Unlocking Success: Equality of Opportunity vs. Equity - The Real Winner -- 🦛 💌 Hippogram #26

Yesterday I posted a conversation between the former Digital Minister of Taiwan, Audrey Tang, and Frederike Kaltheuner at thiss years re:publica in Berlin.

Frederike Kaltheuner, raided the question were we need to invest to increase plurality and suggested that investing in open source AI, we would still need to depend on infrastructure cloud that's neither open, public, nor accountable. It's owned by companies like Amazon, Google, and Microsoft. This idea that open source increases the dependencies from BigTech was wide spread through a paper from David Gray Widder, Sarah Myers West and Meredith Whittaker "Open: Big Tech, Concentrated Power, and the Political Economy of Open AI". This paper was wide spread among policy maker during the EU Act negotiations, and I strongly criticized it as bringing a unilateral view that distorts reality.

Audrey pushed back a bit on the idea that open source would not decrease our dependencies and gave the same answer I need to give people on a weekly basis.
Audrey mentions that the language model she uses to reply to her emails is entirely trained on her MacBook, which runs on Linux.

"The point is, we can now do fine-tuning and training for language models with personal computers. The idea that power has to be concentrated in the cloud is a myth propagated by the largest frontier labs. They want you to believe the edge isn't powerful enough, but that's not true. " - Audrey Tang

Having talked to policy makers and debunking the myth that was spread through the paper, I find it hard to ignore that this distortion is deliberate and stems from a desire for implementing highly centralized, state-controlled monopoly and social policies that prioritize collective ownership and management of resources, with extensive state intervention in various sectors, resembling structures seen in certain historical examples that never went well. Today, these approaches are sold under the banner of equity, which focuses on creating equality of outcome. This stands in hard contrast to open source movements that create equality of opportunity.

Open-source AI minimizes central control, leading to a more decentralized approach to AI development. Over the past 20 months, this has fostered innovation by enabling more individuals and smaller entities to contribute and experiment without the constraints often imposed by large, centralized organizations. To illustrate this, I recently spoke with a student in Harare, Zimbabwe, who, with less than $50, built an AI reading assistant powered by a local multimodal LLM, Llava, running on a 1 GB RAM Raspberry Pi 3B+.

Open source is providing him with equality of opportunity. Without access to these open-source models, he wouldn't have had the chance to build his project. The simple reason is that he doesn't have a credit card to pay for OpenAI APIs to begin with. Last year, while traveling in Guatemala, I had a conversation with some brilliant researchers that made me realize how our current scientific publication practices hinder their chances of becoming part of the global scientific community. Their universities cannot afford the publishing fees, and their primary means of accessing research papers is the controversial and illegal SciHub website. Have we sold out by turning scientific knowledge into assets rather than treating it as digital commons to ensure equality of outcome?

Equality of opportunity also enables social mobility, allowing individuals to improve their socio-economic status based on merit and hard work, reducing income inequality over time. Encouraging a free open source approach can empower individuals and communities to actively participate in global development processes. Open source is fundamentally about the creation of digital commons, not commercial practices. It focuses on building shared resources that everyone can use and contribute to, fostering collaboration and collective growth rather than individual profit. This community-driven approach can help ensure that AI technologies benefit a broad spectrum of society. For instance, the internet itself and the role of internet service providers (ISPs) serve as a useful analogy. The internet is a decentralized network that has grown and thrived because it operates as a global commons, where anyone can contribute and access information. This openness and lack of central control have driven unprecedented innovation and connectivity. Now it's important to note that these are all related to the creation of commons, not the adoption of it commercial practices. Regulation needs to implement safeguards and collaborative governance structures preventing the negative outcomes associated with power imbalances and inequalities. Governments should never attempt to run organizations to build open source themselves; instead, they should create funds to develop and maintain digital commons. The EU failed to do this with cloud technologies, leading to current power asymmetries and market failures.

So, I'm glad that Audrey challenged the conclusions of the well-cited paper, calling them a myth and a lie. I genuinely fear those who deliberately ignore the current progress of decentralized AI, as their arguments don't add up. I understand the appeal of centralized language models as a means of controlling thought, especially in our turbulent environment where the tools to spread narratives have become decentralized. Western governments celebrated these tools during the Arab Spring but fear them when their own societies adopt these new dynamics.

If a political agency could control language, it would also control thought, making dissent not only impossible to express but even to conceive. Orwell's warnings in 1984 serve as a cautionary tale about the dangers of allowing any entity to have unchecked control over information, language, and truth. His work underscores the importance of maintaining a free and open society where information is not monopolized by those in power. It should serve as a warning, not a manual for a totalitarian future.

Shouldn't governments invest in open source? Of course they should, Audrey, highlighted the importance of public investment across every level of the technological stack. In her systematic approach, which is detailed extensively in her new book, Plurality: The Future of Collaborative Technology and Democracy it is essential that communities have the necessary resources to network and support each other locally, within their social context, without relying on abstracted national or foreign entities. This ensures that communities, including civil groups, can share resources and scale up to meet their needs, especially if personal devices are insufficient. This vision empowers communities to implement technology in a way that is relevant and beneficial to their unique contexts.

I have always admired Audrey's work, as she appears to be one of the rare individuals who has found effective solutions while living in a country bombarded with misinformation. During her time as a digital minister the trust within society towards the government significantly increased. Tang advocates for "radical transparency," which involves making all government discussions and meetings accessible to the public. This contrasts with conventional practices where government operations often occur behind closed doors, limiting public insight and engagement. Her response stands in stark contrast to what I have observed in Germany, the EU, and the US.