SlideShare a Scribd company logo

Chat GPT and Generative AI in Higher Education - Empowering Educators and Learners

•Download as PPTX, PDF•
•
Alain Goudey
Alain Goudey

If you failed to join us for this inspiring and groundbreaking conference that explores the transformative potential of ChatGPT and generative AI in higher education at AACSB Innovative Curriculum Conference in 2023. This slidedeck brings together some ideas in education, technology, and artificial intelligence to delve into the exciting possibilities that these innovative technologies hold for educators and learners alike. Discover how ChatGPT and generative AI are revolutionizing teaching methods, enhancing student engagement, and promoting personalized learning experiences. Gain insights into the latest developments in AI-powered educational tools and platforms, and learn how they can help students overcome academic challenges, foster critical thinking, and unlock their full potential. At NEOMA we are at the forefront of integrating AI into the classroom, and explore successful case studies that showcase the immense benefits of this digital transformation. We also address the ethical considerations, best practices, and strategies for harnessing the power of ChatGPT and generative AI to create more equitable and inclusive educational environments. Let's embark together on a thrilling journey that will redefine the way we teach, learn, and grow with AI, connect on social networks with me.

1 of 22
Alain Goudey 03/22/2023 NEOMA Business School Twitter: @AlainGoudey https://alain.goudey.eu Introduction to ChatGPT… and generative AIs
Generative AI The set of algorithms that can take in a prompt (text, audio, images, random noise, or nothing) and output something that looks realistic and reasonable to humans. (Zhou, 2022)
Generative AI Timeline OpenAI released an updated version of ChatGPT using an advanced model based on human feedback 2014 2017 OpenAI’s GPT-1 can generate coherent text (groundwork for the creation of more advanced systems) 2021 2022 Emergence of the Generative Adversarial Networks (GAN) OpenAI launches DALL.E (AI for image generation) 2019 Google starts using BERT’s algorithm for understanding human language
Photo by Pexels News about Generative AIs A huge impact in business
3 major consequences of GenAIs: impacts on job market
3 major consequences of GenAIs: change our link to information and knowledge
From a sketch to a website in 30 sec. 3 major consequences of GenAIs: faster prototyping
Photo by Pexels What is ChatGPT? - Generative AI tool for text with more than a million users in 5 days. 100 millions today (since nov’22). 80% of our students know about it, 30% are using it regularly. - Used for text writing, idea exploration, code prototyping, thought structuring, etc. - ChatGPT (Generative Pre-trained Transformer) is a neural network architecture that uses automatic language processing techniques to understand and generate content based on GPT-3 - Based on the Transformer approach invented by Google in 2017 - Other tools based on the same approach exist, such as Google's LaMDA, GPT-4 (since 14th March 2023) or Bloom (Europe)
ChatGPT It can record and learn from previous conversations Follow-up corrections Based on Reinforcement Learning with Human Feedback Massive databases (knowledge) and 25 capabilities on Natural Language Processing (NPL) Chatbot Human-like conversations Massive data
How does it work? The user provides a ”prompt” (text, image, audio, or a mix) Neural network processing • - Machine learning algorithms • - Datasets for language processing • - Databases and documents • - Human feedback 01 02 03 Output The algorithm generates an example It doesn’t understand for real!
GPT-4 is more precise and powerful How has ChatGPT been trained?
GPT-4 passes in 67% to 90% of cases What about GPT-4 and exams?
Photo by Pexels Generating Content With ChatGPT / generative AIs - Uses information "learned" during training to identify context and generate output - Uses text completion techniques to generate responses using a given context - The "art of prompt engineering" is essential in any generative AI as it gives the priming for the results It is a statistical construct of the content!
Photo by Pexels Generative AIs can generate a lot of various content - texts, images, videos, slides, voices, codes, excel formulas, even objects (!) - 2023 is about to be the Year of Generative AIs Photos by https://iAMoi.art
