Generative AI Utilization Guide: Empowering Education and Research

Generative AI refers to AI systems that can create new content, such as text, images, audio, music, or video, in response to user prompts. Unlike discriminative AI, which classifies or predicts based on labeled data, generative AI learns patterns and distributions from vast amounts of unlabeled data to produce novel outputs. This guide categorizes generative AI into several types: Text Generation AI (e.g., ChatGPT, Bard), Image Generation AI (e.g., DALL-E, Midjourney), Voice Generation AI (e.g., VoiceBox, Tacotron), Music Generation AI (e.g., MusicLM), and Video Generation AI (e.g., Gen2, Make-A-Video). The core principle is learning from existing data to probabilistically generate new content.

Generative AI exhibits remarkable capabilities, including translation, logical reasoning, problem-solving, summarization, coding, and pattern recognition, largely due to the scale of data they are trained on and advancements like transformer architecture. Multimodal generative AI can process and connect different data types, such as text and images. However, generative AI is not without its limitations. 'Hallucination,' where the AI generates factually incorrect information presented as truth, is a significant challenge stemming from its probabilistic nature. Another critical limitation is the potential for biased content generation, as AI models can reflect biases present in their training data, leading to ethical concerns. Users must exercise critical judgment and verify AI-generated information.

Generative AI can be a valuable asset for educators. Professors can leverage AI for efficient lesson planning, simulating student responses, and designing engaging course content. For course design, asking AI for subject-specific utilization methods, adopting flipped learning strategies, seeking textbook recommendations, and using AI to draft syllabi are effective approaches. In lesson preparation, AI can assist in creating lecture outlines, generating presentation slides, developing diverse explanations, finding illustrative examples, and simulating potential student questions. For assessment, oral exams, AI-generated quizzes, and the use of AI detection tools (like Turnitin and GPTZero) can help maintain academic integrity, though human oversight remains essential. AI can also aid in grading and providing feedback, provided the AI's output is carefully reviewed and edited.

Researchers can integrate generative AI into their workflow to streamline tasks and explore new avenues. Key considerations include adhering to specific guidelines set by academic journals and conferences regarding AI use. If generative AI contributes to research, its use should be properly cited. Researchers must diligently verify the sources and factual accuracy of AI-generated content, as AI can produce misinformation or plagiarize existing work. Awareness of potential AI bias is also critical to prevent skewed research outcomes. Most importantly, researchers must safeguard sensitive or proprietary information by not sharing it with generative AI platforms, which may log or analyze user inputs. Ensuring robust cybersecurity measures is also advised.

Generative AI is rapidly evolving, pushing the boundaries of what artificial intelligence can achieve and bringing us closer to artificial general intelligence. As AI technology continues to advance at an unprecedented pace, it is imperative for educational institutions and researchers to continuously analyze its impact, adapt policies, and develop new guidelines. UNIST's Education Innovation TF remains committed to exploring the advantages of AI while mitigating its risks, ensuring the quality and enhancement of research and education. This guide serves as a foundational resource, with the understanding that ongoing adaptation and critical engagement with AI will be necessary to harness its full potential responsibly.