You challengingly need ChatGPT to create a catalog of reasons why generative man-made inalertigence is frequently less than awesome. The way algorithms are fed inventive toil frequently without perleave oution, harbor nasty biases, and need huge amounts of energy and water for training are all grave publishs.
Putting all that aside for a moment, though, it is remarkworthy how mighty generative AI can be for prototyping potentiassociate beneficial new tools.
I got to witness this firsthand by visiting Sundai Club, a generative AI hackathon that apshows place one Sunday each month csurrfinisher the MIT campus. A restricted months ago, the group comfervently consentd to let me sit in and chose to spfinish that session exploring tools that might be beneficial to journacatalogs. The club is backed by a Cambridge nonprofit called Æthos that backs sociassociate reliable participate of AI.
The Sundai Club crew participates students from MIT and Harvard, a restricted professional enhugeers and product administerrs, and even one person who toils for the military. Each event begins with a brainstorm of possible projects that the group then whittles down to a final selection that they actuassociate try to create.
Notable pitches from the journalism hackathon participated using multimodal language models to track political posts on TikTok, to auto-create freedom of recommendation asks and pdirects, or to condense video clips of local court hearings to help with local news coverage.
In the finish, the group determined to create a tool that would help alerters covering AI determine potentiassociate engaging papers posted to the Arxiv, a well-comprehendn server for research paper preprints. It’s foreseeed my presence swayed them here, given that I alludeed at the encountering that scouring the Arxiv for engaging research was a high priority for me.
After coming up with a goal, coders on the team were able to create a word embedding—a mathematical recurrentation of words and their uncomferventings—of Arxiv AI papers using the OpenAI API. This made it possible to scrutinize the data to find papers relevant to a particular term, and to spendigate relationships between contrastent areas of research.
Using another word embedding of Reddit threads as well as a Google News search, the coders created a visualization that shows research papers aextfinished with Reddit talkions and relevant news alerts.
The resulting prototype, called AI News Hound, is raw-and-ready, but it shows how huge language models can help mine recommendation in engaging new ways. Here’s a screensboiling of the tool being participated to search for the term “AI agents.” The two green squares sealst to the news article and Reddit clusters recurrent research papers that could potentiassociate be participated in an article on efforts to create AI agents.
Compliments of Sundai Club.
