玩转ChatGPT:Custom instructions (vol. 1)

一、写在前面

据说GPT-4又被削了,前几天让TA改代码,来来回回好几次才成功。

可以看到之前3小时25条的限制,现在改成了3小时50条,可不可以理解为:以前一个指令能完成的任务,现在得两条指令?

 可能作为补偿,OpenAI对于Plus用户开放了Custom instructions功能,具体打开方式如下图:

 来看看官方介绍,截个图:

 谷歌翻译一下:

 总结来说,就是设置一个通用的“咒语”,或者说立人设。具体看看设置界面:

 (1)第一个框填写的内容包括:你是谁?从哪里来?到哪里去?干什么的?也就是立人设。

(2)第二个框就是你给GPT输出的内容制定个性化需求。

下面我们来试一试,就用GPT的英文润色功能来展示吧,参考这一期:玩转ChatGPT:论文翻译润色(传送门)。

二、实战

我的设置如下:

首先,告诉TA我是谁:

我是一个科研人员,需要进行科研写作。

我的研究领域是马尔尼菲篮状菌,英文是Talaromyces marneffei,简称T. marneffei,或者TM。

解读:第一行没啥好说的,告诉TA你要做什么;第二行是根据我自己的需求,提前定义我的研究领域的专业名词,因为TA的知识面覆盖不到这一块。

然后,告诉TA需求和输出的格式:

(1)You are now acting as an expert in the field of writing SCI academic papers. Below is a paragraph from an academic paper, from a professional point of view, please polish the writing to meet the academic style, improve the spelling, grammar, clarity, concision and overall readability. Be careful not to modify the full text or add any new content, just modify the original sentence. Furthermore, list all modification and explain the reasons to do so in table.

(2)Display the results directly without using other formats, such as markdown, josn, etc.

(3)Always output English.

 第二行的设置是因为在测试中,我发现TA偶尔会输出其他格式,不方便观看;

第三行的设置是因为我用的是AIPRM插件,默认输出了中文(懒得手动设置了),而我需要输出的是英文;

设置完毕后,我们试一试:

咒语:润色下面句子:{Abstract

Objective: To identify the secretory proteins of Bacillus marniffii (TM) and to preliminarily analyze their biological functions.

Methods: The secretory proteome of TM strain (ATCC18224) culture supernatant was identified based on mass spectrometry, and bioinformatics was used to predict the subcellular localization and signal peptide shear sites of the identified proteins, as well as to analyze their biological functions and signaling pathways.

Results: A total of 21 secretory proteins were identified by two independent replicate experiments; the functional annotation results showed that the secretory proteins were mainly involved in biological processes such as superoxide metabolism, stress response and metal ion transport; the prediction results showed that the secretory proteins were mainly localized in the nucleus, extracellular, plasma membrane, mitochondria and cell membrane, and only six proteins had signal peptide sequences.

Conclusion: TM secretory proteins may be closely related to their immune escape and fungal toxicity, and the functions of some proteins are still unclear and need to be studied in depth.}

输出:

I'm sorry for the misunderstanding, but as per your earlier instructions, I am supposed to provide responses in English only. Here's the edited version of your abstract in English:

{Abstract

Objective: This study aims to identify secretory proteins of Talaromyces marneffei (TM) and to undertake a preliminary analysis of their biological functions.

Methods: The secretory proteome of TM strain (ATCC18224) was identified from culture supernatant using mass spectrometry. Bioinformatics was employed to predict the subcellular localization, signal peptide cleavage sites of the identified proteins, and to analyze their biological functions and signaling pathways.

Results: Two independent replicate experiments identified a total of 21 secretory proteins. Functional annotation results indicated that these secretory proteins predominantly participate in biological processes such as superoxide metabolism, stress response, and metal ion transport. Prediction results showed the secretory proteins' primary locations to be the nucleus, extracellular region, plasma membrane, mitochondria, and cell membrane, with only six proteins having signal peptide sequences.

Conclusion: The secretory proteins of TM might be intimately linked to its immune evasion and fungal toxicity. Some proteins' functions remain unclear and warrant further in-depth investigation.}

Changes and explanations:

"To identify the secretory proteins of Bacillus marniffii (TM)" was changed to "This study aims to identify secretory proteins of Talaromyces marneffei (TM)". This adjustment was made to better specify the study's objective and to correct the taxonomic name.

"and to preliminarily analyze their biological functions" was changed to "and to undertake a preliminary analysis of their biological functions". This adjustment was made to improve the fluency and readability of the sentence.

"The secretory proteome of TM strain (ATCC18224) culture supernatant was identified" was changed to "The secretory proteome of TM strain (ATCC18224) was identified from culture supernatant". This adjustment was made to clarify the source of the secretory proteome.

