How to use chatbots well: Ask the right questions, and the answers will be amazing

News on April 13th, the recently popular chatbot ChatGPT can provide very high-quality answers, but only if you know how to ask it the right questions. So, how do we enter the correct prompt? American media provides many proven and effective techniques.

The following is the translation:

Nowadays, we can use generative artificial intelligence tools, such as ChatGPT, and only need to enter some prompt words to get the answer. However, that doesn't mean these answers are always accurate. The key is that you need to enter the prompt information correctly.

Can you communicate with the latest artificial intelligence systems using simple vocabulary? Peter Lin, a computer science professor at the University of Waterloo, said the idea was "very deceptive."

Many companies today are hiring so-called prompt engineers, people with the skills and experience to converse effectively with large-scale linguistic artificial intelligence systems. Many job seekers also hope to master this skill.

Professional social network LinkedIn said that the number of global members using terms such as ChatGPT, prompt engineering, and generative​​AI​​in their profiles in February increased by 55% compared with January. , which increased by 71% in March compared with February.

Peter Lin said: "We now have 'dog whisperers' and 'horse whisperers'. Now we are developing whisperers for large language models."

Generally speaking, users The more detailed the prompts entered, the closer the response given by ChatGPT will be to the results the user wants. But the most effective tips aren’t just about adding details.

Figure 1: Tips for writing better, you can add some specific formatting requirements, such as writing style, list or summary, etc. Here’s an example:

Basic tip: “What are ETFs?”

Better tip: “How do you explain ETFs to a layperson ?Can you give me three examples to help me understand it?"

First of all, in order to get the best response, it is crucial to add the right details. And, there are many ways to instruct an AI system on how to present the information it finds, depending on the goal. For example, you can tell it to provide answers in a certain style or for a specific audience. Each tool also has specific capabilities and limitations that require the user to learn how to use. And the best results often come from long conversations with the AI, reacting to the content it produces and asking for improvements.

Brianne Kimmel, founder of venture capital fund Worklife, said that even in applications that don’t explicitly mention “prompt engineering” job positions, having this skill can make a lot of sense. Required skills for the position. For example, art directors can use generative AI to sketch out ideas, marketers can use it to create new slogans, software engineers can use it to spot problems in code, and lawyers can use it to study the letter of the law.

Worklife is a venture fund that invests in companies using generative artificial intelligence.

"This is going to be an essential skill that all of us need to know," Kimmel said. "Because, once you learn a new skill, you can apply it anywhere you like."

Figure 2: Basic Tip: "Help I write a love poem for my partner."

Better tip: "Write a love poem for my partner who loves mystery. His favorite The poem is "The Raven" by Edgar Allan Poe."

Basic Tip: "Make a five-day vegetarian plan."

Better Tip: “Prepare vegetarian recipes for the next five days and make a shopping list.”

Essential Job Skills

Sophie Antebi will graduate from the University of Southern California in May with a degree in marketing and entrepreneurship and is interested in working at a startup. To make her resume stand out, she learned to use a variety of generative AI tools. Antebi believes these tools can be useful when planning marketing campaigns or creating images for presentations.

Although Antebi hasn’t found a job yet, she said her application has gotten more responses since adding her familiar “prompt engineering” and other artificial intelligence skills to her cover letter. She credits these skills with helping her make it to her second and third round interviews. Speaking of his interviewers, Anteby said: "They asked a lot of relevant questions and were very interested in these skills."

Boston Children's Hospital is hiring Prompt Engineers. The hospital posted job postings in February and is currently interviewing applicants. John Brownstein, the hospital's chief innovation officer, said the prompt engineers will work in the hospital's innovation and digital health accelerator division, which has continued to invest in artificial intelligence over the past few years.

