Bootstrapping the Open Source AI "commons"...
...is a not-so-stealth project, business and obsession of mine these days
Ever since people started to notice that I stepped down from my CTO role at the company I co-founded (still rooting for y’all ZEDEDA peeps, tho!) I felt like I kept re-living that “can’t wait to see what you do next, Joe” scene from Halt And Catch Fire finale. Go ahead — watch it — it is pretty iconic and we’ve got time:
People kept asking about my future company or my next project (right after opening with “are you still in Cyprus?”) but the truth is I don’t have a title or a role to offer you. Yet. What
and I have for you is the following challenge: you see, there's a little trend we've been witnessing lately, and it's got everything to do with those fancy schmancy AI models. But here's the kicker, folks. The trend we're talking about is the centralization of training and deployment of these models within just a handful of big-shot corporations. Yep, you heard me right. It's like a group of monopolistic moguls playing a game of "who's got the biggest AI model" and “where’s my moat, dude?” while the rest of us are left scratching our heads, wondering what the heck happened.Now, hold on a second. You might be thinking, "But Roman, what about the open source community? The fearless innovators who paved the way for AI over the past few decades?" Well, my friend, they're feeling a little left out in the cold. They had all these incredible ideas, pushing the boundaries of what AI could do, only to find themselves on the sidelines as the AI world becomes a playground for the chosen few (with legislators eager to oblige by enshrining them as gatekeepers, because… oooh… Danger! Danger! Danger!).
We believe that economically viable, decentralized, self-governed, developer-centric “Open Source AI” alternatives are not just important but downright critical for the future of business and society. They're like the Gandalfs of AI, fighting against the Saurons of centralization, shouting, "You shall not pass!" to the closed systems that threaten to engulf us all.
While the usual suspects would try to go out of their way to make you believe that there’s absolutely nothing that needs to be created to make this dream a reality (after all we got to GPT-4 just fine using what developers already have). We invite you to consider the following questions: are the following tools and mechanisms really enough
Traditional Open Source licenses for model implementation and weights’ IP management?
Traditional Open Source foundations as governance structures?
GitHub as a collaborative platform for working on model implementations?
Public cloud providers as your only source of compute power for both training and inference?
Traditional events and conferences (after all, where’s FOSDEM of AI?)
AI hackers on-ramps (where’s O'Reilly Media of AI?)
Of course, it isn’t like nothing is happening to address these questions. I’d really like to raise a glass to HuggingFace, DagsHub, TogetherAI and all the Davids to the Goliaths of the AI world. They're here to remind us that we can't let a few bigwigs control the fate of AI. No, no, no. We need diversity, we need competition, and most importantly, we need the little guys to have a say in this crazy AI party. But most importantly, we need to start building all the rest of the ecosystem and that’s where we would like to invite you on a journey with us.
And just as much as we shouldn’t be soothed by BigCo’s assuring us that they “will handle AI on our behalf” I think we shouldn’t assume that we can delegate it fully to the old school Open Source organizations either. Now, don't get me wrong. I'm not here to diminish the expertise of organizations like the Open Source Initiative (OSI). They've undoubtedly accumulated a wealth of knowledge in their pursuit of all things open source: a portion of it will even apply to AI verbatim. At the same time, it seems to me that some of their organizational DNA is just incongruent with the needs of this emerging community (more on that in later posts). The same skepticism applies to the governmental AI efforts. Frankly, the only role I wish governments would play is to step aside and allow the actual hackers among us to figure out our own governance. Let them regulate the companies peddling AI products, sure. But when it comes to the very building blocks of AI that we're constructing, meddling should not be tolerated.
The organization and the ecosystem for these “AI commons” doesn’t really exist yet, but it is high time we start building it: let’s “seize the means of AI production”!
We are looking at all of you students, professors, developers, product managers, customers, investors and entrepreneurs. Actually, regardless of who you are, if any of the above resonate with you please contact us via LinkedIN, email, Twitter or even Instagram and together we'll have a chance to shape the future of AI in a way that benefits everyone. It's like a group hug for algorithms, a dance party for neural networks, and a stand-up comedy show for chatbots. Because when it comes to AI, folks, we're all in this together.
Why should you join us, instead of us joining you? Fair question. Check us out on LinkedIN by browsing through our collective experience. Take a look at what we have built in the past. If after that you still think we should be joining you instead – let us know that too!
P.S. And if it all fails, I will see you in my humanities class at a Cyprus Community College.