Open Source AI and the Two Contracts

The language of openness is everywhere in AI right now. As someone who has spent a long time in open source, open education, open government, and open data, it is increasingly noticeable that the words, and the technology output, are in service of profit and power - losing the origins of human intention and value. Business and community values have always co-existed, but I observe that division getting much, much wider.

I have been paying close attention.

I listen with a lens of someone deeply experienced in context-switching to bring businesses together with community. For example, when I am speaking to community, to humans - I speak about the mechanisms for empowerment that allow their success in open communities (career, justice, safety, control, privacy ); this is also where I am most comfortable.

When I speak to businesses, groups of funders, or others (AKA humans with OKRs) I switch to speak about innovation, cost reduction, risk mitigation, hiring pipelines, brand-value etc. I am less comfortable in this space, but I know that the outcomes can still help humans and thus I have gotten quite good at it.

You might say, that it goes beyond context - that there have always been two different 'contracts' required to sustain and grow open source - one being for humans, and the other business.

While complimentary in some ways, there's always been friction between business contracts, and human contracts - we saw the Ethical Source license rise as part of that tension. Its also not uncommon to hear 'FOSS purists' as a dismissal of attempts to align technology as a contract with humans, and not profit.

I observe that the gap between these two contracts is widening to a point that we are witnessing a rupture. Recent moments brought this into focus for me.

“Doubtless some of you are affronted by my modest use of an LLM. You think that LLMs are “fruits of the poisoned tree” and must be eschewed because they are saturated with the sin of their origins. I think this is a very bad take, the kind of rathole that purity culture always ends up in.”

- Corey Doctorow on his usage of an LLM for his writing (6 years of Pluralistic)

He uses a so-called “open source LLM” and that’s very much how he presents his values but open-source LLMs do not really exist. You can download some weights but cannot understand what went into them or really change or reproduce them. Open source AI is just marketing and openwashing. Acting Ethically in an Imperfect Word, tante who later posted an Olive-branch update.

Corey, has built his esteemed career as a critic of technology, and yet - when criticism applies to his own business contract, he seed to find it a little harder. That's because , for him, context has switched - it's about his livelihood and reputation. There's nothing inherently wrong with caring about your livelihood, - but human need for openness and transparency is a feature not a flaw.

Prof. Emily M. Bender described more clearly how Corey could have averted both is own reaction (and those of others) by describing what LLM he was using (and thereby teaching others) how to use Data Statements to pivot back to the human contract.

There's a lot of great work like this 'out there' but still very difficult to pull together in the moment, I get that too. It also need to be easier.

Another example (though not specifically about openness + AI, includes anything open that Open AI builds), was this statement from CEO Sam Altman, who is clearly not focused on the human contract (environment):

"People talk about how much energy it takes to train an AI model, relative to how much it costs a human to do one inference query. But it also takes a lot of energy to train a human," he said during a Q&A session hosted by The Indian Express. "It takes like 20 years of life and all of the food you eat during that time before you get smart."

More subtle, is the article from Mozilla's CTO Raffi Krikorian making the case for open source AI as the path to national sovereignty. A serious argument, genuinely important to the human contract on some level. I got stuck on this statement however:

Mozilla is building a program that "leverages the open-source community to build at real scale and speed - AI sovereignty won’t come from renting Big Tech’s models, Raffi Krikorian, CTO, Mozilla

The means by which scale gets achieved ]is by 'leveraging' the community; described as a monolith available - with open source AI we know that the community is not a given; there has to be a contract and it must have equal mention .

Other more alarming definitions like this from the White House, is arguably not a human contract. Without the structures of trust we'e built over the years - with bots maintaining projects - like, yikes how easy will this actually be?

“We need to ensure America has leading open models founded on American values.” (White House AI action plan)

If we want to know whether open source AI is actually operating under the human contract 1) that reflects intention, honors labor, respects the environment, and can be trusted 2) we can't wait for it to emerge on its own. Trust in open source is built deliberately, through inclusive leadership, codes of conduct, governance, contribution guidelines, licenses - trust. That infrastructure didn't happen because people hoped for the best. It happened because communities decided to be intentional about what their contracts required.

We need that same intentionality now, before the business contract becomes the only one anyone is building for.

Want to talk about this? Join the March 13th, Open Source Practitioners Community Call : "How do we keep the human contract in open source?"

We are also working on this problem space with the AI Alignment in OSS, CHAOSS Working Group.