Dr. Brownstein said that large-scale language artificial intelligence tools will play a role in many tasks in hospitals, which may include analyzing large amounts of unstructured clinical records, detecting medical patterns that may otherwise be ignored, providing health knowledge to patients, and Improve patient admission efficiency. He added: "There is definitely an element of hype around AI technology like ChatGPT, but that doesn't stop us from adopting it."

London law firm Mishcon de Reya also plans to hire an AI prompt engineer, its chief strategy Chief executive Nick West said large-scale language AI tools will most likely be used in the first steps of litigation, which is summarizing legal documents, finding relevant case law and querying information databases.

West said: "I believe this is a transformative technology. So any help can accelerate our understanding and understanding of it, which also means we can take advantage of it faster." ”

Elvis Saravia, co-founder of Dair.ai, said that even without full-time AI prompt engineers, technical teams need to be familiar with artificial intelligence tools. This is becoming increasingly apparent. Dair.ai is an organization dedicated to democratizing artificial intelligence tools through open source research, educational materials, and training. Sarabia quickly released an engineering guide on how to build and interface with large language models.

Picture 3: Basic question: "Make a plan to stay in Paris for three days."

More Good tip: “Make a complete plan for travelers’ three-day stay in Paris, including hotels for under $500 per night and gluten-free dining options.

Collaboration The Art of

Researchers and users have discovered and documented many hinting methods. One of them is telling a large language model "let's think step by step", which allows the technology to break down complex requests into Smaller problems that are easier to handle. Another approach is to tell the AI ​​that it is an expert in a field and its job is to provide advice to laypeople. This can help simplify the answer to complex problems.

Some researchers and users are publishing guides and tips they find. There are even users selling their tips on the marketplace, and others are developing tools that can optimize tips. But many resources are freely available in Twitter posts and YouTube videos Share. Includes resources posted by Nick St.Pierre, a designer who creates content focused on explaining Midjourney and other art-making AI tools.

For artwork, We need to combine experience with the tool and knowledge of artistic style to refine the prompt. Pierre recommends starting with a basic concept and then adding to it as needed. For example, he said the prompt might start by asking the AI ​​to provide a 40 Photo of a 20-year-old brunette woman sitting at a bar. If the resulting image is not what the user wanted, they can add terms to change factors such as camera angle, clothing style, and lighting to get a different result.

Pierre has found that the elements that come first in the cue usually stand out more in the final image. So he tends to list what he thinks is most important in the visual first. “We need to iterate a lot,” he says. , testing, and working closely with AI tools. ”

How to give better tips on ChatGPT?

1) Understand the functions of ChatGPT

Although ChatGPT can do a lot of things, but it's not omnipotent. For example, it's just a language model and not good at math. Furthermore, ChatGPT is currently only trained until 2021, so any events after that are unknown to it .

OpenAI didn’t disclose what data it used to train ChatGPT, but the company’s science communicator Andrew Mayne recommended that people read what others are doing with ChatGPT and where they are at. has been successful. He also encourages people to try it out for themselves to figure out how to make it reliably do what people want it to do.

2) Make sure the prompts are accurate and clear

Inaccurate wording can confuse ChatGPT just as much as it confuses people. Vague wording, typos, and missing words can all contribute. Homophones can be particularly troublesome: without clear context Under this situation, artificial intelligence may be able to recognize the word, but it will create ambiguity. For example, "Right" can mean both direction and "right".

3) Adjust the prompts

ChatGPT will refine its responses based on new input from users, from requests for additional content to specific instructions on formatting. For example, users can ask it to display its responses in summary form, or try to present its responses in a more serious or well-known format. Answers in the tone of writers (like Mark Twain).

4) Ask for help

If your initial query doesn't work, you can let ChatGPT help you design Better tips. For example, you could say: “What do I need to tell ChatGPT to help it write a good cover letter for me? "This artificial intelligence tool will tell you the information it needs. (Xiao Xiao)

The above is the detailed content of How to use chatbots well: Ask the right questions, and the answers will be amazing. For more information, please follow other related articles on the PHP Chinese website!