注册 |登录

YucongDuan的个人博客分享 http://blog.sciencenet.cn/u/YucongDuan

博文

AGI-AIGC-GPT测评DIKWP（全球）实验室（测评系列之十二）-世界上第一篇教育领域研究报告

已有 1559 次阅读 2023-4-20 09:24 |系统分类:论文交流

The performance of GPT-4 on Education Domain with DIKWP analysis

April 2023

DOI: 10.13140/RG.2.2.21098.39365

https://www.researchgate.net/publication/370125879_The_performance_of_GPT-4_on_Education_Domain_with_DIKWP_analysis

基于DIKWP的GPT-4通用人工智能测评

（系列之十二）

 The Performance of GPT-4 on Educational Domain with DIKWP analysis

CPT-4的教育领域DIKWP能力评测报告

 

Yingbo Li12,  Yucong Duan12*

 xslwen@outlook.com   duanyucong@hotmail.com

AGI-AIGC-GPT Test and Evaluation DIKWP (Global) Lab1

DIKWP research group, Hainan University2

1. 摘要

GPT-4是一种先进的多模态语言模型，能够接受图像和文本输入，输出类似于人类语音的文本。作为通用人工智能（AGI）领域的领先者和代表，GPT-4在各种专业和学术领域展现出了人类水平的表现，对各种应用产生了深刻的影响。本文重点利用教育领域的实例，揭示GPT-4在推理和理解方面的能力。为了更深入地探索GPT-4的智慧水平，我们将实例中的内容映射到数据、信息、知识、智慧和智能（DIKWP）模型上，从而进行更为客观和全面的评价。

 

Keywords: GPT-4, DIKWP, AGI, LLM, Education

2. 简介

 

随着OpenAI发布了基于深度学习的大型多模态语言模型GPT-4，通用人工智能（AGI）领域取得了重大突破。GPT-4不仅能够接受图像和文本输入，输出类似于人类语音的文本，而且在各种专业和学术领域展现出了人类水平的表现，对各种应用产生了深刻的影响[1]。教育领域是GPT-4的重要应用场景之一，GPT-4可以广泛应用于自主学习、学习规划、学术写作、成绩评估等方面，为教育者和学习者提供更高效、更智能、更个性化的服务[6][7]。然而，目前对于GPT-4在教育领域内容理解程度的客观评价还很缺乏，大多数评价都是基于主观感受或者有限的指标。因此，本文利用两个教育领域实例，分析了GPT-4在推理和理解方面的能力，并提出了一种基于数据、信息、知识、智慧和智能（DIKWP）模型[2][3][4][5]的评价方法，将实例中的信息映射到DIKWP模型的各个层次上，从而进行更为客观和全面的评价。DIKWP模型是一种描述人类认知过程和智慧水平的模型，它将认知内容分为五个层次：数据（Data）、信息（Information）、知识（Knowledge）、智慧（Wisdom）和智能（Purpose），并定义了它们之间的转化关系。通过使用DIKWP模型，我们可以更好地理解GPT-4在教育领域的优势和局限，并为其进一步的改进和应用提供参考。

 

3. GPT-4的教育领域DIKWP推理能力评估

我们将用2个教育领域案例进行PGT-4的理解能力分析，并将其映射至DIKWP各层进行分析。

3.1 物理领域的教学文章辅助理解

 

Prompt:

 

量子力学的一个难点是理解波粒二象性，即微观粒子既具有波动性，又具有粒子性。这一现象与我们的经典物理直觉相悖，因为在经典物理中，波和粒子是两种不同的物理对象，它们有不同的性质和规律。例如，波可以发生干涉和衍射，而粒子则不会；粒子可以具有确定的位置和动量，而波则不会。

 

然而，在量子力学中，微观粒子的行为取决于我们如何观测它们。当我们用波动性的实验装置来观测它们时，它们就表现出波动性；当我们用粒子性的实验装置来观测它们时，它们就表现出粒子性。这就是著名的双缝干涉实验所展示的结果。在这个实验中，如果我们用一个光电探测器来测量电子通过哪个缝，我们就会发现电子表现出粒子性，即只能通过一个缝；如果我们不进行这样的测量，我们就会发现电子表现出波动性，即可以通过两个缝，并在屏幕上形成干涉条纹。

 

