In Short:
Generative AI often faces criticism for issues like copyright concerns, biases, and high resource consumption. However, at a recent MIT-affiliated hackathon, participants developed innovative tools for journalists. They created “AI News Hound,” a prototype that helps reporters find relevant research papers by analyzing AI-related data from Arxiv, Reddit, and news articles. This highlights the potential of AI for valuable applications.
Generative artificial intelligence presents a range of challenges, including the unauthorized use of creative work, inherent biases, and significant resource requirements for training. These issues cannot be overlooked; however, the potential of generative AI to facilitate the development of innovative tools is noteworthy.
Experience at Sundai Club
My firsthand experience at the Sundai Club, a generative AI hackathon held monthly near the MIT campus, illustrated this potential. A few months ago, I attended a session where members focused on developing tools beneficial to journalists. This initiative is supported by a Cambridge nonprofit organization called Æthos, which advocates for the socially responsible use of AI.
Collaboration and Project Development
The participants included students from MIT and Harvard, along with several professional developers and product managers, as well as a representative from the military. Each meeting begins with a brainstorming session to propose potential projects, which the group subsequently narrows down to a final concept for development.
Innovative Journalism Tools
During the hackathon, notable project suggestions included:
- Utilizing multimodal language models to monitor political discourse on TikTok.
- Automatically generating freedom of information requests and appeals.
- Summarizing local court hearing video clips to enhance local news coverage.
Ultimately, the team opted to create a tool aimed at assisting reporters covering AI by identifying pertinent research papers from Arxiv, a well-known repository for research paper preprints. My interest in this area may have influenced their decision, as I expressed the importance of discovering intriguing research during our meeting.
Prototype Development
Once the project’s objective was established, the coding team leveraged the OpenAI API to create a word embedding—essentially a mathematical representation of words and their meanings—focused on Arxiv AI papers. This enabled them to analyze the data for research papers related to specific terms and to examine the interconnections between various research domains.
By integrating another word embedding derived from Reddit threads and conducting a Google News search, the developers produced a visualization tool that correlates research papers with online discussions and relevant news articles.
Introducing AI News Hound
The prototype, named AI News Hound, is in its initial stages but demonstrates the capabilities of large language models to facilitate the discovery of information in new and engaging ways. A screenshot from the tool illustrates a search for the term “AI agents,” highlighting research papers alongside related news and Reddit discussions.
This initiative by the Sundai Club exemplifies a promising direction for leveraging AI in journalism, showcasing its utility while acknowledging the challenges that lie ahead.