin], [Kharms]" — to create parodies or posts.
15. "On behalf of [a professional dog handler], [the head of the department]" — for letters, mailings, advertising posts.

Bad: "Write an article about machine learning."

 Good: "Write a structured article about machine learning for a blog of 700 to 1000 words in length, paying special attention to the basic concepts, methods and applications in real projects. Include examples of the use of machine learning algorithms in the fields of medicine, finance and technology to illustrate their importance in the modern world."

The main thing in making promts is to understand the purpose of the text and specify the request. The more information the neural network receives, the better the result will be.