In the aftermath of the OpenAI governance battle, the narrative has been dominated by the idea that “the capitalists won.” While this conclusion is catchy, it misses the point.
The real (and bigger) story is: The founders of OpenAI set out to create a nonprofit public institution to ensure that AI’s evolution aligns with the interests of humanity at large. Last month, it became clear that they failed.
What is also clear: Public institutions that prioritize the public good over short-term profits are desperately needed for the AI era.
History has shown that transformative technology created by the private sector needs to be balanced out by public option. Options should be serving needs not met by the wonderful (yet often one-dimensional) offerings of private players.
A century ago, the spread of devices needing electric power led to the establishment of public electric authorities, thus ensuring that everyone could benefit from the technology. Similarly, as commercial television grew into a pervasive entertainment medium, the creation of the Public Broadcasting Service (PBS) ensured that educational content wasn’t neglected. This pattern has played out during most eras in the modern history of technological innovation.
The same has been true in the internet era. Wikipedia emerged to ensure that the “sum of all human knowledge” was both created by and held in public hands as we digitized our society. The Linux and Apache foundations emerged to ensure that the core plumbing of the internet is owned and managed collectively. And the Mozilla Foundation was formed to ensure that the technology that runs the web was open to everyone, and not controlled by a single company.
The success of these organizations serves as a counterpoint to the narrative that emerged from the Open AI fracas last week: that nonprofits shouldn’t be trusted to create the technology that underpins our digital lives. As the examples of Wikipedia, Linux, and Mozilla show, this argument is just silly. Nonprofit public institutions have been a critical part of creating a vibrant, open, and innovative digital society.
Seemingly, the founders of OpenAI were prescient—they knew we’d need institutions to ensure that AI would be built in ways that would continue to prioritize the public interest. Seemingly, they wanted to create the institution that owned this turf, that defined this category. Either they got it wrong—or more likely they changed their minds. However, it is also clear that institutions like this are still desperately needed. Someone needs to fill this void.
On a positive note, there is a burgeoning movement to establish institutions that aim to do exactly this. Eleuther AI, a community-driven initiative, and the Allen Institute for AI, a nonprofit created by Microsoft cofounder Paul Allen, are making significant strides in the field of open source AI. On November 17, a $300M French nonprofit AI lab named Kyutai was launched to “democratize AI through open science.” Similarly, we launched Mozilla.ai earlier this year with a comparable mission.
All of this gives me hope that there is a deeper and longer story to be told than the one we heard the past few weeks. Yes, OpenAI’s nonprofit governance model failed—and OpenAI failed to become the public institution its founders envisaged. Regardless, we are still well on our way to building the public technology institutions we need for the AI era—institutions that will ensure AI truly serves humanity by acting as a complement and counterbalance to whatever the private players build.