Narratives Around High-Speed Rail: An AI Analysis
General description of the project
We are engaged in a cutting-edge project that leverages advanced AI technology, specifically the GPT-3.5 language model, to conduct a comprehensive analysis of narratives surrounding high-speed rail systems in the United States. This innovative approach involves processing extensive textual data from academic research, policy documents, and public discourse to extract key narratives, insights, and trends.
One notable success of our project lies in the identification of underrepresented narratives within the high-speed rail discourse. We have successfully uncovered narratives emphasizing the crucial aspect of equitable transportation access, particularly for minority communities, including the Hispanic population. By highlighting these narratives, our project has given voice to previously marginalized perspectives and contributed to a more inclusive and equitable transportation discourse.
Our approach proves to be highly cost-effective. Unlike traditional, time-consuming manual content analysis, our AI-driven methodology significantly reduces the resource and time investments required for discourse analysis. This cost-saving aspect is particularly beneficial for projects in resource-constrained environments, such as academic research within higher education. Using the GPT-3.5 technology, our preliminary results indicate that a total of $42 was spent, showcasing the potential for cost-effective research.
The insights generated by our project have directly contributed to informed decision-making processes, especially in the realm of transportation policy. By identifying dominant narratives, we’ve provided stakeholders with a clearer understanding of public sentiment and policy priorities, aiding in the formulation of more effective, responsive policies.
The project’s usefulness extends to a diverse range of stakeholders, from policymakers to researchers and educators. It facilitates better-informed decisions, academic research, and discourse analysis, all while remaining cost-effective due to the efficiency of AI-driven analysis.
In line with a commitment to diversity and inclusivity, our project has actively focused on narratives that impact Hispanic communities. By recognizing the unique transportation needs and challenges faced by this demographic, we’ve contributed to a more inclusive and equitable transportation discourse.
Through our project, we have learned the importance of continuous adaptation and refinement in AI-driven analysis. Staying abreast of evolving narratives and societal priorities is crucial to maintaining the project’s relevance. Additionally, it has reinforced the significance of interdisciplinary collaboration, as addressing complex transportation issues often necessitates diverse expertise.
In summary, our AI-enhanced project explores and analyzes high-speed rail policy narratives with a specific focus on the Hispanic community. It showcases the effectiveness of leveraging GPT-3.5 for cost-effective, informed discourse analysis, contributing to more inclusive transportation policies and decision-making processes.
Technologies
We used OpenAI, the creators of ChatGPT, to fine-tune a powerful language model that has capabilities to work with human language specific to the domain of high speed rail policy in the United States. A language model is what ChatGPT is essentially but the difference between our model and ChatGPT is that our models are trained differently. ChatGPT was trained through endless text data from the internet, this method is called few-shot learning. For our model, we used a method called fine-tuning which trains the model within a certain domain, in our case text documents relating to high speed rail policy. ChatGPT has clear benefits but the internet does not always have the best information. Our model differs because it was able to be specialized through our selected high quality data, it became specialized in high speed rail discourse.
Davinci is an example of a language model, specifically one of the versions of OpenAI’s GPT-3.5 (Generative Pre-trained Transformer 3). GPT-3.5 models are designed for natural language understanding and generation tasks, making them language models that excel in a wide range of text-related tasks, such as text generation, summarization, translation, question answering, and more. Davinci, like other GPT-3.5 models, is a practical tool for working with human language and is widely used in various applications and services.
The Davinci model at the time of original use was typically one of the most advanced and capable versions of GPT-3.5 that OpenAI offers. It can generate high-quality and coherent text and is suitable for a wide range of natural language understanding and generation tasks. It is important to note that models are constantly being improved by Open AI and whenever a new version of a GPT is updated a model can become a legacy model or become extinct. Thus the user must regenerate their data onto an updated model equivalent. This essentially means that the nature of the finetuning process demands it not be finite. We as researchers are in the process of updating our fine tuned model with the new GPT 3.5 Turbo Instruct.
Explain project results
Focusing this research on a dimension of how it benefits and hurts hispanic communities was vital to our understanding of A.) how policy functions and who it functions for and B.) empowering and uplifting stories we don’t traditionally get to hear C). understanding that on the heels of every innovative tech there is also a trail of untold stories of the people most negatively impacted by that – how to improve practices. We believe that high-speed rail policy analysis, particularly with a Hispanic focus, can have several meaningful benefits for your school and its students. By focusing on the Hispanic community’s transportation needs and concerns within the high-speed rail policy context, our project can help increase the representation and visibility of this demographic. This is especially important in educational settings that strive for diversity and inclusivity.
Our project can provide valuable insights into the specific transportation challenges and opportunities facing Hispanic communities. This information can be used to advocate for more inclusive and equitable development and implementation of transportation policies which can benefit both students and the community.
