A brave new world of artificial intelligence (AI) opened in the classroom this year so we sat down with ChatGPT-4 and two Franklin & Marshall College professors to discuss the future.
What do they think that brave new world will look like? Nicholas Montemarano, Alumni Professor of Creative Writing and Belle Lettres, and Associate Professor of Computer Science Ed Novak offered some sobering and hopeful insights.
Montemarano, who teaches creative writing, also invited ChatGPT-4 to participate in the discussion. The professors and the machine responded to a series of questions asked by the human writing this. Their responses have been edited for brevity.
Q: Can AI write better than humans?
Montemarano: It depends on what we mean by "better." My hypothesis is that AI writes "better" than some humans, but far worse than the best human writers. To be clear, I'm thinking in terms of creative writing. When I think of George Saunders, a contemporary fiction writer who stands out to me for his unique style and voice, there's no way AI could produce writing so idiosyncratic. I suppose you could feed AI everything Saunders has ever written and ask it to write a story in the style of Saunders, but my guess is that it would be a very watered down and laughable attempt to write like Saunders. Let's say George Saunders never existed. Do I believe that AI ever could have produced something that sounds like Saunders' writing? Not a chance! Only a human being could. I could say the same for many writers—Lydia Davis, Raymond Carver, Virginia Woolf, etc.
Novak: “Better" is very subjective in most writing. In expository writing, I think that current AI can write as well as a well-motivated, but perhaps haphazard undergraduate student. In terms of persuasive, descriptive, and narrative writing, the AI does not do as well. It is comparable to a lower-class-year high school student. Current AI writes much faster, and can be coerced via elaborate prompt-writing to perform better or worse according to the user's desired criteria.
ChatGPT-4: It depends on how you define "better.” AI can certainly generate a large amount of text quickly, and in some cases, it can produce text that is grammatically correct and coherent. In fact, AI language models like GPT-3 have shown impressive capabilities in generating human-like text. However, when it comes to producing text that is creative, engaging, and emotionally resonant, humans still have the upper hand. AI lacks the ability to understand the nuances of human emotions, experiences, and cultural contexts, which are essential for producing truly compelling writing.
Q: As a professor of English and a published author, how do you imagine AI being used in your teaching two years from now?
Montemarano: It's hard to say, but my best guess is that I won't use it at all in my creative writing courses. I'm very happy teaching human writers. I have no desire to involve AI writing—not even to discuss the ways it may lack certain qualities we see in the best human writing. Maybe I should be more curious, but the truth is, I'm not.
Q: How do you imagine AI being used to teach English two years from now?
ChatGPT-4: In the next two years, AI could be used in several ways to enhance the teaching of English: AI can be used to automate the assessment of written and spoken English, providing instant feedback on grammar, vocabulary, pronunciation, and other language skills. This could help teachers save time and provide more individualized feedback to students. Create personalized learning plans for students based on their individual needs, interests, and learning styles. Provide students with real-time support and guidance as they practice speaking and listening in English. Develop language models that analyze large amounts of text and provide insights into grammar, vocabulary, and usage patterns. These models can be used to create more effective language learning materials, such as textbooks, exercises, and instructional videos. While AI cannot replace human teachers, it can complement their efforts and provide students with more opportunities to practice and improve their language skills.
Q: As a professor of computer science, how do you imagine AI being used in your teaching two years from now?
Novak: I have been thinking a lot about both how my teaching and how the computer science industry in general are likely to change based on these new AI agents. My inclination is to accept the presence of AI and to teach students to effectively utilize it. AI is particularly threatening to programmers. But, as more time passes, the more it is clear to me that current AI is not sophisticated enough to replace programmers.
Q: How do you imagine AI being used to teach computer science two years from now?
ChatGPT-4: In the next two years, AI could be used in several ways to enhance the teaching of computer science: Automated Grading. Create personalized learning paths for students based on their individual strengths, weaknesses, and learning styles. Virtual teaching assistants can be used to provide students with real-time support and guidance as they work through coding exercises and assignments. Tutoring systems can be used to provide students with feedback and guidance in real-time as they work through coding exercises and assignments. These systems can analyze a student's code and provide suggestions for improvement, as well as provide explanations for concepts they may be struggling with. Overall, AI has the potential to revolutionize the teaching of computer science by providing more personalized, efficient, and effective learning experiences for students
That Human Touch: A Chat with AI
Q: Will AI ever become creative, empathetic and intuitive?
ChatGPT-4: AI has already shown some capacity to emulate aspects of creativity, empathy, and intuition, but whether it will ever truly possess these traits in the way humans do is still an open question. AI has shown some ability to generate creative works such as music, art, and literature, although these are often based on patterns learned from existing examples rather than truly original creations. Similarly, AI systems can be programmed to respond to human emotions and display empathy, but it is not clear whether they truly understand or experience emotions in the same way humans do. As for intuition, AI systems have shown some capacity to make decisions based on complex patterns and relationships, but this is still based on logical algorithms rather than the kind of intuitive decision-making that humans often rely on. It is important to recognize that human intelligence and AI intelligence are fundamentally different, and it is unclear whether AI will ever fully replicate the complexities of human creativity, empathy, and intuition. While AI systems can assist humans in various ways, it is unlikely that they will completely replace human intelligence in these areas. Nonetheless, continued research in AI could lead to further advances in these areas and new insights into the nature of human intelligence.
Q: You're not just placating me, are you?
ChatGPT-4: As an AI language model, I don't have feelings or intentions, and I don't placate or deceive anyone. I provide responses based on the algorithms and data used to train me, and I strive to provide accurate and informative answers to the best of my abilities. If you have any concerns about my responses or want more information, please feel free to ask me further questions.
Q: If AI does not have a mother or a father, how could it become empathetic, intuitive or creative?
ChatGPT-4: AI does not have a biological mother or father, but it can be designed and programmed by human beings to possess qualities such as empathy, intuition, and creativity. These qualities are not exclusive to humans and can be programmed into an AI system through various techniques such as machine learning, neural networks, and natural language processing. For example, an AI system can be trained to recognize patterns in human behavior and emotions, allowing it to respond empathetically to human needs and emotions. Overall, while AI may not have a biological mother or father, it can still possess qualities that we typically associate with human beings, including empathy, intuition, and creativity, through careful programming and design.
Professor Montemarano responds:
I found several things about the AI responses to be noteworthy:
First, its writing is very competent, clean, and grammatically correct. It sounds like it could have been written by a human, albeit a human whose individual personality isn't coming through in the writing. It's very clinical, and of course that makes sense! Note that I chose to end my previous sentence with an exclamation point. Even in my responses to your initial questions, when I was writing about George Saunders and his singular writing style, I used an exclamation point: "Not a chance!" That also happens to be a sentence fragment. Used for stylistic purposes. I don't believe that the AI responses to your questions include sentence fragments—because they're not "correct." So its writing is competent, but has little personality. That could be fine when used for certain kinds of writing, but not for creative writing.
My second observation is how honest the AI response is when articulating its limitations compared with human writers. There are so many "rogue AI" narratives out there in books and films, and have been for years, that I might have come to expect the AI response to be defensive or self-serving. I was struck by how it listed human advantages such as life experiences, emotions, unique perspectives, etc.
Finally, one phrase among the AI responses stood out to me—to the writer in me: "AI does not have a biological mother or father..." It may seem strange to say this, but that phrase "moved" me. In other words, I felt empathy for a nonhuman that itself has no real empathy. That said, you were the one who introduced the phrase in your question, so AI did not come up with that on its own.