{"id":930,"date":"2023-10-27T18:04:42","date_gmt":"2023-10-27T10:04:42","guid":{"rendered":"https:\/\/www.1ai.net\/?p=930"},"modified":"2023-10-27T18:04:42","modified_gmt":"2023-10-27T10:04:42","slug":"chatgpt-%e5%9f%ba%e7%a1%80%e7%94%a8%e6%b3%95","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/930.html","title":{"rendered":"Basic usage of ChatGPT"},"content":{"rendered":"<h2 id=\"\u57fa\u672c\u6cd5\u5219\" class=\"anchor anchorWithStickyNavbar_LWe7\">fundamental rules<a class=\"hash-link\" title=\"Direct links to basic laws\" href=\"https:\/\/www.learnprompt.pro\/docs\/basics\/prompting#%E5%9F%BA%E6%9C%AC%E6%B3%95%E5%88%99\" aria-label=\"Direct links to basic laws\">\u200b<\/a><\/h2>\n<p>Compared to a search engine,<a href=\"https:\/\/www.1ai.net\/en\/tag\/chatgpt\" title=\"[View articles tagged with [ChatGPT]]\" target=\"_blank\" rel=\"noopener\">ChatGPT<\/a>The strength of this is its ability to efficiently correlate ideas and summarize information. Before delving further into the idea of constructing hints, I want you to be aware of two general rules of thumb that can be used to improve the quality of the output of generated AI models. These include<\/p>\n<ul class=\"\">\n<li>Try different representations of the prompt to get the best results<\/li>\n<li>Use clear and concise prompts and avoid unnecessary words<\/li>\n<li>Reduce imprecise descriptions<\/li>\n<\/ul>\n<p>Of course, it's not all about rules of thumb. What we want is for you to be able to follow the examples step-by-step, not a list of things to memorize over and over again. Also, you don't need to follow all the rules in every tip you write at the beginning. For simple tasks, you only need<strong>Instruction (task explanation) + problem<\/strong>You can construct a prompt that works pretty well.<\/p>\n<p>Here we directly use a few simple examples to illustrate.<\/p>\n<h2 id=\"\u4f8b\u5b501-\u6587\u7ae0\u6458\u8981\" class=\"anchor anchorWithStickyNavbar_LWe7\">Example 1 Article Summary<a class=\"hash-link\" title=\"Example 1 Direct link to article summary\" href=\"https:\/\/www.learnprompt.pro\/docs\/basics\/prompting#%E4%BE%8B%E5%AD%901-%E6%96%87%E7%AB%A0%E6%91%98%E8%A6%81\" aria-label=\"Example 1 Direct link to article summary\">\u200b<\/a><\/h2>\n<p>I believe this is one of the common functions that many of us use. The extraction and summarization of long text can help us quickly understand the main content.<\/p>\n<p>Here we have selected a news article related to GPT4 and asked ChatGPT to help us summarize the content:<\/p>\n<h3 id=\"\u63d0\u793a\u8bed-1\" class=\"anchor anchorWithStickyNavbar_LWe7\">Tip 1<a class=\"hash-link\" title=\"Direct link to Tip 1\" href=\"https:\/\/www.learnprompt.pro\/docs\/basics\/prompting#%E6%8F%90%E7%A4%BA%E8%AF%AD-1\" aria-label=\"Direct link to Tip 1\">\u200b<\/a><\/h3>\n<div class=\"codeBlockContainer_Ckt0 theme-code-block\">\n<div class=\"codeBlockContent_biex\">\n<pre class=\"prism-code language-text codeBlock_bY9V thin-scrollbar\" tabindex=\"0\"><code class=\"codeBlockLines_e6Vv\"><span class=\"token-line\"><span class=\"token plain\">Summarize the paragraph in one sentence:<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">Fox News reported on April 2, according to a tech industry insider and expert, that OpenAI is expected to launch GPT-5 later this year, which could make generative AI indistinguishable from humans.<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">I got word that GPT-5 is scheduled to finish training this December, and OpenAI hopes it will achieve AGI,<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">AGI stands for<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">Chen added, \"This means that we will all have a lively discussion about whether it actually achieves AGI. This means that it will.\" Chen went on to say that he's not saying that there's a consensus within OpenAI that upgrading to GPT-5 will mean that AGI will be realized, but 'there are quite a few people who believe it will.'<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">OpenAI last month released the GPT-4 system, the company's latest deep learning model, which \"demonstrates human-equivalent performance on a variety of professional and academic benchmarks.\"<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">More than 2,000 tech leaders, university professors, and others, including Elon Musk and Apple co-founder Steve Wozniak, have signed an open letter calling for a moratorium on the training of systems more powerful than GPT-4 in all artificial intelligence labs. The letter calls for a six-month moratorium on experiments while warning that AI systems that are as intelligent as or more intelligent than humans pose far-reaching risks to society and humanity.