Will the AI revolution be democratized? It’s a long trek, but possible

Does AI have to be controlled by tech oligarchs?

The optimism about FOSS as a transition point to a more democratic economy has dimmed, to put it mildly. On the one hand, FOSS is inescapable today. Linux is as pervasive on the internet as microplastics in the ocean. On the other hand, Big Tech is far more powerful and concentrated than it was even a decade ago. The Magnificent Seven make up one-third of the value of the S&P500 (compared to 12.5% in 2016). The advent of generative AI hinted that new entrants like OpenAI and Anthropic might challenge the dominance of incumbents such as Google and Microsoft, but their capital and computing needs are so vast that they have little choice but to collaborate with the old guard. For now, Big Tech has made itself indispensable.

But maybe not forever. Claude Code and its inevitable successors mean that almost anyone can be a coder, speaking software into existence like Captain Picard with his tea. Many who have sampled the wares of advanced vibe coding see it as a Promethean moment, making amazing superpowers readily available to average humans.

When my kids were going to college in the 2010s, I told them, “Study what you want, but if you don’t learn to code, you will die penniless, homeless, and alone.” (This may be why there are no “World’s best dad” coffee mugs in my house.) They ignored me, and now they have their revenge: anyone who can follow a few simple directions can code now, just by chattering about what they want into an AI dictation program and feeding the transcript to Claude. It’s not much more complicated than that.

Imagine a world where every entrepreneur in Detroit could speak into existence customized tools for everything they need. Plumbers, furniture designers, coffee shop owners, graphic artists, accountants, home health co-ops – they can all be coders now, self-reliant and self-employed.

But, of course, they need something like Claude Code, and Claude’s parent company Anthropic has investment deals and cloud contracts with Google and Amazon, along with other partnerships with Microsoft and Nvidia. Anthropic is densely connected with Big Tech, perhaps inextricably so.

What would a more democratic alternative look like? Trebor Scholz, an American scholar-activist, and Swiss social scientist Mark Esposito propose the creation of a full “solidarity stack“. They write: “A democratic AI cannot simply rent space on the extraction stack. It requires that workers, communities, cooperatives, and public institutions reclaim ownership of the infrastructure itself, layer by layer, from the earth to the cloud.”

Some of the components are already in place. The Swiss AI Initiative, a collaboration between several Swiss research institutions, recently released Apertus, a “fully open foundation model for sovereign AI”. All of its central components are open access, the LLM meets Swiss data privacy and copyright protection laws, and it was trained on the Alps supercomputer using 100% carbon-neutral electricity. It was released under an Apache 2.0 license (comparable to copyleft) and is available on Hugging Face, an AI repository similar to GitHub.

More broadly, a social movement is emerging to point AI’s development in a more sustainable and democratic direction. This movement has the wind at its back. As Bernie Sanders put it, AI is “the most transformational economic revolution in world history”. The public does not want the direction of this revolution guided by a handful of peculiar tax-dodging billionaires who are busy building escape bunkers on distant Pacific islands. Only by enrolling the public in its development will AI avoid some of the more apocalyptic scenarios.