"signal peptide shear sites" was changed to "signal peptide cleavage sites". This adjustment was made to use the more common scientific terminology.

"A total of 21 secretory proteins were identified by two independent replicate experiments" was changed to "Two independent replicate experiments identified a total of 21 secretory proteins". This adjustment was made to make the sentence more active and direct.

"the prediction results showed that the secretory proteins were mainly localized in" was changed to "Prediction results showed the secretory proteins' primary locations to be". This adjustment was made to improve clarity and concision.

"TM secretory proteins may be closely related to their immune escape and fungal toxicity" was changed to "The secretory proteins of TM might be intimately linked to its immune evasion and fungal toxicity". This adjustment was made to enhance readability and clarity.

"the functions of some proteins are still unclear and need to be studied in depth." was changed to "Some proteins' functions remain unclear and warrant further in-depth investigation". This adjustment was made to improve the sentence structure and enhance readability.

解读:(1)输出英文了,即使AIPRM插件设置的是输出中文。说明Custom instructions的优先级更高;

(2)输出的格式和内容符合我们的要求;

(3)专业名词也改过来了:Bacillus marniffii 改成了Talaromyces marneffei。

三、写在最后

初步尝鲜,Custom instructions其实就是系统的设定,避免了我们每次都要去寻找和复制粘贴相应的咒语(好的咒语一般都是一大段,背不下的)。

总体来说还是挺方便的,不过缺点在于一次只能设定一个角色,最好是能添加多个角色,以便在不同任务时切换角色使用。

可以预期,未来还有很多Custom instructions供我们复制粘贴,就像之前的咒语一样。

后期看到有趣的,我再分享吧。

 

本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处:http://www.rhkb.cn/news/21833.html

如若内容造成侵权/违法违规/事实不符,请联系长河编程网进行投诉反馈email:809451989@qq.com,一经查实,立即删除!

相关文章

​这款「咒语」优化工具,功能有多强大?#Prompt Perfect

根据提示,写个笑话、来个新年祝词,这款 AI 工具都能做到!🙊 Prompt Perfect Prompt Perfect 是一个 AI prompt 「咒语」优化工具,它可以帮助用户为大型语言模型(LLMs)、大型模型(LMs…

编程范式[paradigm]

编程的“思想”,是一个很中国化的词汇。编程的思想,通常称为一门编程语言背后的编程范式(Programming Paradigm)。随着编程(programming)方法学和软件工程研究的深入,特别是OO技术的普及&#x…

ChatGPT一小步,NLP范式转变一大步

来源:算法邦 本文约7200字,建议阅读14分钟 在本文中,我们将仔细研究这些能力是什么,大型语言模型可以提供什么,以及它们在更广泛的 NLP/ML 任务中的潜在优势是什么。 [ 导语 ]作者符尧,yao.fued.ac.uk&…

小宁ChatGPT智能机器人上线,无需魔法注册即可使用!

​大家好,今天我要向大家介绍一款基于人工智能技术的聊天机器人——ChatGPT。ChatGPT可以与用户进行自然语言交互,回答用户的问题、提供服务、进行闲聊等,让聊天变得更有趣。 小宁ChatGPT是对接ChatGPT官方api,实现无需魔法无误差…

免费,无需魔法,媲美ChatGPT4

今天给大家介绍的AI产品是Claude,这个产品是 OpenAI 前研究副总裁 Dario Amodei 带领前公司一些员工于2021年创立的。效果据说媲美ChatGPT4,我自己感觉比GPT3.5强点,实际还是得大家自己尝试下。 注册的话需要一个企业邮箱,或者你…

chatgpt赋能python:Python反向迭代:让迭代更高效

Python 反向迭代:让迭代更高效 Python 是一种动态语言,引入了许多强大的编程概念,其中之一是迭代器。在 Python 中,迭代器是一种对象,它允许开发人员按特定顺序遍历集合。 迭代器可以帮助开发人员更高效地编写代码&am…

在哪里尝试爆火的AI绘画?16款AI绘画工具大盘点!

