Greetings, young wizards, witches, and muggles! Born in a bygone era, yet learning in today's world, set to thrive in tomorrow's challenges, it is the duty of our little shop Wit & Wisdom Wares to equip you with the invaluable knowledge and tools essential for your future triumphs. All it cost is your time and dedication. How about an AI-based Wand? Whenever you come across a term, concept, or simply a statement that is unclear to you, I strongly encourage you to use it for AI-enhanced learning. (Don't forget to refer to ChatGPT in the Glossary section.) It is crucial for you to hone your skills in utilizing AI tools to elevate your learning.
In the classroom, I will share with you what I learned to write better prompts. It's essential to remember a few key points: The outputs given by AI are influenced by the data it was trained on, so always verify their accuracy before you publish them as they will exhibit errors and/or bias. As you delve deeper into the world of AI, stay firmly rooted in AI ethical practices. It is your responsibility to employ AI responsibly and considerately, always adhere to the ethical guidance outlined in the Ursinus Student Handbook. You MUST embrace this tool's potential but do so with discernment and responsibility. You are not alone on this journey. Reach out to me anytime you are uncertain.
ChatGPT is an advanced computer program that can understand and generate human-like text. The "Chat" part is just like chatting with a friend. The "GPT" part is a bit more technical - it stands for "Generative Pre-training Transformer."
This might sound like a lot of mumbo jumbo, but let's break it down:
- "Generative" is just like it sounds. ChatGPT can generate or create its own responses to whatever you type. It's not just choosing from pre-written replies - it's making up new ones on the spot.
- "Pre-training" is like the homework it did before it started chatting. It studied a huge amount of text (like, imagine reading the internet) to learn how words and sentences flow together.
- "Transformer" is the name of the super-smart architecture that powers the model. It's like the engine of a car, helping ChatGPT to run and generate those conversations.
Taken these together, it is a huge leap in artificial intelligence, especially in the field of natural language processing. Let's see how ChatGPT is developed: first, the ChatGPT model is trained with a massive amount of data to learn the word patterns from all these materials and gradually calibrate all its parameters, so that, when you are feeding a sentence (prompt) to it, it could use these calibrated parameters to predict the most likely next word of the sentence. It would then combine this newly generated word with the previous feeding sentence together to predict the next most likely word. This model is constantly learning and adjusting its parameters to form a complete reply sentence. Sometimes, these newly created sentences may contain incorrect information. After this faulty information are usually corrected by human reviews, the parameters are further calibrated. Therefore the program is evolving.
ChatGPT as a program does not store any of the previous trained information, rather it stores calibrated parameters that it could use to calculate out the answers. In ChatGPT-3.5 version, it has about 175 billion independent parameters to adjust. While, in the newer ChatGPT-4 version, this number jumps up near 1000 times (971 times, to be precise), and reaches 170 trillion parameters. When comparing the performance of ChatGPT3.5 version with ChatGPT4 version, people often referred that "intelligence explosion" occurred in the ChatGPT4 version as it could provide far more understanding of the prompts and produce far better answers (intelligence, or creativity). How come? In an average human brain, we have around 86 billion neurons, which form an intricate network of connections. These cells have somewhere between 100 trillion to 1000 trillion synapses. These synapses are the points of communication between neurons, allowing them to transmit simple signals to each other in order for brain to THINK. You must have noticed that the number of parameters in ChatGPT-4 is similar in number to the ones of synapses in human brains, which correlates with the "intelligence explosion".
I will stop here, and you can carry on the thinking. Does our brain function with any similarities as a large language-based AI model? Both brains and AI models can process and interpret information. Human brain processes information in parallel across various interconnected regions, allowing for complex functions such as perception, memory, reasoning and understanding. It has the abilities to comprehend and intercept information, make connections and form meaningful representations. While AI model is a product of machine learning algorithms and processes text or input data sequentially, apply learned patterns and associations to generate responses. What is "understanding", what is "intelligence"? Read this if you are interested.
Throughout history, knowledge has been passed down through generations via verbal communication and/or written sources like books and papers. Within any given generation, it's impossible for any individual to read every book or fully grasp the vast knowledge spanning across all fields. Consider the expertise in ancient Aztec history - someone with sufficient knowledge in this niche could potentially become a specialized professor due to the rarity of such expertise. However, as large language-based artificial intelligence models continue to develop and advance, the transmission of knowledge is bound to undergo inevitable changes. No longer will expertise be easily attained through possessing specific knowledge alone. Imagine a world where you could gain access to ALL accumulated knowledge in human history simply by asking the right questions. The way we learn is on the verge of transformation, so get ready to embrace these exciting changes.
Let me provide you with an example of how the world would be different. I serve as an Associate Editor of a peer-reviewed journal, and my role involves seeking advice from a diverse group of experts to assess the quality and scientific validity of submitted manuscripts. In the past, I often received complaints about grammar mistakes and English misuse in manuscripts authored by individuals from non-English-speaking countries. However, I believe that with the widespread use of ChatGPT or similar AI models, this issue could become a thing of the past. By leveraging the capabilities of language models like ChatGPT, authors from non-English-speaking countries could receive real-time assistance in improving their English proficiency and writing skills. They could utilize these models to refine their work, ensuring better readability and adherence to language standards. This would not only enhance the overall quality of their work but also foster greater inclusivity and diversity in scientific research.
In this section, I will share with you what I learned about how to integrate the AI-assisted tools into your study. During the time that I am still writing about this, I would like you to think about this: what information would you need in order to grasp a FULL understanding of ChatGPT-assisted learning?
I am filling in with more information now, so check back occasionally..
HOW TO USE CHATGPT FOR LINEAR REGRESION ANALYSIS
In ChatGPT-4, you can either upload your raw data into ChatGPT-4, or typed it in as following:
I typed 2
You may ignore the Python codes, or you may copy and paste into RStudio to run python. You can also ask ChatGPT to plot this equation.
You can change every viisual elements of the figure, all you need to do is just ask.
I encountered the following statement from Publisher about their policy on using AI-assisted writing.