Browse
Recent Submissions
Now showing 1 - 5 of 188
- ItemScene and Unseen: GPT Bias in Script Writing(2024) Crawford, Charlie; Friedler, SorelleAs Large Language Models and generative AI become more and more prevalent, the question of how to measure bias in these systems becomes that much more crucial. This paper delves into a background on how these systems came to be, the biases they have been shown to carry, and the strategies researchers and developers can use to monitor these biases. Alongside this, I will seek to pull from the literature to understand the different strategies which researchers use to define fairness and bias in generative AI as a way to contextualize their audits. This literature review will focus on OpenAI’s publicly available generative pre-trained transformer model, ChatGPT, as an example of these themes throughout the paper. Following this literature review is an overview of research conducted on OpenAI’s content moderation system, the “moderation endpoint". This research was conducted in the form of an algorithm audit, using television data as input to the moderation endpoint to determine how frequently texts were flagged as violating OpenAI’s content moderation rules. For this input text, we compared real television scripts to scripts we had asked GPT-3.5 and GPT-4 to generate to determine any trends in content moderation. We ultimately found that the moderation endpoint flagged a high proportion of scripts, both GPT and human generated, but had a much higher flagging rate for real scripts.
- ItemInfluence and Equity in Faculty Co-Authorship Networks(2024) Ellis-Einhorn, Mia; Friedler, SorelleFaculty co-authorship networks reflect the connections between faculty members in academia. These connections open individuals to information access. Faculty co-authorship networks, like other social networks, can create a power imbalance between faculty who have access to information and those who don’t. In this thesis, I seek to understand this power structure and develop methods to make information access in computer science more equitable. Computer Science (CS) suffers from a lack of diversity and underrepresentation of women and minority groups. Mentorship programs are a proven, effective strategy to improve diversity in these fields. I seek to develop a mentorship pairing algorithm based on a CS co-authorship network and demographic information of faculty members to increase diversity in the field. To work towards this goal, we collect demographic information on faculty members and analyze edge augmentation strategies focused on improving information access.
- ItemAn Analytical Approach to Higman's Lemma(2024) McMahon, Eleanor; Lindell, StevenHigman’s lemma is a fundamental result in computer science and mathematics, providing applications in areas such as graph theory, and other data structures. One proof of this lemma provides an interesting use of the minimally bad sequence proof technique, originated by Nash Williams. Furthermore, this lemma sets the foundation for the finite basis property, a property focused on the potential definable nature of infinite sets. However, it is not commonly known to look at this lemma analytically rather than algebraically. By taking an analytical approach to Higman’s lemma through topological and analytical notions, new perspectives and insights may be uncovered, shedding light on other possible applications of this lemma.
- ItemImplementation and Utilization of Hardware Parallelism in Modern CPU Architectures(2024) Reed, Justin; Wonnacott, David G.For at least the past three decades, nearly all computer architectures have made use of pipelining, a concept that has proven indispensable in improving the efficiency of computer processors, regardless of the application. Pipelining improves the throughput of an instruction sequence by running instructions in mutually independent stages, thus achieving what is called instruction level parallelism (ILP). This paper provides an overview of how ILP has been utilized to develop more efficient processors, reviewing a selection of papers influential to its use. To this end, the tradeoffs between out-of-order and in-order execution are considered. Finally, the most recent work being done to improve instruction level parallelism is discussed.
- ItemQ*Bird: An Exploration into Augmented Reality Controllers(2024) Thumu, Neha; Normoyle, AlineIn this thesis, we analyze various types of controllers in a mobile augmented reality environment. These controllers are joystick, laser, and tilt. We conducted an experiment in Q*Bird, a game we developed, in order to test these controllers and find which is the most optimal in this environment.