Posts

Annually, the Mozilla free software community recognizes 25 people who are leading the next wave of the internet with the Rise25 Awards, which were awarded in Dublin, Ireland, on August 13. Aaron Gokaslan, who received both his undergraduate and Master’s degrees in computer science with Brown CS and is currently a PhD student at Cornell University, was nominated and chosen as a Rise25 honoree for the 2024 cohort.

Aaron traveled to Dublin to participate in the award ceremony and was invited to an exclusive welcome dinner with Mozilla leadership and Rise25 cohorts. The theme for the 2024 Rise25 was artificial intelligence, and the committee received hundreds of nominations of change-makers ensuring that the future of AI is responsible, trustworthy, inclusive, and centered around human dignity.

This year’s honorees, Mozilla says, are the “architects of trustworthy AI, including engineers and data scientists dedicated to the principles of open source, open data and open science. They focus on technical proficiency and responsible and ethical construction. Their work ensures AI is secure, accessible and reliable, aiming to create tools that empower and advance society”.

Aaron’s award honors his work in attempting to make artificial intelligence more accessible to everyone, including open-source projects like PyTorch and open-source datasets like OpenWebText. The latter is a project used by millions of people on a daily basis, including companies such as Facebook.

“This award acknowledges a commitment to not only doing good AI research, but doing research that really benefits the openness, transparency, and researchability of the technology,” Aaron says. “This type of contribution was historically not something celebrated in academia or industry – often you don’t get papers out of doing this kind of work, and when you do, some of the contributions can be thought of as obvious [to academics].” 

But, as Brown CS faculty member James Tompkin says, “hindsight is 20/20, and we all know that academics excel at finding ways to feel superior to others – Aaron’s accomplishments are necessary to sustain a research community.” James is an Associate Professor of Computer Science and was Aaron’s research advisor while at Brown. They continue to collaborate: “Aaron is always contributing to more projects than I can count – he never seems to stop.”

Some of Aaron’s other accomplishments include co-authoring the RAIL license, a type of open-source license designed specifically for ensuring that open-source AI can remain accessible while also putting reasonable usage restrictions on the model. As shown in a recent paper (“On the Standardization of Behavioral Use Clauses and Their Adoption for Responsible Licensing of AI”), this license is used by over 40,000 models on HuggingFace, a global platform for the machine learning community. 

Aaron also recently released CommonCanvas, a series of diffusion models trained entirely on Creative Common Images, which are images that people have effectively donated in the public domain, and showed that these types of text-to-image models can be trained on less data and more cheaply than previously thought.

“We curated a dataset of 100m images and released them on HuggingFace,” Aaron explains. “At the time of this writing, this is the largest dataset on HuggingFace Hub, and it’s also the first diffusion model trained with openly licensed images. Now, anyone can train Stable Diffusion-quality models for as little as $10-20k, which will open up whole new avenues of research.”

Aaron has also worked on the policy side, advocating to ensure that pending AI regulations do not make it harder for people to open-source their models for others to build off of. Aaron believes that keeping AI open also allows us to apply lessons learned from large language models (LLMs) to other fields. 

“If we did not make these LLM techniques available outside of large tech companies, these types of research projects would be impossible in academia,” Aaron says. “This kind of advocacy work is historically not the kind of work that academics have placed a high value on, but we should. Our research futures depend on the ability to make contributions that have an impact – an essential aspect of an engineering discipline.”

Brown CS regularly publishes news articles about our pioneering and innovative alums. We have no financial involvement in any of the companies mentioned above and have not been compensated in any way for this story. The views and opinions expressed above are those of individuals, and do not necessarily state or reflect those of Brown CS, nor does their publication here constitute an endorsement of them.

For more information, click the following link to contact Brown CS Communications Manager Jesse C. Polhemus.