Back to Syllabus.
If you work in a team, design and coding responsibilities must be split relatively evenly, and your coding contributions should be modular so as to easily identifies each member’s coding contributions, though you should be working on a single overriding design. In many case, software will be your primarily deliverable. Grading of individuals will be comparable to those who work individually or in group. You aren’t simply allowed to use AI tools for any and all of the coding — you are required to. The use of chatGPT as a programming tool by itself satisfies the requirement that Project 3 be relevant to AI creativity and human-AI co-creativity.
Here are some project 3 ideas, but typically you will have to scale it back to address one particular functionality that you imagine would be needed.
- Business: Intelligreetings: Consider a company that designs and creates greeting or note cards using AI tools to create the visual art and AI to create the text. What should the original software do? (This company name suggested by chatGPT: “‘IntelliGreetings’ could be an unused title for a greeting card company that uses AI to generate images and text. The name combines “intelligence” and “greetings,” which conveys the idea that the cards are smart and innovative. Additionally, it has a catchy and memorable ring to it, making it easy for customers to remember and recognize the brand.“)
- Research, Business: Prompts R Us: Using Professor White’s et all prompts, formalize their description, in say first order logic, and design software that will automatically combine various basic prompts into composite prompts. This might be done without actually needing data, other than the skeletal prompts themselves.
- Business: Old and in the Way Programmatics: Create a company that specializes in legacy software in still-used languages such as Cobol and Fortran and Common Lisp, as well as creating documentation for the software.
- Education: VUCS assignments. Design programming assignments for any courses in the VU CS curriculum, with solutions (created specifically for CS 3891) in a number of programming languages. If you have done any of these solutions before, then compare your solutions to your solutions with chatGPT. We must be careful of honor violations on this depending on the course, so it may be best to create new assignments and solutions for them.
- Business, Research, Education: Finds and Evaluates topics for shows like “This is Nashville“
- Education, Research: Serious games. Implement a serious game — a simulated world trade game — that is to be played by AIs, each representing a virtual country, for “international relations” that is used in my AI project courses. Well suited to a team project where implementing the AI(s) and the environment would occupy different members of a team.
- Research: Breaking the rules through symbolic/subsymbolic interaction: It is sometimes said that creativity involves “breaking the rules”. This project investigates how creativity can be introduced at the interface of symbolic processing and subsymbolic processing. Our premise, stemming from earlier work, is that symbolic AI offers the appropriate language for developers (or AIs) to specify intelligent processes, but that subsymbolic AI assumes a finer grained space in which more accurate processing can occur. As an analog, symbolic AI offers a high level programming language and subsymbolic AI offers a finer-grained machine language. By translating a program, say a classifier, from a symbolic representation to a subsymbolic representation, the classifier can be revised in the subsymbolic space, and potentially be translated back into the symbolic space for reasons on comprehensibility, but also now capturing the improved performance that was realized in the subsymbolic revision. The revisions that occur at the subsymbolic level can represent incremental improvements even when translation is made back to symbolic descriptions, or the translation back may possibly reflect a revolutionary change in the symbolic space. (The concept of classifier “instability” is relevant here.). This project is concerned with facilitating these ‘revolutionary” changes that “break the rules” relative to the original symbolic program, this uncovering and encouraging creativity in the interaction between representational levels.
- Research: Identifying Deep Fakes by Content Inconsistency with the Known. Work on identifying deep fakes by imperfections in the rendering are a dominant applied research topic, but looking at inconsistency between the content of and what is known and factual, as represented by semantic networks and the like, is a novelty.
Back to Syllabus.