Chang Liu of Ohio University, Athens, OH

I. INTRODUCTION 

Large Language Models (LLM) such as ChatGPT1, enabled by Transformers [1], are capable of responding fluently in many human natural languages as well as most common programming languages. They can help introduce programming basics through natural language processing of user inputs and AI generation of computer code [2,3]. This means that users can learn programming concepts and syntax by interacting with LLMs like ChatGPT using their own native languages. For example, users can ask questions about programming concepts and receive answers in computer code along with natural language explanations. This makes it easier for people who are new to programming to learn the basics without having to learn a new syntax of a new programming language first. 

II. WORKSHOPACTIVITIES 

We propose a workshop activity in which the participants are invited to join an interactive adventure into the world of LLMs and learn how to code with ChatGPT as a co-pilot. In this interactive, hands-on session, the participants will take a guided tour to see live actions of generative AI tools such as ChatGPT in code generation and problem solving. 

The event will start with an introduction of key concepts behind large language models and more generally, natural language processing (NLP), including: 

Language models 
Word embeddings 
Skip gram 
1 GPT stands for Generative Pre-trained Transformer. 
• Transformers 

• Large Language Models 

Additionally, constraints and caveats of LLMs and their relationship to AGI (Artificial General Intelligence) will be discussed. Hallucination examples will be shown to illustration the limitations of LLMs. 

Next, we will show how LLM Chatbots can assist in coding. For example, we can get started by asking ChatGPT to build a basic “Hello, World!” program using Python. Then we can ask ChatGPT to improve the program to greet a user with more specific information and also ask for something from the user as input. In particular, we can ask the program to be expanded so that it will help a user to compute their BMI (Body Mass Index). Furthermore, we can ask the co-pilot to help create a motivating game centered around the BMI concept so that player will not only learn about BMI calculation but also be motivated to achieving healthier BMI numbers. 

The downside of this approach will also be demonstrated in the process. Inaccuracy and bugs in the code will be shown in live testing sessions to illustrate current limitations of the tools. 

If participants bring a laptop computer, they can follow along and learn coding in their own OpenAI accounts. Otherwise, they can watch the live interactive session and contribute their own ideas about how to guide the AI to help with coding. Python and other programming languages will be used, although no prior programming experience is required. 

There is no specific hardware requirement for the laptop computers because all activities will take place on the cloud. A common web browser on most computers, including Chromebook, will be sufficient. 

III. SUMMARY 

LLMs have great potential in enabling end user programming and assisting in rapid prototyping. The proposed workshop is just an initial attempt to bring awareness of this potential to the participants. 

REFERENCES 
[1] Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz 
Kaiser, and Illia Polosukhin. "Attention is all you need." Advances in neural information processing systems 30 
(2017). 
[2] Fiannaca, Alexander J., Chinmay Kulkarni, Carrie J. Cai, and Michael Terry. "Programming without a 
Programming Language: Challenges and Opportunities for Designing Developer Tools for Prompt Programming." 
In Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems, pp. 1-7. 2023. 
[3] Vaithilingam, Priyan, Tianyi Zhang, and Elena L. Glassman. "Expectation vs. experience: Evaluating the 
usability of code generation tools powered by large language models." In CHI conference on human factors in 
computing systems extended abstracts, pp. 1-7. 2022. 

Biography:

Professor of Computer Science, School of Electrical Engineering and Computer Science, Ohio University, Athens, Ohio