The second annual Sushi Hackathon gathered at Stanford technologists, entrepreneurs, scholars, and students to explore the application of generative AI for societal benefit and encourage the development of AI-powered solutions to advance sustainability and productivity in the fisheries sector.

Held on October 3, 2025, and hosted by APARC’s Japan Program in partnership with GDX Co., Ltd. and SMBC,  the event was headlined by Audrey Tang, Taiwan’s cyber ambassador-at-large and its first digital minister. Tang, named one of TIME’s “100 Most Influential People in AI” and a champion of digital freedoms, shared insights on the civic, regulatory, and ethical dimensions of artificial intelligence in a polarized digital era.

Following her keynote remarks, Tang joined in a fireside chat with Gita Wirjawan, former Minister of Trade of the Republic of Indonesia and a visiting scholar at Stanford’s Precourt Institute for Energy, delving into the political and regulatory implications of generative AI.

Sign up for APARC newsletters to receive our experts' updates >

Epistemic Security and Portable AI

Tang positioned epistemic security, the preservation of shared realities, as a cornerstone of democratic resilience in the age of generative AI. Referencing Silicon Valley as an “epicenter of innovation,” Tang urged attendees to think beyond disruption toward horizontal alignment between humans and AI agents.

She framed the global challenge of “PPM” (Polarization Per Minute), a phenomenon in which digital engagement intensifies social division, as a central threat to civic cohesion. In response, Tang advocated for a design philosophy rooted in civic care and public, portable, and pluralistic AI systems that enable communities to retain agency across technological platforms. Concepts such as AI portability and human context protocols were introduced as tools to ensure interoperability and continuity in digital governance.

“Whatever super-information highway we have can have an offramp,” Tang remarked, emphasizing the need for forward-compatibility in AI services, from social networks to national governance systems. “Remember that pluralism is here,” Tang said, calling for AI to be a force for cooperation, not fragmentation.

Regulation, Resilience, and Digital Sovereignty

Wirjawan opened the fireside chat by challenging the premise of the internet as a democratizer of ideas, asking whether the abundance of intelligence would deepen existing polarization. Tang responded by highlighting intentionality and “creation care” as antidotes to algorithmic fragmentation. While cautioning against regulatory overreach, she pointed to initiatives like the Engage California Project as efforts to recalibrate policy-making to better reflect demographic diversity and public needs.

Both speakers discussed the monopolized dynamics of today’s information ecosystem and emphasized the importance of public options as alternatives to corporate platforms. Tang noted that Taiwan’s approach, eschewing bans in favor of preference-driven design, offers a compelling model, as seen in the declining usage of TikTok in favor of locally grounded alternatives.

In a wide-ranging dialogue that touched on small language models, geothermal sustainability, and the geopolitics of AI, Tang warned against replicating the arms races of the Cold War. Instead, she proposed a “third horizon” vision for AI: one that centers pluralism, safety, and a clear understanding of the promises technologies are built to uphold.

“Building AI without a vision of the social contract,” she cautioned, “is like developing a nuclear weapon without understanding nuclear safety.”

AI for Sustainable Fisheries

The second part of the event was devoted to presentations by the teams of university students and young engineers selected to showcase their AI-enabled solutions for sustainable fishing practices in the Global South. 

Teams presented solutions for fish stock monitoring, catch optimization, supply chain transparency, and market hedging, tackling key pain points for small-scale fisheries. Projects like PillSnap’s CARP and Kapybara’s SonarSync showcased the potential of AI to bridge ecological sustainability with economic empowerment in coastal communities. 

The winning team, Sushinnovation, was awarded the top prize for Polaris, a machine learning system that uses sensor data from fishing boats to detect early signs of engine failure. These student innovations underscored the Hackathon’s goal of making AI socially attuned, adaptable, and accessible. 

Read the Stanford Daily’s coverage of the event, below.

Teams Roll Out for Second Annual Sushi Hackathon

By Jack Quach and Kate Quach

Two weeks of intense research, coding and testing came to a peak in the Arrillaga Alumni Center on Friday for the second annual Sushi Hackathon. Sixteen teams made live presentations with closing pitches to compete for a $30,000 grand prize.

GDX Co. Ltd., a Japanese e-commerce company, partnered with the Stanford Shorenstein Asia-Pacific Research Center (APARC) to host the event, which invited participants to develop innovative artificial intelligence (AI) projects for supporting productive and sustainable fishing. Project groups included students from Stanford and universities nationwide, as well as engineers from leading technology companies including Nvidia, Google and OpenAI.

“I believe deeply in the power of collaboration across borders, where diverse ideas and perspectives come together to create long-lasting engagement,” GDX Co. Ltd. COO Kenjiro Ikawa said.