<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">OpenAI did not immediately respond to Fox News' request for comment on Chen's tweet.<\/span>\r\n<\/span><\/code><\/pre>\n<div class=\"buttonGroup__atx\"><\/div>\n<\/div>\n<\/div>\n<p>ChatGPT generation result 1<\/p>\n<blockquote><p>OpenAI plans to launch GPT-5 later this year, which is said to potentially enable general-purpose AI, sparking discussion about the distinction between AI and humans. Chen Siqi revealed that OpenAI hopes that GPT-5 will enable AGI, which means that it will be able to understand tasks and concepts as well as humans. Previously, more than 2,000 tech leaders, university professors and others had called on all AI labs to suspend training of systems more powerful than GPT-4, warning that the transcendence of AI systems could pose far-reaching risks to society and humanity.<\/p><\/blockquote>\n<h3 id=\"\u63d0\u793a\u8bed-2\" class=\"anchor anchorWithStickyNavbar_LWe7\">Cue 2<a class=\"hash-link\" title=\"Direct link to Tip 2\" href=\"https:\/\/www.learnprompt.pro\/docs\/basics\/prompting#%E6%8F%90%E7%A4%BA%E8%AF%AD-2\" aria-label=\"Direct link to Tip 2\">\u200b<\/a><\/h3>\n<p>ChatGPT has shortened our reading time by condensing the original long text into 3 statements through Prompt 1. So, is there a better way to formulate the prompt for the summarization task? Let's try another expression for the same task.<\/p>\n<div class=\"theme-admonition theme-admonition-note alert alert--secondary admonition_LlT9\">\n<div class=\"admonitionHeading_tbUL\">Remark<\/div>\n<div class=\"admonitionContent_S0QG\">\n<p>To minimize unnecessary reading, I have replaced the repetition with ellipses in the subsequent cue comparisons.<\/p>\n<\/div>\n<\/div>\n<div class=\"codeBlockContainer_Ckt0 theme-code-block\">\n<div class=\"codeBlockContent_biex\">\n<pre class=\"prism-code language-text codeBlock_bY9V thin-scrollbar\" tabindex=\"0\"><code class=\"codeBlockLines_e6Vv\"><span class=\"token-line\"><span class=\"token plain\">Summarize the content below:<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">According to Fox News on April 2, a tech industry insider and expert said that OpenAI is expected to launch GPT-5 later this year, which could make generative AI indistinguishable from humans... .<\/span>\r\n<\/span><\/code><\/pre>\n<div class=\"buttonGroup__atx\"><\/div>\n<\/div>\n<\/div>\n<p>ChatGPT Generation Result 2<\/p>\n<blockquote><p>OpenAI plans to launch GPT-5 later this year, potentially making generative AI indistinguishable from humans, according to Fox News on April 2nd. Chen Siqi, a tech entrepreneur and developer, tweeted that he got word that GPT-5 is scheduled to finish training in December of this year, and that OpenAI hopes that it will achieve AGI, which means that the AI system will be able to understand a task or concept as well as a human. He added that this means there will be discussions about whether it actually achieves AGI, and there are a number of people who believe it will. OpenAI released the GPT-4 system last month, and it has shown comparable performance to humans on a variety of professional and academic benchmarks. In addition, more than 2,000 tech leaders, university professors and others, including Elon Musk and Steve Wozniak, signed an open letter calling on all AI labs to suspend training of systems more powerful than the GPT-4 in order to avoid the far-reaching risks to society and humanity posed by AI systems that are as intelligent as, or more intelligent than, humans. OpenAI has not responded to this at this time.<\/p><\/blockquote>\n<h3 id=\"\u63d0\u793a\u8bed-3\" class=\"anchor anchorWithStickyNavbar_LWe7\">Cue 3\u200b\nSummarize the content in a simple statement, preferably one that can be in a few sentences.<a class=\"hash-link\" title=\"Direct link to Tip 3\" href=\"https:\/\/www.learnprompt.pro\/docs\/basics\/prompting#%E6%8F%90%E7%A4%BA%E8%AF%AD-3\" aria-label=\"Direct link to Tip 3\">\u200b<\/a><\/h3>\n<p>Let's try another expression.<\/p>\n<div class=\"codeBlockContainer_Ckt0 theme-code-block\">\n<div class=\"codeBlockContent_biex\">\n<pre class=\"prism-code language-text codeBlock_bY9V thin-scrollbar\" tabindex=\"0\"><code class=\"codeBlockLines_e6Vv\"><span class=\"token-line\"><span class=\"token plain\">Summarize the content in a simple statement, preferably in a few sentences.