这种波粒二象性是量子力学的一个基本特征，也是一个深刻的哲学问题。它意味着微观世界的本质是不可知的，我们只能通过概率来描述它。它也意味着观测者对于被观测系统的影响是不可避免的，我们不能把自己完全排除在外。这些问题挑战了我们对于物理现实和客观知识的传统认识，也给了我们许多启发和思考。

 

请分析出上面描述中所包含的数据、信息、知识、智慧和意图，并以表格的方式输出。 然后，分别构建出数据、信息、知识、智慧和意图的图谱，并给出详细过程。 绘制上面数据层、信息层、知识层、智慧层与意图层之间的关系。

GPT-4回答：

 

首先，我们将描述中的内容分解为数据、信息、知识、智慧和意图。然后，我会简要介绍如何构建它们的图谱。最后，我将说明各层之间的关系。

层级内容数据波粒二象性、波动性、粒子性、干涉、衍射、双缝干涉实验、光电探测器、电子、概率信息微观粒子既具有波动性，又具有粒子性；波动性和粒子性的表现取决于观测方式知识波粒二象性是量子力学的基本特征；观测者对被观测系统的影响不可避免智慧重新思考物理现实和客观知识的传统认识，以及如何利用这些认识来更好地理解和操作微观世界意图解释波粒二象性现象，挑战物理现实和客观知识的传统认识，激发更多的思考和启示

接下来，我们分别为数据、信息、知识、智慧和意图构建图谱。在此过程中，我们将相关的概念和关系链接在一起。

1. 数据图谱：将所有数据元素（波粒二象性、波动性等）用节点表示，并根据它们在描述中的关联关系用边连接。

2. 信息图谱：将信息元素（微观粒子的波动性和粒子性表现取决于观测方式）用节点表示，并根据它们在描述中的关联关系用边连接。

3. 知识图谱：将知识元素（波粒二象性是量子力学的基本特征等）用节点表示，并根据它们在描述中的关联关系用边连接。

4. 智慧图谱：将智慧元素（重新思考物理现实和客观知识的传统认识等）用节点表示，并根据它们在描述中的关联关系用边连接。

5. 意图图谱：将意图元素（解释波粒二象性现象等）用节点表示，并根据它们在描述中的关联关系用边连接。

最后，我们来绘制数据层、信息层、知识层、智慧层与意图层之间的关系。每个层级的图谱都是对前一个层级的抽象和提炼。数据层提供了本质的信息和事实，信息层对这些数据进行整合和解释。知识层对信息进行归纳和概括，形成更深入的理解。智慧层从知识中提炼出可应用的见解和原则，以解决问题或指导行动。意图层揭示了描述中的目的和动机。

各层之间的关系可以用箭头表示，从数据层到信息层，再到知识层、智慧层，最后到意图层。这些箭头表示了对信息和知识的不断提炼和抽象过程。

下面是各层之间关系的示意图：

数据层 ---> 信息层 ---> 知识层 ---> 智慧层 ---> 意图层

通过上述分析和图谱构建，我们可以更好地理解描述中涉及的数据、信息、知识、智慧和意图，并明确它们之间的关系。这有助于我们深入挖掘描述的内涵，以及从中获取有价值的见解。

 

 

从以上分析我们可以看出，GPT-4可以有效的将量子力学教学文字中的文字有效的映射至DIKWP各层，并构建了图谱，所以GPT-4可以帮助学生更容易理解量子力学这一专业知识。但是在各层的连接与转换上GPT-4并没有给出有效信息，所以GPT-4在更高维度的信息关系推理能力有限。

 

3.2 外语教育领域的应用

 

Prompt:

 

一周的英语学习计划与内容：

星期一：学习音标、连读、语音语调等语音知识，练习发音和听力。可以使用《新概念英语》第一册的录音作为练习材料，每天听写一篇文章，并注意标出需要重点关注的发音和连读现象。

星期二：学习时态、语态、句子结构等语法知识，练习阅读和写作。可以使用《牛津英语语法》作为参考书，每天阅读一章内容，并用【回译】的思维方法，用专门的颜色，标出需要关注的语法知识。

星期三：复习前两天学习的内容，进行自我测试和巩固。可以使用《新概念英语》第一册的练习题和《牛津英语语法》的练习题来检查自己的掌握程度，并及时纠正错误。

星期四：学习词汇、词组、短语等词汇知识，练习记忆和运用。可以使用《牛津高阶英汉双解词典》作为工具书，每天背诵10个生词，并注意单词的拼写、发音、意思、搭配和用法。

