Increase Your Creativity With ChatGPT – Psychology Today
Hope is double-edged; false hope can set you on a collision course with despair. Know when your hopes are well-founded and how to turn your deep desires into results. 
Verified by Psychology Today
Posted June 29, 2023 Reviewed by Devon Frye
Brainstorming usually means getting a group of people together to generate new ideas. I have found a novel way of brainstorming by getting the artificial intelligence program ChatGPT to generate interesting questions and answers, leading to more questions and answers.
ChatGPT is serving as a useful research assistant for the book I am now writing on consciousness, despite its tendency to make serious mistakes, such as fake references. But compared to Google searches, ChatGPT’s answers to my queries tend to be more organized and articulate.
I recently noticed that ChatGPT is more than a question answerer, because it can generate original questions, concepts, hypotheses, and even methods. Creativity is the ability to produce ideas that are novel, surprising, and valuable. In my 2021 book Bots and Beasts, I reviewed current AI programs and concluded that most are not creative at all, although a few (e.g. IBM Watson and AlphaFold) have some creativity.
One day in May, I muttered to ChatGPT: “You’re so smart at answering questions based on the billions of web documents you’ve been trained on, but the real creativity comes from generating smart questions.” But then I wondered if ChatGPT could generate questions, and immediately discovered that ChatGPT can in fact do this, which suggests a powerful new way of brainstorming.
I am now working on a chapter on musical consciousness and ChatGPT has generated great questions in response to queries like “What are good questions about the role of consciousness in music?” Typically, ChatGPT produces 10-20 new questions including ones that probably would not have occurred to me even with a lot of research. I have found ChatGPT to be terrific at generating questions in many other domains as well, from philosophy and cognitive science to gardening.
Google searches suggest that the lists are specially generated, not just copied from other lists, in keeping with the way ChatGPT works with an artificial neural network based on more than a billion parameters (weighted connections between neurons). Hence the questions generated by ChatGPT are novel, many of them are surprising, and some of them are highly valuable in suggesting new directions of inquiry. Therefore, ChatGPT’s question generation seems to me to qualify as creative.
After ChatGPT generates a list of questions, you can get it to answer them one at a time. This process of getting ChatGPT to provide questions and answers could go on indefinitely, although some repetition will turn up. On musical consciousness, ChatGPT gives me new questions and then answers to these questions, and I then turn the answers into more questions. The questions are always interesting, but I don’t trust the answers until I can check them out in reputable sources because of ChatGPT’s notorious tendency to make stuff up.
I call this new method “ChatStorming,” but it raises some difficult questions.
The good news is that these answers are not subject to copyright which can only be held by persons: unlike many of the websites that ChatGPT was trained on, it has no copyright to its material. ChatGPT produces responses based on learning from training data and does not store the work of individual authors. However, it would still be intellectually dishonest to pass off writing produced by ChatGPT as your own.
An old jokes say that if you steal from one person, it’s called plagiarism; but if you steal from many people, it’s called research. The punch line isn’t true, because good research always indicates the sources it draws on for ideas and evidence, including AI models.
ChatGPT is great at producing creative questions, but how about concepts, hypotheses, and methods? To my astonishment, ChatGPT fluently answers prompts such as these:
The results produced are definitely novel and surprising, but I don’t yet know whether they can also be valuable. Most new ideas turn out to be useless, and it will take much testing and evaluation to determine whether ChatGPT and even better models like GPT-4 can be creative with new ideas as well as questions. Another open question is what it would take for a large language model to dominate humans. Large language models are major advances in AI that can enhance human creativity, but they are not to be trusted.
Paul Thagard, Ph.D., is a Canadian philosopher and cognitive scientist. His latest book, published by Columbia University Press, is Balance: How it Works and What it Means.
Get the help you need from a therapist near you–a FREE service from Psychology Today.
Psychology Today © 2023 Sussex Publishers, LLC
Hope is double-edged; false hope can set you on a collision course with despair. Know when your hopes are well-founded and how to turn your deep desires into results.
