In this age of ever more pervasive and persuasive AI, one story I want to promote is the story of the Source Layer. The Source Layer is what LLMs draw on: fact-checked reputable sources such as independent media outlets, peer-reviewed scientific papers, and not least, Wikipedia. Not only are LLMs trained on these sources, but they also surface, aggregate and summarise from these sources - crucially, without adequate attribution - unless you make this a requirement in your LLM settings (which I recommend you do).  Already on Google search results, unsourced AI output is given primacy over any other kind of information. This is another death blow to struggling media business models. Legal work is being done to develop licencing models to ensure that media sources are paid for the information services they provide to LLMs. But even if financing becomes more just, a fundamental problem remains:  as the Source Layer becomes less visible, it becomes less valued. As it becomes less valued, it dries up. 

I find it useful to think of trustworthy knowledge like fresh drinking water: something that is necessary for human thriving, but that is only possible if you continue to work hard at cleaning and protecting your sources, and maintaining the distribution and social structures that ensure everyone has access to it. In a scenario where the Source Layer is dead or dying, we continue to get information that has the appearance of truth - beautifully phrased or visualised LLM output - but it is drawing on a river system that is silting up. There's not enough fresh water coming in. 

This issue crystallised for me on a recent project I did with the Wikimedia Foundation, the noprofit organisation that hosts and operates Wikipedia and other Wikimedia projects. The Wikipedia community has provided an essential service in making all kinds of knowledge accessible and digestible. Wikipedia editors draw on a rich and diverse series of reliable sources, depending on the topic that is being covered. They do this for no commercial or political gain, keeping each other accountable to a strict set of editorial values - which helps keep the information trustworthy and the topics broad  - much wider than click-bait news agenda or algorithm-driven infotainment. Wikipedia itself is on a continual journey to ensure that the range of contributors and topics becomes ever more representative of human knowledge as a whole. The more people who participate in this activity, the better the knowledge becomes. I’ve come to see Wikipedia as a critical part of a functioning online Source Layer, a global asset that becomes more important as truth becomes more challenged, a work in progress that we can all participate in. 

So how do we keep the Source Layer healthy? We continue to support independent media journalism and organisations that work to support a just and fair knowledge economy. We campaign for greater transparency in the use of AI in social and knowledge contexts. We continue to ask questions about what is true and what is not. We ask, Who has written this? What is their point of view? What other perspectives might be useful here?  What are the facts, and what is only hearsay? We teach our children how to question everything they read and see. We help them understand what AI is and what it is not.  And we inspire them to become contributors to the sources themselves: to become researchers, journalists, Wikipedia editors, to join in the search for truth, and not be satisfied until they are sure they’ve found it.