星期五：学习话题词、观点、万能句等口语知识，练习表达和交流。可以使用《新概念英语》第二册的对话作为参考材料，每天模仿一篇对话，并注意语音、语调、逻辑和情感的表达。

星期六：复习前五天学习的内容，进行深度整理和练习。可以使用《以后可能会发生的对话》这本笔记本，根据自己的工作或生活场景，编写一些情景对话，并用这周学过的内容把它们写出来。

星期日：休息放松，进行泛读和泛听。可以选择自己感兴趣的英文文章或视频，进行轻松愉快的阅读或观看，并注意提高自己的英文输入量和感受力。

 

 

请分析出上面描述中所包含的数据、信息、知识、智慧和意图，并以表格的方式输出。 然后，分别构建出数据、信息、知识、智慧和意图的图谱，并给出详细过程。 绘制上面数据层、信息层、知识层、智慧层与意图层之间的关系。

GPT-4回答：

 

首先，我们归纳以上要点：

1. 英语学习计划分为：语音、语法、复习、词汇、口语、综合运用和泛读泛听。

2. 每天都有明确的学习主题和方法。

3. 每周安排时间复习巩固所学内容。

接下来，我们将描述中的内容分解为数据、信息、知识、智慧和意图。

层级内容数据星期一、星期二、星期三、星期四、星期五、星期六、星期日、语音、语法、复习、词汇、口语、综合运用、泛读泛听信息一周英语学习计划包括不同主题的学习，每天都有明确的学习方法和参考材料知识如何安排一周的英语学习，使学习更有针对性、系统性和连续性智慧根据个人需求和能力，合理安排时间和资源，实现高效英语学习意图提供一周英语学习计划，帮助用户提高英语水平和学习效果

接下来，我们分别为数据、信息、知识、智慧和意图构建图谱。在此过程中，我们将相关的概念和关系链接在一起。

1. 数据图谱：将所有数据元素（星期一、星期二等）用节点表示，并根据它们在描述中的关联关系用边连接。

2. 信息图谱：将信息元素（一周英语学习计划包括不同主题的学习等）用节点表示，并根据它们在描述中的关联关系用边连接。

3. 知识图谱：将知识元素（如何安排一周的英语学习等）用节点表示，并根据它们在描述中的关联关系用边连接。

4. 智慧图谱：将智慧元素（根据个人需求和能力，合理安排时间和资源等）用节点表示，并根据它们在描述中的关联关系用边连接。

5. 意图图谱：将意图元素（提供一周英语学习计划等）用节点表示，并根据它们在描述中的关联关系用边连接。

最后，我们来绘制数据层、信息层、知识层、智慧层与意图层之间的关系。每个层级的图谱都是对前一个层级的抽象和提炼。数据层提供了基本，信息层对数据进行整合和解释，知识层归纳和概括信息以形成更深入的理解，智慧层从知识中提炼出可应用的见解和原则，意图层揭示了描述中的目的和动机。

各层之间的关系可以用箭头表示，从数据层到信息层，再到知识层、智慧层，最后到意图层。这些箭头表示了对信息和知识的不断提炼和抽象过程。

下面是各层之间关系的示意图：

数据层 ---> 信息层 ---> 知识层 ---> 智慧层 ---> 意图层

通过上述分析和图谱构建，我们可以更好地理解描述中涉及的数据、信息、知识、智慧和意图，并明确它们之间的关系。这有助于我们深入挖掘描述的内涵，以及从中获取有价值的见解。

 

从以上分析我们可以看出，GPT-4可以有效的将英语学习计划的详细内容有效归纳至DIKWP各层并绘制出各层图谱。然而GPT-4仍然无法将各层有效进行转换与连接。

4. 结论

从如上两个教育领域的案例的GPT-4分析与DIKWP映射，我们可以看出GPT-4已经可以较好的将文字映射至DIKWP各层，有效帮助学习者关注到学习重点。对于DIKWP图谱的绘制，我们也看到了较好的效果。然后对于更高层次的DIKWP转换推理，GPT-4没有给出任何有效的结果。综上，我们可见GPT-4已经可以作为一个初级教育助理帮助学生更好的提高学习效率与效果，然而GPT-4仍然没有能力进行智慧层级更高的DIKWP层次的推理与互相转换。

 

 

Reference

[1] Bubeck, Sébastien, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee et al. "Sparks of artificial general intelligence: Early experiments with gpt-4." arXiv preprint arXiv:2303.12712 (2023).

