AI Nonprofit CEO: Closed Research Hinders Innovation
February 24, 2025
The Allen Institute for Artificial Intelligence (AI2), a nonprofit research laboratory established over a decade ago, is advocating for a more open approach to artificial intelligence development. According to CEO Ali Farhadi, this openness is critical to fostering innovation and addressing significant global challenges.
Farhadi, who has led AI2 since 2023, previously held a position at Apple. He spoke with The Associated Press about the institute’s mission to push the boundaries of AI through open-source principles.
“The biggest threat to AI innovation is the closed nature of the practice,” Farhadi explained. He emphasized the institute’s commitment to open-source ideals, mirroring the philosophy behind open-source software.
“If you think about open-source software, the core essence was, ‘I should be able to understand what you did. I should be able to change it. I should be able to fork from it. I should be able to use part of it, half of it, all of it. And once I build my thing, I put it out there and you should be able to do the same,”” he said.
Farhadi clarified AI2’s definition of an open-source AI model, contrasting it with what he considers “open weights” models, such as those from Meta. He stated that true open source requires access to every component of the AI model, not just its parameters.
One of the barriers to greater openness, according to Farhadi, is the proprietary nature of the training data used to develop AI models.
“Some of these training data have a little bit of questionable material in them,” he remarked. “But also the training data for these models are the actual IP. The data is probably the most sacred part.” He noted the significant investment companies make in gathering and curating this data, explaining why they are hesitant to share it.
Looking ahead, Farhadi expressed excitement about AI’s potential to revolutionize scientific discovery, particularly in complex areas like gene and cell structure research. He argued that AI can be instrumental in formulating and solving these computational problems.
“Our hope is to empower scientists to be able to actually train their own model,” he explained, highlighting the need for researchers to have the tools necessary for AI development.
© 2025 The Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed without permission.