Uses of ChatGPT in general - Used to generate content that is logical and fluent - Can generate sentences or responses using a context, a prompt: answers to questions, summaries of scientific articles, paragraphs of content for course materials, stories, articles, etc. - ChatGPT can be used to create interactive dialogues, tutorials and self-assessment exercises for students - ChatGPT can automatically correct spelling and grammar mistakes in students’ work It solves the blank page syndrome
Limitations of ChatGPT - High dependence from the quality of the prompt - When a data doesn’t exist, it uses an ungraceful degradation to « replace it »… - … leading to “hallucinations” and fake news - Answers highly depend on the original dataset used for training (GPT3, 3.5 and are highly US based, non transparent and stopped in nov’21) ChatGPT can « hallucinate » and provide fake news with a huge assertivity
- Source of: information, debates, examples, grading for students - Start for thoughts - Structure your thinking / document - Summarize - Personalize the content - Ask for coaching Use it as another « expert » to ask, as another source of information… Uses of ChatGPT in the classroom
To work around ChatGPT's limitations, changing the form of the assignment or exam is important: • Some possibilities include having students write by hand, using MCQs or high/complex logic, using graphics, mind maps, images or video to convey concepts • Video is an effective means of expression and can help students learn how to express themselves intelligently • Specific questions, logic or original analysis will render the tool useless • To work around this, apply concepts to a specific case, use less mainstream or recent topics, and focus on critical analysis rather than memorization • Make your students think about why they made their dissertations evolve, vary in their angles / ways of envisioning the topic • Develop strategies for monitoring, curation and critical thinking, which will become essential in an era of continuous and automated content • Place students as reviewers of AI answers If many tools are announced, evaluating the use of GPT after the exam is not reliable If Generative AIs passe, change the exams
1. Purpose: Clearly describe the purpose or end goal of the content you want to generate. For example, do you want a blog post, a story, a poem, etc.? 2. Tone: specify the tone you want for the generated content. For example, do you want a formal, informal, serious, humorous, etc. tone? 3. Target Audience: Describe the target audience for whom you want the content generated. This will help ChatGPT tailor the style and tone of the writing to that target audience. 4. Additional information: Provide any additional information that can help ChatGPT understand the context, such as topics or themes, character information, events, dates, numbers, etc. For example, you can “put ChatGPT in the shoes of a character”, such as a salesperson or an advertising copywriter, etc. 5. Constraints: specify any constraints or restrictions that may apply to the generated content, such as length, keywords to include or exclude, etc. 6. Discuss: Construct it as a discussion to refine the answers provided The more information you get, the better is the answer For new skills: the art of prompt engineering
A conclusion https://www.gartner.com/en/arti cles/what-s-new-in-artificial- intelligence-from-the-2022- gartner-hype-cycle ChatGPT For me: Painting is not (only) reproducing, writing is not (only) describing, creating is not (only) duplicating, coding is not (only) writing codes. Thinking is not (only) producing outputs. « Thinking is the greatness of humanity » (Blaise Pascal)
Resources for more - https://medium.com/@AlainGoudey (articles in English) - https://www.linkedin.com/in/goudey/ (articles in French + regular curation) - https://www.linkedin.com/groups/12775320/ (185+ members for a collective curation) - https://er.educause.edu/articles/2023/2/educause-quickpoll- results-did-chatgpt-write-this-report - https://platform.openai.com/docs/chatgpt-education - https://www.timeshighereducation.com/campus/collections/ai- transformers-chatgpt-are-here-so-what-next - https://hbsp.harvard.edu/inspiring-minds/chatgpt-and-ai-text- generators-should-academia-adapt-or-resist - https://ed.stanford.edu/news/stanford-faculty-weigh-new-ai- chatbot-s-shake-learning-and-teaching - 15 AI tools for the classroom - Ditch That Textbook - Lecturers urged to review assessments in UK amid concerns over new AI tool | Artificial intelligence (AI) | The Guardian - Intelligence Artificielle et impact en enseignement supérieur (padlet.org) - Réflexions sur les AI génératives et l'enseignement supérieur (padlet.com) - (PDF) Chatbot Prompting: A guide for students, educators, and an AI- augmented workforce (researchgate.net)
Alain.Goudey@neoma-bs.fr Twitter: @AlainGoudey https://alain.goudey.eu Feel free to contact

