My principles for building and using a custom ChatGPT as a quantum AI researcher
This is the post where I shared how to set up your ChatGPT with customized prompts and data available to accelerate your research projects.
Have you ever dreamed of someday you could have an AI chatbot to help you with your literature review process? Have you ever wanted a research partner to do literature with you? With customized Chat-GPT, you can build your own version of a personalized peer researcher with background knowledge from your domain. Today, I want to share how I use a customized chatbot using ChatGPT for my quantum machine learning research. Without further ado, let's get started.
Define your data policy before you start to do anything
Defining your data policy is about what to collect and what to put on the OpenAI's platform. Especially since this technology is still in the very early stage of development, there are significant risks and problems regarding the potential data leakage issues. Therefore, a great data policy and guidelines are two of the most important things to design before configuring the app for the problems you are addressing.
My policy is to set a boundary for my data usage; having public and private domains is essential to keep your most important thing to let AI understand and what is the most personal thing you need to keep from AI systems and potentially other users to know about. For example, I will put the editorial policy and submission guidelines, which are publically available through the Internet, on my chatbot to help edit the manuscript. However, I will avoid uploading my unpublished manuscript to the publically available chatbot on ChatGPT, which might expose me to the dangers of potential prompt attacks from others.
Collect your custom dataset for your projects.
Next, you need to collect the dataset, aka the knowledge base for your customized ChatGPT. Take this project, Quantum Researcher, for example. I collected essential editorial information from the typical quantum computing journal to help my agent supervise and discuss the manuscript with me.
Use Chat-GPT as a peer researcher, not someone to finish your critical thinking.
Once I have built this GPT, I can work with this technology to help accelerate my learning process, augment my imagination, and advance my research quality. But we need to remember that we need to be careful about using this technology to maximize our benefits, not just be an intelligent auto-complete tool. And here are some common prompts I will ask the ChatGPTs.
Be cautious about this technology. Alignment is equally vital to acceleration, augmentation, and advancement.
I am a big believer and advocate for effective accelerationism. "Advancement is inevitable". However, in advancing human knowledge, it remains critical to think about designing these technologies to align with human values while safeguarding our data and privacy.
Overall, I still view this technology with a positive outlook and perspective. I can't imagine what you can build and learn with this Quantum Researcher GPT and your own version of GPTs!