Our project’s findings can serve as a foundation for further academic research. Students and faculty interested in transportation policy, urban planning, or related fields can use this research as a starting point for their own studies, creating opportunities for academic growth and innovation.The data and analysis generated by our project can be used as educational material in relevant courses. Professors can incorporate this real-world case study into their curricula, providing students with practical examples of policy analysis and research. Students who engage with our project can gain valuable experience in policy analysis, data collection, and research. This experience can be instrumental in their career development and future job opportunities, especially if they are interested in policy-related fields.
Our project can foster community engagement and partnerships. Collaborating with local Hispanic communities and organizations to gather data and share findings can build valuable relationships and strengthen the school’s ties to the community. Encouraging a focus on Hispanic issues in our project can help raise awareness and promote cultural competence among students and faculty. Understanding the unique challenges faced by Hispanic communities is a step toward a more inclusive and diverse educational environment.
Why it should be considered best practice?
As we reflect on the process, certain best practices emerge that proved instrumental in achieving optimal results. Understanding the importance of a well-curated dataset has been paramount. The data used for fine-tuning should align closely with the specific task or domain at hand, ensuring the model captures the intricacies and nuances required for accurate outputs. Cleaning and preprocessing the data, including careful removal of noise and irrelevant information, contributed significantly to the model’s performance.
Effective prompt engineering emerged as another cornerstone of successful fine-tuning. Crafting prompts that are clear, concise, and aligned with the task’s objectives facilitates the model’s comprehension and responsiveness. Experimenting with different prompts, variations, and contextual cues allowed for a nuanced understanding of how the model responds to various inputs, ultimately refining its capabilities.
Regular model evaluation and iteration played a pivotal role in honing the fine-tuned model. Continuous assessment of performance on validation sets and real-world examples allowed for the identification of strengths and weaknesses. This iterative process empowered me to make targeted adjustments, optimizing the model’s behavior and ensuring its adaptability to a range of inputs.
Ethical considerations were woven into every stage of the fine-tuning process. Being mindful of biases within the data and refining prompts to mitigate potential ethical concerns fostered responsible AI development. Striking a balance between achieving task-specific accuracy and upholding ethical standards became a guiding principle, underscoring the importance of ethical AI practices.
Documentation emerged as a practice that cannot be overstated. Comprehensive documentation detailing dataset specifics, fine-tuning parameters, and notable observations during the process proved invaluable. This documentation serves as a reference point for future endeavors, fostering knowledge transfer and ensuring continuity in the event of model updates or team changes.
As I conclude this fine-tuning journey, the importance of community engagement stands out. Sharing insights, challenges, and lessons learned with the broader community fosters a collaborative environment for knowledge exchange. OpenAI’s vibrant community forums provided a platform for discussion, feedback, and collective learning, enriching the fine-tuning experience.
In retrospect, fine-tuning OpenAI GPT-3.5 has been a dynamic and enlightening process. By embracing these best practices, I’ve not only achieved a finely tuned model but also contributed to the evolving landscape of responsible AI development. The journey has instilled a deep appreciation for the intricacies of model refinement and the ethical considerations inherent in shaping the capabilities of advanced language models.
Highlights of your proposed presentation
In this presentation, we utilize OpenAI’s GPT-3.5 model to analyze narratives surrounding high-speed rail systems in the United States. Leveraging artificial intelligence, specifically GPT-3.5, enabled us to fine-tune and analyze extensive textual data, extracting key narratives, insights, and trends from academic research, policy documents, and public discourse. A notable success emerged in the identification narratives for and against the development and implementation of high speed rail systems. The project’s methodology proved highly cost-effective, showcasing the efficiency of AI-driven analysis in contrast to traditional manual content analysis. Importantly, the project’s insights directly contributed to informed decision-making processes in the realm of transportation policy, providing stakeholders with a clearer understanding of public sentiment and policy priorities. The project’s usefulness extended to a diverse range of stakeholders, and its commitment to diversity and inclusivity was evident through a dedicated focus on narratives impacting Hispanic communities. Lessons learned throughout the project included the critical importance of dataset curation, effective prompt engineering, regular model evaluation, ethical considerations, comprehensive documentation, and community engagement. These lessons underscored the dynamic and adaptive nature of AI-driven analysis and the significance of interdisciplinary collaboration in addressing complex transportation issues effectively. As we reflect on this journey, we envision future directions that involve further fine-tuning, expansion into related domains, and collaboration with other research initiatives, marking the continuous evolution of responsible AI development.
The Evaluation Committee will evaluate submitted proposals based on the following criteria. Each area will be rated on a scale from 1 to 7 (1= non-satisfactory; 7 =outstanding), for a maximum of 63 points.