摘要
Exploring the Potential of Generative AI (ChatGPT) for Foreign Language Instruction: Applications and Challenges Antonio Pérez-Núñez What is ChatGPT? ChatGPT is currently the most powerful generative AI model available for public use. It recently gained significant media attention for its unprecedented capacities to produce human-like text and engage in natural conversations with users (Rudolph, Tan and Tan 2023). ChatGPT is a natural language processing (NLP) system that was released by OpenAI on November 30, 2022, for free research preview. An NLP system is a branch of artificial intelligence (AI) that enables computers to comprehend, manipulate, and generate human language (Chowdhary 2020). NLP systems use a combination of techniques from computer science, linguistics, and machine learning to process and analyze natural language text and speech data (Deng and Lin 2022). ChatGPT employs a deep learning (DL) model constructed using artificial neural networks and designed to analyze input data and extract increasingly complex features (Hu and Buehler 2023). What sets ChatGPT apart from other AI-powered models is its large scale. GPT-4, which is ChatGPT’s current AI model, is reportedly the largest artificial neural network ever produced with over one trillion parameters (Ray 2023). This enormous network was trained using a massive dataset that consisted of web crawl data from the internet, books, conversations, and the English Wikipedia (Brown et al. 2020). This extensive training allows ChatGPT to comprehend and respond to a wide range of topics. ChatGPT has several other features that stand out. Its chatbot interface makes it easy to use, requiring just a prompt to generate a response from the model. It is capable of producing fluent text in multiple languages, including English, Spanish, and Portuguese, across a broad range of genres such as fiction, poetry, news articles, and academic papers (Chatterjee and Dethlefs 2023; Deng and Ling 2022). Finally, ChatGPT is able to learn from previous conversations and feedback, enabling the model to provide more accurate or contextually relevant responses (Tate et al. 2022). Due to its impressive capabilities, an increasing number of studies are exploring the potential use of ChatGPT across multiple fields, including education (Rudolph, Tan, and Tan 2023), medicine (King 2022), journalism and mass media (Pavlik 2023) and computer science (Chatterjee and Dethlefs 2023). The following section outlines and describes specific uses of ChatGPT’s capabilities to support foreign language teaching and learning. Next, it highlights some of the current limitations of the AI system and identifies potential issues that may arise from using generative AI systems in foreign language learning and teaching. Using ChatGPT to Support Language Instruction The potential applications of ChatGPT to support foreign language teaching and learning include: (1) generation of language teaching materials, (2) provision of automatic feedback in writing, and (3) conversation practice. The description of the uses of ChatGPT and the content [End Page 355] provided come from the author’s interaction with the January 9th version of ChatGPT, a free research preview available to the public (https://chat.openai.com/). The examples provided are in Spanish due to the author’s background in teaching the language. Generation of Teaching and Learning Materials Teachers can prompt ChatGPT to automatically generate customized activities and tasks for their classes. To illustrate these capabilities, ChatGPT was prompted to use the content from an article in Spanish published at elpais.com entitled “La suerte de hablar español” to generate tasks targeting the ACTFL interpretive mode. ChatGPT was instructed to generate an interpretive task (figure 1). This was achieved by entering the prompt: “use article ‘La suerte de hablar español’ to create an Interpretive Task for advanced-mid level Spanish students following ACTFL standards and proficiency guidelines. It should include four activities. Activity 1. Main ideas. Activity 2. Supporting details. Activity 3. Meaning from context. Activity 4 Inferences.” Click for larger view View full resolution Figure 1. Interpretive task generated by ChatGPT in English The ChatGPT output displays the AI’s outstanding ability to generate activities using an external input (e.g., newspaper article) and a simple prompt. In this particular case, ChatGPT created a detailed task with a series of activities that require students to engage with the article in a variety of...