More Related Content

Chat GPT and Generative AI in Higher Education - Empowering Educators and Learners

  • 1. Alain Goudey 03/22/2023 NEOMA Business School Twitter: @AlainGoudey https://alain.goudey.eu Introduction to ChatGPT… and generative AIs
  • 2. Generative AI The set of algorithms that can take in a prompt (text, audio, images, random noise, or nothing) and output something that looks realistic and reasonable to humans. (Zhou, 2022)
  • 3. Generative AI Timeline OpenAI released an updated version of ChatGPT using an advanced model based on human feedback 2014 2017 OpenAI’s GPT-1 can generate coherent text (groundwork for the creation of more advanced systems) 2021 2022 Emergence of the Generative Adversarial Networks (GAN) OpenAI launches DALL.E (AI for image generation) 2019 Google starts using BERT’s algorithm for understanding human language
  • 4. Photo by Pexels News about Generative AIs A huge impact in business
  • 5. 3 major consequences of GenAIs: impacts on job market
  • 6. 3 major consequences of GenAIs: change our link to information and knowledge
  • 7. From a sketch to a website in 30 sec. 3 major consequences of GenAIs: faster prototyping
  • 8. Photo by Pexels What is ChatGPT? - Generative AI tool for text with more than a million users in 5 days. 100 millions today (since nov’22). 80% of our students know about it, 30% are using it regularly. - Used for text writing, idea exploration, code prototyping, thought structuring, etc. - ChatGPT (Generative Pre-trained Transformer) is a neural network architecture that uses automatic language processing techniques to understand and generate content based on GPT-3 - Based on the Transformer approach invented by Google in 2017 - Other tools based on the same approach exist, such as Google's LaMDA, GPT-4 (since 14th March 2023) or Bloom (Europe)
  • 9. ChatGPT It can record and learn from previous conversations Follow-up corrections Based on Reinforcement Learning with Human Feedback Massive databases (knowledge) and 25 capabilities on Natural Language Processing (NPL) Chatbot Human-like conversations Massive data
  • 10. How does it work? The user provides a ”prompt” (text, image, audio, or a mix) Neural network processing • - Machine learning algorithms • - Datasets for language processing • - Databases and documents • - Human feedback 01 02 03 Output The algorithm generates an example It doesn’t understand for real!
  • 11. GPT-4 is more precise and powerful How has ChatGPT been trained?
  • 12. GPT-4 passes in 67% to 90% of cases What about GPT-4 and exams?
  • 13. Photo by Pexels Generating Content With ChatGPT / generative AIs - Uses information "learned" during training to identify context and generate output - Uses text completion techniques to generate responses using a given context - The "art of prompt engineering" is essential in any generative AI as it gives the priming for the results It is a statistical construct of the content!
  • 14. Photo by Pexels Generative AIs can generate a lot of various content - texts, images, videos, slides, voices, codes, excel formulas, even objects (!) - 2023 is about to be the Year of Generative AIs Photos by https://iAMoi.art
  • 15. Uses of ChatGPT in general - Used to generate content that is logical and fluent - Can generate sentences or responses using a context, a prompt: answers to questions, summaries of scientific articles, paragraphs of content for course materials, stories, articles, etc. - ChatGPT can be used to create interactive dialogues, tutorials and self-assessment exercises for students - ChatGPT can automatically correct spelling and grammar mistakes in students’ work It solves the blank page syndrome
  • 16. Limitations of ChatGPT - High dependence from the quality of the prompt - When a data doesn’t exist, it uses an ungraceful degradation to « replace it »… - … leading to “hallucinations” and fake news - Answers highly depend on the original dataset used for training (GPT3, 3.5 and are highly US based, non transparent and stopped in nov’21) ChatGPT can « hallucinate » and provide fake news with a huge assertivity
