Code-recommendation systems, such as Copilot and CodeWhisperer, have the potential to improve programmer productivity by suggesting and auto-completing code. However, to fully realize their potential, we must understa...
详细信息
ISBN:
(纸本)9798400703300
Code-recommendation systems, such as Copilot and CodeWhisperer, have the potential to improve programmer productivity by suggesting and auto-completing code. However, to fully realize their potential, we must understand how programmers interact with these systems and identify ways to improve that interaction. To seek insights about human-ai collaboration with code recommendations systems, we studied GitHub Copilot, a code-recommendation system used by millions of programmers daily. We developed CUPS, a taxonomy of common programmer activities when interacting with Copilot. Our study of 21 programmers, who completed coding tasks and retrospectively labeled their sessions with CUPS, showed that CUPS can help us understand how programmers interact with code-recommendation systems, revealing inefficiencies and time costs. Our insights reveal how programmers interact with Copilot and motivate new interface designs and metrics.
As Genai technologies, particularly Large Language Models (LLMs), continue to revolutionize programming and data science, it is increasingly vital for educators to adapt computer science curricula. This paper presents...
详细信息
ISBN:
(纸本)9798350362060;9798350362053
As Genai technologies, particularly Large Language Models (LLMs), continue to revolutionize programming and data science, it is increasingly vital for educators to adapt computer science curricula. This paper presents a review of recent technical books on ai-assisted programming and utilizes the findings to guide curriculum changes in higher education. Our analysis underscores the necessity for novel teaching strategies, emphasizing skills like problem decomposition, top-down design, and advanced debugging. Furthermore, it emphasizes the crucial expansion of curricula to encompass courses on developing applications based on LLMs, utilizing libraries such as LangChain and incorporating Retrieval Augmented Generation functionality. Our analysis reveals a significant gap in technical literature regarding the ethical and societal impacts of Genai, highlighting the urgent need for programming curricula to evolve and equip students with the skills required to ethically develop ai-enhanced software products. This paper advocates for curriculum development that not only aligns with the latest industry trends but also contributes to research on ai-assisted coding and its future impact.
Artificial intelligence (ai) permeates modern society and is poised for further integration across various domains. However, there exists a notable deficiency in equipping K-12 students with foundational ai understand...
详细信息
ISBN:
(纸本)9798400706332
Artificial intelligence (ai) permeates modern society and is poised for further integration across various domains. However, there exists a notable deficiency in equipping K-12 students with foundational ai understanding. This paper introduces a novel learning framework that leverages large language models (LLMs) and strategic gaming to teach K-12 students about the inner workings of ai. The framework consists of a chatbot programming and testing IDE that enables K-12 students to construct ai from scratch, engage in strategic gameplay to generate instant training data, and improve the ai heuristics with a data-driven learning mechanism. With a tiered curriculum catering to diverse proficiency levels and fostering synchronous collaboration, this framework efficiently adapts learning experiences to suit various groups of students, thereby facilitating learning at scale. Preliminary experiments validate the feasibility and vast potential of this approach, promising to revolutionize ai education in K-12 education.
The advent of generative ai tools presents novel opportunities and challenges in computer science education, particularly in introductory programming courses. This study explores the implementation of ai-Lab, a framew...
详细信息
While current chat-based ai assistants primarily operate reactively, responding only when prompted by users, there is significant potential for these systems to proactively assist in tasks without explicit invocation,...
详细信息
This paper delves into the capabilities of Large Language Models (LLMs) in supporting beginner programmers in crafting virtual reality (VR) content for gaming and simulation applications. Our research group, comprisin...
详细信息
ISBN:
(纸本)9798350374537;9798350374544
This paper delves into the capabilities of Large Language Models (LLMs) in supporting beginner programmers in crafting virtual reality (VR) content for gaming and simulation applications. Our research group, comprising both seasoned and novice VR developers, engaged with versions of ChatGPT, including 3.5, 4, and Bard with Gemini Pro, during the development of various VR projects, spanning games to simulations. The findings, reflected through a comprehensive analysis of code quality and usability metrics, reveal that LLMs offer significant advantages for newcomers in the field. However, we also identify several challenges that must be addressed. This study enriches the discourse on obstacles faced in VR development and elucidates the impact of LLMs on coding and software creation, shedding light on both the advantages and hurdles of integrating LLMs into VR game production.
Evaluating conversational assistants, such as GitHub Copilot Chat, poses a significant challenge for tool builders in the domain of Software Engineering. These assistants rely on language models and chat-based user ex...
详细信息
ISBN:
(纸本)9798400706851
Evaluating conversational assistants, such as GitHub Copilot Chat, poses a significant challenge for tool builders in the domain of Software Engineering. These assistants rely on language models and chat-based user experiences, rendering their evaluation with respect to the quality of the Human-ai conversations complicated. Existing general-purpose metrics for measuring conversational quality found in literature are inadequate for appraising domainspecific dialogues due to their lack of contextual sensitivity. In this paper, we present RUBICON, a technique for evaluating domain-specific Human-ai conversations. RUBICON leverages large language models to generate candidate rubrics for assessing conversation quality and employs a selection process to choose the subset of rubrics based on their performance in scoring conversations. In our experiments, RUBICON effectively learns to differentiate conversation quality, achieving higher accuracy and yield rates than existing baselines.
In June 2023, the Imperial College London Graduate School's Research Computing and Data Science group invited thirty PhD students for a one-day Hackathon on ai-assisted programming. This poster abstract presents o...
详细信息
暂无评论