With more than double the number of applicants seeking to compete at the Sushi Hackathon compared to last year, Ikawa acknowledged the growing urge to tackle current-day problems through AI.

Ikawa said that fishers and sushi chefs face a multitude of challenges, including declining fish captures, poor quality of catches, climate change and supply chain inefficiencies. He hoped that the hackathon would bring engineers ready to confront obstacles in the fishing industry and help GDX Co. Ltd. “maximize opportunity to find treasure” and innovative solutions.

Throughout the day, teams rose to present and answer questions from the audience and six judges. Projects included dashboards for Peruvian fishing co-operatives to plan trips, market analysis of fish prices and on-board voice assistants that used machine learning to read sonar data aloud to fishers.

Attendees reconvened in the evening to hear the announcement of the top three prizes. After nearly four hours of presentations and judging, Sushinnovation was awarded the top prize.

The winning team, which was comprised of four students from the University of California (UC) Santa Cruz, UC Davis and San Jose State University (SJSU), used a sensor installed on fishing boats to detect signals indicating potential signs of engine failure or mechanical issues. Their project, Polaris, is a machine learning program designed to interpret these warning signals, notify fishers about problems and guide them through the repair process. The Sushinnovation team used transformer architecture, similar to the backbone of large language models (LLMs) like ChatGPT, to create Polaris.

Rome Drori, a senior from SJSU, likened the feeling of winning to “relief.” The team faced a variety of technical challenges, even in the minutes leading up to their live presentation.

This year’s win was also a story of redemption for the team: during the first Sushi Hackathon in 2024, Sushinnovation finished in second place.

The team said that this year, they focused on fishers’ needs first by driving to Half Moon Bay to meet with fishers before conceptualizing their idea. From there, they decided to solve a problem many fishers faced: dealing with mechanical failures that prevented them from spending time fishing.

“I listened to a series of great presentations with creative, innovative ideas and cool, cool demos,” Kiyoteru Tsutsui, sociology professor and APARC director, said.

Tsutsui described Sushinnovation’s reaction to winning — leaping into the air with loud cheers — as “pure joy.”

The members of Sushinnovation plan to continue their studies, with many saying they would use their split of the grand prize for their tuition. Drori said he would soon return to studying for his midterm — but only after attending a sushi dinner hosted for the winners.

Team Pill Snap, composed of UC San Diego students Brian Liu, Shawn Pana, Caylin Canoy and Reagan Hsu, earned third place. They focused on addressing the health and mobility of fishermen. “I was very shocked to learn that one in three fishers experience carpal tunnel syndrome,” Liu said.

“We believe that for the welfare of fishers, we care not only just about the income they make, but also their health in general,” Canoy said. Their collective concern for the fishers’ nerve conditions motivated them to build a specialized glove that reduces risk of carpal tunnel syndrome.

“Knowing the needs of fishers in their daily life from their requests is something that should be taken more seriously, especially in AI solutions,” said Zijian (Carl) Ma, a first-year Stanford Ph.D. candidate in bioengineering and first-time hackathon participant.

The event also featured highlighted speakers who presented on the broader implications of AI on global public speech, economy and democracy.

“You are the first generation, if you are a young person, to be born in an environment of very high PPM,” said Audrey Tang, a former minister of digital affairs in Taiwan and keynote speaker of the event. PPM, she explained, stands for “polarization per minute.”

Tang emphasized the importance of “public, portable and pluralistic” developments in AI as the technology becomes increasingly integrated in everyday life.

She contrasted the “vertical alignment” of companies using AI to drive up engagement and clicks with the need for “horizontal alignment,” which emphasizes strengthening people’s relationships with each other, AI and the natural world.

“Are we building AI systems to supercharge profit, or are we building it to foster cooperation?” Tang said.

Tang also acknowledged the surging energy demands of massive LLMs like Claude or ChatGPT, which can require up to 10 times as much energy as a traditional search. The solution, she said, may lie in using much smaller models specialized for certain tasks, like language translation. These models would demand significantly less energy and could be more accurate than those built for a wide range of general tasks.

Tsutsui shared that learning about the hackers’ solutions “gave [him] hope that the future will be filled with a lot of new innovations that take advantage of and leverage generative AI.”

“When we see machine learning, let’s make it collective learning,” Tang said to participants at the end of her keynote. “When we see user experience, let’s make it about human experience.”

Jack Quach ’27 is a News staff writer and was a Vol. 266 desk editor for science and technology. He is from San Francisco, CA, and in his free time loves cheering for his hometown sports teams, exploring the outdoors, learning new recipes and being the official™ S.F. expert/tour guide for his friends.

Kate Quach is a high schooler writing as part of The Daily’s Summer Journalism Workshop.

Media Coverage