As Baidu raced to develop its ChatGPT equivalent, employees working to advance AI technologies for drug discovery had to suspend their research and cede their computer chips to develop the chatbot instead. The current AI paradigm is also choking off alternative paths to AI development. The number of independent researchers not affiliated with or receiving funding from the tech industry has rapidly dwindled, diminishing the diversity of ideas in the field not tied to short-term commercial benefit. Companies themselves, which once invested in sprawling exploratory research, can no longer afford to do so under the weight of the generative AI development bill. Younger generations of scientists are falling in line with the new status quo to make themselves more employable. What was once unprecedented has become the norm.

During my conversation with Brockman and Sutskever, I had referenced one of them: the carbon footprint of training such models. In June 2019, Emma Strubell, a PhD candidate at the University of Massachusetts Amherst, had been the first to coauthor a paper showing that the footprint for developing large language models was growing at a startling rate. Where neural networks could once be trained on powerful laptops, their new scale meant their training was beginning to require data centers drawing significant amounts of energy from carbon-based sources. In the paper, Strubell estimated that training the version of the Transformer that Google used in its search for just a single cycle—in other words, feeding it some data and letting it compute a statistical model of that data—could consume roughly 1,500 kilowatt hours of energy.

Within weeks of the white paper, OpenAI and Microsoft formed a strategic alliance with Google and Anthropic to launch the Frontier Model Forum, a group for advancing relevant research and influencing the policy agenda on AI safety risks. It was a rare issue in which the interests of Doomers, Boomers, and profit-motivated corporates aligned: Keeping frontier models front and center in government regulatory discussions was ideologically imperative to Doomers and convenient to Boomers and corporates for shifting attention away from regulating existing AI models and their problems. Everyone also advanced their causes by arguing against opening up the weights of cutting-edge models. In the Forum’s first year, Meta was conspicuously absent. (It would join a year later to gain a seat at the table.)

The formula for dissolving empire thus requires the redistribution of power along each axis. The suggestions and recommendations I lay out here are exemplary but by no means comprehensive. First, to redistribute knowledge, we need greater funding to support its production outside the empire. That involves supporting researchers who can conduct independent evaluations of corporate models so we are not solely reliant on companies to understand their capabilities. It involves supporting organizations like DAIR that can pursue completely new directions of research, such as new forms of AI beyond large language models that are more efficient with data and energy. It involves supporting organizations like Te Hiku that can pursue task-specific, community-driven AI applications that strengthen marginalized communities.

With this transparency, we would also begin to redistribute power along our second axis: resources. By hiding the ingredients of their models as their intellectual property, the empires of AI have thus far been able to get away with seizing other people’s IP without credit, consent, or compensation. Visibility into company training data would make such extractive and exploitative behavior far more difficult. So, too, would visibility into company supply chains, including where they contract their labor and where they’re negotiating new leases of land to build more power plants and data centers, which so often happens under shell entities.