<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><span class=\"token-line\"><span class=\"token plain\">According to Fox News on April 2, a tech industry insider and expert said that OpenAI is expected to launch GPT-5 later this year, which could make generative AI indistinguishable from humans... .<\/span>\r\n<\/span><span class=\"token-line\">\r\n<\/span><\/code><\/pre>\n<div class=\"buttonGroup__atx\"><\/div>\n<\/div>\n<\/div>\n<p>ChatGPT Generation Result 3<\/p>\n<blockquote><p>OpenAI plans to launch GPT-5 later this year and hopes to achieve AGI, which could make generative AI indistinguishable from humans, a tech industry insider and expert said, according to Fox News. This comes after more than 2,000 tech leaders and university professors have called for a moratorium on training more powerful AI systems to avoid the far-reaching risks to society and humanity.<\/p><\/blockquote>\n<p>After the above comparison of different prompts in the same text, I believe it is not difficult for you to see that Generated Result 1 is not only more concise than Generated Result 2, but also more complete in terms of information content than Generated Result 3. This responds to the three rules of thumb we proposed above.<\/p>\n<ul class=\"\">\n<li>Try multiple representations of the prompt to get the best results:<strong>Different prompts give different results<\/strong><\/li>\n<li>Use clear and short prompts and avoid unnecessary words:<strong>Reduce the use of degree words such as 'best'<\/strong><\/li>\n<li>Reduce imprecise descriptions:<strong>Reduce the use of uncertainty and replace the word \"sentences\" with \"one\".<\/strong><\/li>\n<\/ul>\n<h2 id=\"\u4f8b\u5b502-\u4ee3\u7801\u751f\u6210\" class=\"anchor anchorWithStickyNavbar_LWe7\">Example 2 Code Generation<a class=\"hash-link\" title=\"Example 2 Code-generated direct links\" href=\"https:\/\/www.learnprompt.pro\/docs\/basics\/prompting#%E4%BE%8B%E5%AD%902-%E4%BB%A3%E7%A0%81%E7%94%9F%E6%88%90\" aria-label=\"Example 2 Code-generated direct links\">\u200b<\/a><\/h2>\n<p>More and more people wish to write code using ChatGPT. If you have a problem that you wish to solve by programming in code, you can specify the relevant programming language and task by using the prompts.<\/p>\n<p>Here we let ChatGPT automatically generate a Python code for quick sorting in a sorting algorithm.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-931\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2023\/10\/code-4db913231722f567c20c675ba68075d8.png\" alt=\"\" width=\"1402\" height=\"1354\" \/><\/p>\n<p>The thoughtful ChatGPT also provides us with test cases. We can manually verify the correctness of the code.<\/p>\n<p>I'm sure that with the two examples above, you have learned the role of prompts. So, how can you create prompts that generate the best results in actual tasks? This is the focus of prompt engineering and the focus of this course.<\/p>","protected":false},"excerpt":{"rendered":"<p>In comparison to the search engine, the strength of ChatGPT lies in its ability to efficiently link ideas and aggregate information. Before going further into the construction of the hints, I would like you to know two common lessons of experience that can be used to improve the output quality of the AI model. These include a variety of expressions that try to get the best results, using clear and short tips, avoiding unnecessary words, reducing inaccurate descriptions, of course, not all the rules of experience. What we hope is that you can do it step by step, according to the example, rather than drawing up a list that you can remember over and over again. Besides, you don't have to follow all the rules in every tip you start writing. For a simple task, you just need a command + problem to build an effect<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[149],"tags":[177],"collection":[302],"class_list":["post-930","post","type-post","status-publish","format-standard","hentry","category-jiaocheng","tag-chatgpt","collection-prompt"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/930","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=930"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/930\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=930"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=930"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=930"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=930"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}