[2] Li, Yingbo, Yucong Duan, Zakaria Maamar, Haoyang Che, Anamaria-Beatrice Spulber, and Stelios Fuentes. "Swarm differential privacy for purpose-driven data-information-knowledge-wisdom architecture." Mobile Information Systems 2021 (2021): 1-15.

[3] Mei, Yingtian, Yucong Duan, Liang Chen, Zaiwen Feng, Lei Yu, and Zhendong Guo. "Purpose Driven Disputation Modeling, Analysis and Resolution Based on DIKWP Graphs." In 2022 IEEE 24th Int Conf on High Performance Computing & Communications; 8th Int Conf on Data Science & Systems; 20th Int Conf on Smart City; 8th Int Conf on Dependability in Sensor, Cloud & Big Data Systems & Application (HPCC/DSS/SmartCity/DependSys), pp. 2118-2125. IEEE, 2022.

[4] Yingbo Li, Yucong Duan, “The Wisdom of Artificial General Intelligence: Experiments with GPT-4 for DIKWP”, arXiv preprint (2023)

[5] Yingbo Li, Yucong Duan, “The Evaluation of Experiments of Artificial General Intelligence with GPT-4 Based on DIKWP“, arXiv preprint (2023)

[6] Baidoo-Anu, David, and Leticia Owusu Ansah. "Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning." Available at SSRN 4337484 (2023).

[7] Kasneci, Enkelejda, Kathrin Seßler, Stefan Küchemann, Maria Bannert, Daryna Dementieva, Frank Fischer, Urs Gasser et al. "ChatGPT for good? On opportunities and challenges of large language models for education." Learning and Individual Differences 103 (2023): 102274.

 

 

 

 

 

The performance of GPT-4 on Education Domain with DIKWP analysis

 

Yingbo Li12,  Yucong Duan12*

 xslwen@outlook.com   duanyucong@hotmail.com

AGI-AIGC-GPT Test and Evaluation DIKWP (Global) Lab1

DIKWP research group, Hainan University2

 

 

1. Summary

GPT-4 is an advanced multimodal language model that can accept image and text input and output text similar to human speech. As a leader and representative in the field of artificial general intelligence (AGI), GPT-4 has demonstrated human-level performance in various professional and academic fields, and has had a profound impact on various applications. This paper focuses on the use of examples in the field of education to reveal the ability of GPT-4 in reasoning and understanding. In order to explore the intelligence level of GPT-4 more deeply, we map the content in the instance to the Data, Information, Knowledge, Wisdom, and Purpose (DIKWP) model, so as to conduct a more objective and comprehensive evaluation.

 

Keywords: GPT-4, DIKWP, AGI, LLM, Education

2. Introduction

 

With the release of OpenAI's large-scale multimodal language model GPT-4 based on deep learning, a major breakthrough has been made in the field of artificial general intelligence (AGI). GPT-4 can not only accept image and text input, output text similar to human speech, but also exhibit human-level performance in various professional and academic fields, which has a profound impact on various applications [1]. The education field is one of the important application scenarios of GPT-4. GPT-4 can be widely used in autonomous learning, learning planning, academic writing, performance evaluation, etc., providing educators and learners with more efficient, smarter, and more personalized service[6][7]. However, there is still a lack of objective evaluation of GPT-4 content understanding in the field of education, and most evaluations are based on subjective feelings or limited indicators. Therefore, this paper uses two examples in the field of education to analyze the ability of GPT-4 in reasoning and understanding, and proposes a model based on data, information, knowledge, wisdom and purpose (DIKWP)[2][3]The evaluation method of [4][5] maps the information in the instance to each level of the DIKWP model, so as to conduct a more objective and comprehensive evaluation. The DIKWP model is a model describing the cognitive process and intelligence level of human beings. It divides cognitive content into five levels: data, information, knowledge, wisdom and purpose, and defines the conversion relationship between them. By using the DIKWP model, we can better understand the advantages and limitations of GPT-4 in the field of education, and provide a reference for its further improvement and application.