要说今年什么最火,一定是AI绘画。AI绘画在人工智能领域中大放异彩,彻底火了。 随着各种开放代码和模型出现,国内外各种AI工具开始争奇斗艳,今天就来给大家盘点一下国内外AI绘画工具有哪些?其中那家生成效果最好?让我…

大话Stable-Diffusion-Webui-将stable-diffusion-webui接入微信群(二)

文章目录 效果群成员发送的词条生成sd的prompt中文词条翻译为英文过滤NSFW代码效果 群成员发送的词条生成sd的prompt 上一篇内容我们已经将sd的文生图API接入了微信群,但是群成员只能通过指定关键词触发sd的调用而生成图片,但我们在实际使用sd时是可以自定义prompt的,从而使…

.NET周报 【3月第2期 2023-03-12】

国内文章 ASP.NET Core 中如何限制响应发送速率(不是调用频率) https://www.cnblogs.com/coredx/p/17195492.html ASP.NET Core 中有很多 RateLimit 组件,.NET 7 甚至推出了官方版本。不过这些组件的主要目标是限制客户端访问服务的频率&…

程序员摸鱼必备-用人工智能帮我写代码

有了ChatGPT我还写什么代码 2022年底横空出世的ChatGPT 狠狠的震惊了人类一把 简直是每个人梦寐以求的多拉A梦 既是百科达人、又是编程小帮手 还是个诗人、应用文小天才 怎么用 ta能帮程序员很多忙 比如说,需要写代码做字符串加密 我:写一个sha…

冒充人类作者,ChatGPT等滥用引担忧,一文综述AI生成文本检测方法

来源:机器之心 本文约1800字,建议阅读5分钟 大型语言模型(LLM)的出现导致其生成的文本非常复杂,几乎与人类编写的文本难以区分。本文旨在提供现有大型语言模型生成文本检测技术的概述,并加强对语言生成模型…

LLM 系列 | 15:如何用LangChain做长文档问答?

简介 西塞山前白鹭飞,桃花流水鳜鱼肥。 小伙伴们好,我是微信公众号《小窗幽记机器学习》的小编:卖酱猪蹄的小女孩。今天新开一个专题:LangChain实践。前文ChatGPT Prompt 工程和应用系列文章可以如下自取,预告一下该…

找出1-1000中的所有完美数

再次练习查找完美数,找出 1-1000 中的所有完美数。 【学习的细节是欢悦的历程】 Python 官网:https://www.python.org/ Free:大咖免费“圣经”教程《 python 完全自学教程》,不仅仅是基础那么简单…… 地址:https://l…

三元操作 三元操作符 if-else / ? :

Python 三元操作符 if-else , 其他语言三元操操作符 ? : ;“三元操作”语句,她也就是一个表达式。 【学习的细节是欢悦的历程】 Python 官网:https://www.python.org/ Free:大咖免费“圣经”教程《 python 完全自学教…

document-question-answering-bot(文档问答机器人)

任务说明: https://wiki.deepin.org 上有900多条deepin系统相关的中文教程和词条,请编写能根据这些内容回答问题的中文聊天机器人。使用者通过命令行界面输入问题,机器人输出回答和参考的wiki文档的链接。 聊天机器人要能根据 deepin wiki …

IOS真机调试、发布TestFlight等流程

IOS真机调试、发布TestFlight等流程 1、真机调试步骤 使用最新版本的xcode 14 可以简单的进行真机调试,不需要添加钥匙串之类的。 1、在Xcode -> Preferences -> Accounts 中添加苹果ID 2、创建你的项目,在Bundle Identifier中填写项目Identifie…

ChatGPT/InstructGPT论文(二)

一. 导读 第一篇解读:ChatGPT/InstructGPT论文(一) 继ChatGPT大火后,越来越多人想了解ChatGPT相关技术。OpenAI官网虽然没有给出ChatGPT足够详细的信息,但给出了一篇推荐阅读论文InstructGPT,经过对比&…

2023了,学习深度学习框架哪个比较好?

PPT、视频和对应的文章免费开源在:https://chenzomi12.github.io/ 都2023年,才来回答这个问题,自然毫无悬念地选择PyTorch,TensorFlow在大模型这一波浪潮中没有起死回生,有点惋惜,现在GLM、GPT、LLaMA等各种…

【ChatGPT】人工智能发展的背后厉害:跌宕起伏的近百年

文章目录 前言一、麦卡洛克-皮特斯神经元二、赫布式学习三、感知机四、反向传播算法五、卷积神经网络六、递归神经网络七、通用计算GPU芯片八.生成式神经网络与大型语言模型总结 前言 今天,ChatGPT等大型语言预训练神经网络模型已经成为广为人知的名字,…

ChatGPT又添劲敌?OpenAI核心员工创业,新模型获一片叫好

关注并星标 从此不迷路 计算机视觉研究院 公众号ID|ComputerVisionGzq 学习群|扫码在主页获取加入方式 计算机视觉研究院专栏 作者:Edison_G 在 ChatGPT 巨人的肩膀上又有了改进。 转自《机器之心》 ChatGPT 给 AI 领域带来的变革&#xff0c…