Description: Leslie Joe (left), Czarina Argana (middle-left), Kai Ni (middle-right),
Daphne Fabella (right), and Chanel Fraikin (not pictured) attend the SDxAI Hackathon.
San Diego, the world of AI is here! The SDxAI Hackathon held at UC San Diego took place over the weekend of July 15, and two DSA members attended as mentors for the event. Joined by several DSA interns, the team learned more about what a hackathon is, what resources the event provides, the types of projects participants create, and what sets SDxAI apart from the rest.
Perhaps waking up for an 8 a.m. start time is not everyone’s ideal weekend kick-off, but for Saturday and Sunday, July 15 and 16, I was raring to go! That weekend marked the first SDxAI Hackathon, hosted inside UC San Diego’s beautiful Design & Innovation Building. The Data Science Alliance (DSA) team was invited to attend as mentors and onlookers, and so with DSA Graphic Designer Czarina Argana and several of our interns, I headed inside.
After checking in and mingling with other participants and staff, the event was underway, and we learned more about what a hackathon is and what sets this one apart from the rest. To begin with, a hackathon is a competitive event in which teams of 1 to 4 participants—also known as “hackers”—use the weekend to start and finish a project. These teams then present their final products to judges and potentially win big prizes provided by event sponsors.
There is often a theme within each hackathon, and this one was about the booming and blooming field of AI. Given the explosive rise of this technology, it was no surprise an event around generative AI models and their applications was brought to San Diego so quickly.
Unlike other non-hackathon competitions I'd attended, SDxAI hosted workshops and supplied resources that the hackers could use. Four workshop hosts and sponsors were Replit, Qualcomm, Log10, and a combination of ETH SD and Polygon, who all offered “bounties,” or bonus prizes, for the team who used their material most successfully and creatively in their hackathon project.
As mentors, DSA guided hackers in applying responsible data science practices to their projects. Want a tool that will assist with ensuring secure data storage and privacy? We got that! Need help determining what dataset type will be most ethical and lead to an accurate outcome? We can help! I was also given the opportunity to hold a short 15-minute talk of my own for the hackers where I shared the Responsible Data Science Framework and how they too can apply its guiding principles to their projects, both in the hackathon and anywhere else. It was excellent to link up with so many enthusiastic community members and talk about the framework that should exist around this rapidly growing field, and see the incredible projects made by these passionate hackers in just one weekend.
In addition to the attractive prize pools, the pure experience of the event was enough of a draw to many. Meal times and an after-party provided opportunities for hackers, hosts, and anyone else to mingle and network. Many attendees were there without the intent to submit a project for judging, which was entirely understandable. DSA is no stranger to attending events outside of our field, networking with diverse folks, and seeking experiences that can provoke ideas and perspectives which can help startups and projects grow. Throughout the event, there was a strong emphasis on having a meaningful and purposeful experience as an attendee.
At the after-party, I had the pleasure of chatting with some of the winners and bounty titleholders, including the first place winner, Basu Jindal. He created “Control Pre-trained Transformers,” which allowed users who talked to Large Language Models (LLMs)—such as ChatGPT—to provide larger prompts (i.e. provide more context) and still receive adequate answers from the LLM. When prompts given to LLMs are longer than the size of prompts that they were trained on, users will get an inaccurate reply. Jindal’s project sought to resolve this by using the LLM’s existing methodology and freezing the original weights, but then adding a layer by fine-tuning some additional weights that would allow the long context to be “stitched together” in a fashion.
Read More About:
The Takeaway
Beyond the inspiring enthusiasm and success of the hackers, something that stuck with me is the story of a past hackathon that one of the other mentors shared with us: Years ago, there was a nonprofit that hosted a hackathon. They provided several prompts that they genuinely needed help resolving but couldn’t afford a team of full-time data scientists and researchers. Instead, they used the hackathon as an opportunity to tackle those prompts. Upon hearing that, my mind started racing with ideas of how DSA can help our nonprofit partners and enable them to do even more amazing work. I remember all four of us looking at each other with a gleam in our eyes after hearing that story. It was only about a year ago that DSA began one of our biggest projects with fellow San Diego nonprofits, blossoming from an innovative idea at an event into collaborative work for helping others. So perhaps you’ll want to keep an eye out for what events DSA has in store for the future, and we’ll hopefully see you there!