Innovations in Journalism: AI Hackathon at MIT Focuses on New Tools
CAMBRIDGE, MASSACHUSETTS—In a world where technology is rapidly changing the landscape of industries, a group of innovative minds gathered for the monthly Sundai Club hackathon near the Massachusetts Institute of Technology (MIT) campus to explore the intersection of artificial intelligence (AI) and journalism. The hackathon, supported by the Cambridge-based nonprofit Æthos, dedicated to promoting socially responsible AI, provided a platform for students, developers, and professionals to brainstorm and prototype cutting-edge tools on a Sunday afternoon.
Held monthly, this intellectual gathering sees participation from students of prestigious institutes such as MIT and Harvard, professional developers, product managers, and even individuals from the military sector. The process begins with a brainstorming session where participants present potential project ideas. These ideas are reviewed, and one is selected to be brought to life over the course of the day.
During this particular session, the focus was on creating tools that could significantly aid journalists. Notable ideas included leveraging multimodal language models to monitor political posts on TikTok, the automatic generation of freedom of information requests, and summarization of local court hearing video clips to facilitate local news reporting.
Ultimately, the chosen project was the development of a tool tailored to assisting reporters in identifying noteworthy research papers related to AI on Arxiv, a preprint server widely used for disseminating research. The idea for this project may have been influenced by the presence of a journalist who highlighted the importance of tracking developments in AI research.
To realise this project, the team employed the OpenAI API to generate a word embedding, creating a mathematical representation of text data in the field of AI as shared on Arxiv. This tool enabled the analysis of data to find papers relevant to specific terms and allowed for the exploration of relationships between various research areas. In addition, the team utilized word embeddings from Reddit discussions and Google News to craft a visualization that mapped research papers alongside relevant news stories and online dialogue.
The end result was a prototype named AI News Hound. Though still in its developmental stages, AI News Hound demonstrates the potential of large language models in unearthing and linking information from diverse sources, offering a dynamic way for journalists to engage with AI-related research.
A screenshot of AI News Hound in action illustrates its functionality: when the user searches for "AI agents," the tool reveals two green squares that correlate research papers with discussions and news articles, suggesting them as potential components for a journalist crafting a story on AI agent development.
The Sundai Club hackathon serves as a testament to how creative collaboration between diverse expertise can generate tools that may very well shape the future of industries such as journalism, especially as AI continues its transformative march into everyday life.
Source: Noah Wire Services