  • 17. - Source of: information, debates, examples, grading for students - Start for thoughts - Structure your thinking / document - Summarize - Personalize the content - Ask for coaching Use it as another « expert » to ask, as another source of information… Uses of ChatGPT in the classroom
  • 18. To work around ChatGPT's limitations, changing the form of the assignment or exam is important: • Some possibilities include having students write by hand, using MCQs or high/complex logic, using graphics, mind maps, images or video to convey concepts • Video is an effective means of expression and can help students learn how to express themselves intelligently • Specific questions, logic or original analysis will render the tool useless • To work around this, apply concepts to a specific case, use less mainstream or recent topics, and focus on critical analysis rather than memorization • Make your students think about why they made their dissertations evolve, vary in their angles / ways of envisioning the topic • Develop strategies for monitoring, curation and critical thinking, which will become essential in an era of continuous and automated content • Place students as reviewers of AI answers If many tools are announced, evaluating the use of GPT after the exam is not reliable If Generative AIs passe, change the exams
  • 19. 1. Purpose: Clearly describe the purpose or end goal of the content you want to generate. For example, do you want a blog post, a story, a poem, etc.? 2. Tone: specify the tone you want for the generated content. For example, do you want a formal, informal, serious, humorous, etc. tone? 3. Target Audience: Describe the target audience for whom you want the content generated. This will help ChatGPT tailor the style and tone of the writing to that target audience. 4. Additional information: Provide any additional information that can help ChatGPT understand the context, such as topics or themes, character information, events, dates, numbers, etc. For example, you can “put ChatGPT in the shoes of a character”, such as a salesperson or an advertising copywriter, etc. 5. Constraints: specify any constraints or restrictions that may apply to the generated content, such as length, keywords to include or exclude, etc. 6. Discuss: Construct it as a discussion to refine the answers provided The more information you get, the better is the answer For new skills: the art of prompt engineering
  • 20. A conclusion https://www.gartner.com/en/arti cles/what-s-new-in-artificial- intelligence-from-the-2022- gartner-hype-cycle ChatGPT For me: Painting is not (only) reproducing, writing is not (only) describing, creating is not (only) duplicating, coding is not (only) writing codes. Thinking is not (only) producing outputs. « Thinking is the greatness of humanity » (Blaise Pascal)
  • 21. Resources for more - https://medium.com/@AlainGoudey (articles in English) - https://www.linkedin.com/in/goudey/ (articles in French + regular curation) - https://www.linkedin.com/groups/12775320/ (185+ members for a collective curation) - https://er.educause.edu/articles/2023/2/educause-quickpoll- results-did-chatgpt-write-this-report - https://platform.openai.com/docs/chatgpt-education - https://www.timeshighereducation.com/campus/collections/ai- transformers-chatgpt-are-here-so-what-next - https://hbsp.harvard.edu/inspiring-minds/chatgpt-and-ai-text- generators-should-academia-adapt-or-resist - https://ed.stanford.edu/news/stanford-faculty-weigh-new-ai- chatbot-s-shake-learning-and-teaching - 15 AI tools for the classroom - Ditch That Textbook - Lecturers urged to review assessments in UK amid concerns over new AI tool | Artificial intelligence (AI) | The Guardian - Intelligence Artificielle et impact en enseignement supérieur (padlet.org) - Réflexions sur les AI génératives et l'enseignement supérieur (padlet.com) - (PDF) Chatbot Prompting: A guide for students, educators, and an AI- augmented workforce (researchgate.net)