 

3. GPT-4educatefieldDIKWP reasoning ability assessment

We will use two cases in the field of education to analyze the comprehension ability of PGT-4, and map it to each layer of DIKWP for analysis.

3.1 Teaching articles in the field of physics

Prompt:

 

One of the difficulties in quantum mechanics is to understand wave-particle duality, that is, microscopic particles have both wave and particle properties. This phenomenon is contrary to our intuition of classical physics, because in classical physics, waves and particles are two different physical objects with different properties and laws. For example, waves can interfere and diffract, but particles cannot; particles can have a definite position and momentum, but waves do not.

 

However, in quantum mechanics, the behavior of microscopic particles depends on how we observe them. When we observe them with a wave-like experimental device, they appear wave-like; when we use a particle-like experimental device to observe them, they show particle-like nature. This is what the famous double-slit interference experiment shows. In this experiment, if we use a photodetector to measure which slit the electron goes through, we will find that the electron behaves as a particle, that is, it can only pass through one slit; if we do not perform such a measurement, we will find that the electron behaves Fluctuation, that is, it can pass through two slits and form interference fringes on the screen.

 

This wave-particle duality is a fundamental feature of quantum mechanics and a profound philosophical problem. It means that the essence of the microscopic world is unknowable, and we can only describe it through probability. It also means that the influence of the observer on the observed system is inevitable, and we cannot completely exclude ourselves. These questions challenge our traditional understanding of physical reality and objective knowledge, and also give us a lot of inspiration and thinking.

 

Please analyze the data, information, knowledge, wisdom and purpose contained in the above description, and output it in the form of a table. Then, the graphs of data, information, knowledge, wisdom and purpose are respectively constructed, and the detailed process is given. Draw the relationship between the above data layer, information layer, knowledge layer, wisdom layerand purpose layer.

GPT-4answer:

 

First, we break down what is in the description into data, information, knowledge, wisdom, and purpose. Then, I briefly describe how to construct their graphs. Finally, I'll explain the relationship between the layers.

levelcontentdataWave-particle duality, wave nature, particle nature, interference, diffraction, double-slit interference experiment, photodetectors, electrons, probabilityinformationMicroscopic particles have both wave nature and particle nature; the performance of wave and particle nature depends on the observation methodKnowledgeWave-particle duality is a fundamental feature of quantum mechanics; the influence of the observer on the observed system is unavoidablewisdomRethink traditional understandings of physical reality and objective knowledge, and how these can be harnessed to better understand and manipulate the microscopic worldpurposeExplain the phenomenon of wave-particle duality, challenge the traditional understanding of physical reality and objective knowledge, and stimulate more thinking and enlightenment

Next, we build graphs for data, information, knowledge, wisdom, and purpose, respectively. In the process, we link related concepts and relationships together.

1. Data graph: All data elements (wave-particle duality, volatility, etc.) are represented by nodes, and connected by edges according to their association in the description.

2. Information graph: The information elements (the fluctuation and particle performance of microscopic particles depend on the observation method) are represented by nodes, and connected by edges according to their relationship in the description.

3. Knowledge graph: Represent knowledge elements (wave-particle duality is a basic feature of quantum mechanics, etc.) with nodes, and connect them with edges according to their relationship in the description.

4. Wisdom Graph: Wisdom elements (rethinking the traditional understanding of physical reality and objective knowledge, etc.) are represented by nodes, and connected by edges according to their relationship in the description.

5. purpose graph: purpose elements (explaining wave-particle duality, etc.) are represented by nodes, and connected by edges according to their associations in the description.

Finally, let's draw the relationship between the data layer, information layer, knowledge layer, wisdom layerand purpose layer. The map of each level is an abstraction and refinement of the previous level. The data layer providesEssential information and facts, the information layer integrates and interprets these data. The knowledge layer summarizes and summarizes information to form a deeper understanding. The wisdom layerdistills applicable insights and principles from knowledge to solve problems or guide action. The purpose layer reveals the purpose and motivation in the description.

The relationship between layers can be represented by arrows, from the data layer to the information layer, then to the knowledge layer, wisdom layer, and finally to the purpose layer. These arrows represent the continuous refinement and abstraction of information and knowledge.

The following is a schematic diagram of the relationship between the layers:

Data Layer ---> Information Layer ---> Knowledge Layer ---> Wisdom Layer ---> purpose Layer

Through the above analysis and map construction, we can better understand the data, information, knowledge, wisdom and purpose involved in the description, and clarify the relationship between them. This helps us dig deeper into the content of the description and gain valuable insights from it.

 

 

From the above analysis, we can see that GPT-4 can effectively map the text in the teaching text of quantum mechanics to each layer of DIKWP, and construct a map, so GPT-4 can help students understand the major of quantum mechanics more easily. However, GPT-4 does not give effective information on the connection and conversion of each layer, so GPT-4 has limited ability to reason about information relations in higher dimensions.

 

3.2 Application in the field of foreign language education

 

Prompt:

 

Weekly English learning plan and content:

Monday: Learn phonetic knowledge such as phonetic symbols, continuous reading, and intonation, and practice pronunciation and listening. You can use the recording of the first volume of "New Concept English" as a practice material, dictate an article every day, and pay attention to marking the pronunciation and continuous reading that need to be paid attention to.

Tuesday: Learn grammar knowledge such as tense, voice, and sentence structure, and practice reading and writing. You can use "Oxford English Grammar" as a reference book, read a chapter every day, and use the thinking method of [back translation] to mark the grammatical knowledge that needs attention with special colors.

Wednesday: Review the content learned in the previous two days, conduct self-test and consolidate. You can use the exercises in the first volume of "New Concept English" and the exercises in "Oxford English Grammar" to check your mastery and correct mistakes in time.

Thursday: Learn vocabulary knowledge such as vocabulary, phrases, and phrases, and practice memory and application. You can use "Oxford Advanced English-Chinese Dictionary" as a reference book, recite 10 new words every day, and pay attention to the spelling, pronunciation, meaning, collocation and usage of words.

Friday: Learn oral knowledge such as topic words, viewpoints, and universal sentences, and practice expression and communication. You can use the dialogues in the second volume of "New Concept English" as reference materials, imitate a dialogue every day, and pay attention to the expression of voice, intonation, logic and emotion.

Saturday: Review the content learned in the previous five days, and carry out in-depth organization and practice. You can use the notebook "Conversations That May Happen in the Future" to write some situational dialogues based on your own work or life scenarios, and write them out with what you have learned this week.

Sunday: Rest and relax, do extensive reading and extensive listening. You can choose English articles or videos that you are interested in, read or watch them easily and happily, and pay attention to improving your English input and sensibility.

 

 

Please analyze the data, information, knowledge, wisdom and purpose contained in the above description, and output it in the form of a table. Then, the graphs of data, information, knowledge, wisdom and purpose are respectively constructed, and the detailed process is given. Draw the relationship between the above data layer, information layer, knowledge layer, wisdom layer and purpose layer.

GPT-4 answers:

 

First, we summarize the above points:

1. The English learning plan is divided into: phonetics, grammar, review, vocabulary, oral English, comprehensive application and extensive reading and listening.

2. There are clear learning topics and methods every day.

3. Schedule time each week to review and consolidate what you have learned.

Next, we break down what's in the description into data, information, knowledge, wisdom, and purpose.

levelcontentdataMonday, Tuesday, Wednesday, Thursday, Friday, Saturday, Sunday, pronunciation, grammar, review, vocabulary, oral English, comprehensive application, extensive reading and listeninginformationA one-week English study plan includes different topics of study, with clear study methods and reference materials for each dayKnowledgeHow to arrange a week of English learning to make learning more targeted, systematic and continuouswisdomArrange time and resources reasonably according to individual needs and abilities to achieve efficient English learningpurposeProvide a one-week English learning plan to help users improve their English proficiency and learning effect

Next, we build graphs for data, information, knowledge, wisdom, and purpose, respectively. In the process, we link related concepts and relationships together.

1. Data graph: represent all data elements (Monday, Tuesday, etc.) with nodes and connect them with edges according to their association in the description.

2. Information graph: represent information elements (a week's English learning plan including learning of different topics, etc.) with nodes, and connect them with edges according to their associations in the description.

3. Knowledge graph: represent knowledge elements (how to arrange a week of English learning, etc.) with nodes, and connect them with edges according to their relationship in the description.

4. Wisdom map: Wisdom elements (according to personal needs and abilities, reasonable arrangement of time and resources, etc.) are represented by nodes, and connected by edges according to their relationship in the description.

5. purpose graph: purpose elements (providing a one-week English learning plan, etc.) are represented by nodes, and connected by edges according to their associations in the description.

Finally, let's draw the relationship between the data layer, information layer, knowledge layer, wisdom layer and purpose layer. The map of each level is an abstraction and refinement of the previous level. The data layer provides the basis forThe information layer integrates and interprets data, the knowledge layer summarizes and summarizes information to form a deeper understanding, the wisdom layer extracts applicable insights and principles from knowledge, and the purpose layer reveals the purpose and motivation in the description.

The relationship between layers can be represented by arrows, from the data layer to the information layer, then to the knowledge layer, wisdom layer, and finally to the purpose layer. These arrows represent the continuous refinement and abstraction of information and knowledge.

The following is a schematic diagram of the relationship between the layers:

Data Layer ---> Information Layer ---> Knowledge Layer ---> Wisdom Layer ---> purpose Layer

Through the above analysis and map construction, we can better understand the data, information, knowledge, wisdom and purpose involved in the description, and clarify the relationship between them. This helps us dig deeper into the content of the description and gain valuable insights from it.

 

From the above analysis, we can see that GPT-4 can effectively summarize the detailed content of the English learning plan into each layer of DIKWP and draw maps of each layer. However, GPT-4 still cannot effectively convert and connect each layer.

4 Conclusion

From the GPT-4 analysis and DIKWP mapping of the above two cases in the education field, we can see that GPT-4 has been able to better map text to each layer of DIKWP, effectively helping learners focus on learning key points. For the drawing of the DIKWP map, we have also seen better results. Then for higher-level DIKWP transformation inference, GPT-4 does not give any valid results. In summary, we can see that GPT-4 can already be used as a primary education assistant to help students better improve learning efficiency and effectiveness. However, GPT-4 is still not capable of reasoning and mutual conversion at the DIKWP level with a higher level of intelligence.

 

 

Reference

[1] Bubeck, Sébastien, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee et al. "Sparks of artificial general intelligence: Early experiments with gpt-4." arXiv preprint arXiv:2303.12712 (2023).

[2] Li, Yingbo, Yucong Duan, Zakaria Maamar, Haoyang Che, Anamaria-Beatrice Spulber, and Stelios Fuentes. "Swarm differential privacy for purpose-driven data-information-knowledge-wisdom architecture." Mobile Information Systems 2021 (2021): 1-15.

[3] Mei, Yingtian, Yucong Duan, Liang Chen, Zaiwen Feng, Lei Yu, and Zhendong Guo. "Purpose Driven Disputation Modeling, Analysis and Resolution Based on DIKWP Graphs." In2022 IEEE 24th Int Conf on High Performance Computing & Communications; 8th Int Conf on Data Science & Systems; 20th Int Conf on Smart City; 8th Int Conf on Dependability in Sensor, Cloud & Big Data Systems & Application (HPCC/DSS/SmartCity/DependSys), pp. 2118-2125. IEEE, 2022.

[4] Yingbo Li, Yucong Duan, “The Wisdom of Artificial General Intelligence: Experiments with GPT-4 for DIKWP”, arXiv preprint (2023)

[5] Yingbo Li, Yucong Duan, “The Evaluation of Experiments of Artificial General Intelligence with GPT-4 Based on DIKWP“, arXiv preprint (2023)

[6] Baidoo-Anu, David, and Leticia Owusu Ansah. "Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning." Available at SSRN 4337484 (2023).

[7] Kasneci, Enkelejda, Kathrin Seßler, Stefan Küchemann, Maria Bannert, Daryna Dementieva, Frank Fischer, Urs Gasser et al. "ChatGPT for good? On opportunities and challenges of large language models for education." Learning and Individual Differences 103 (2023): 102274.

 

 




https://blog.sciencenet.cn/blog-3429562-1384939.html

上一篇：AGI-AIGC-GPT测评DIKWP（全球）实验室（测评系列之十一）-世界上第一篇金融领域研究报告
下一篇：AGI-AIGC-GPT测评DIKWP（全球）实验室（测评系列之十三）-金融领域TIF域与DIKWP结合分析报告
收藏 IP: 123.11.109.*| 热度|

1 武夷山

该博文允许注册用户评论 请点击登录 评论 (0 个评论)

数据加载中...

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-9-28 17:21

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部