Updates to the Internet's Navigational Infrastructure: Index to Open World
Relational content and visible context from the application of AI (Large Language Models) to internet navigation. Summarising recent research and prototype for designing with AI.
How can AI’s development become more tangible for a wider audience to facilitate increased participation at all levels of influence in this ongoing project?
I’ve spent the past year immersed in a Californian rush of philosophy, art and technology. The groups of people I encountered often have a deep sense of excitement despite, or perhaps due to, working on what they perceive to be some of the most pressing challenges we face as a planet.
This can be an intoxicating feeling but risks seeming smug, entitled or disconnected in the face of the many other people who are deeply concerned by the potential negative effects of proposed ‘solutions’ in their political, environmental, and day-to-day lives.
So, how to intertwine this empowered excitement into the worlds of those who may feel alienated by what they are told is ‘necessary progress’? How could this inform and shape the direction of technical, regulatory, and cultural development?
I don’t have any neat answers but I will keep chipping away.
Using one part of the practice developed on the fellowship I recently participated in, I'll interpret the previously mentioned, sometimes amorphous and esoteric, Californian excitement. One means to understand the current potentials and limitations of technology (or any other institution for that matter) is by first bringing into view the ideas, assumptions and resulting implications that define how it is understood and built.
experimenting
In a recent project, I applied the above approach by using a combination of research into contemporary design literature and experimental approaches to building AI applications. I embodied this into a historical and conceptual lineage of digital design alongside opportunities to imagine products built from alternative perspectives.
The project is intended to provide an insight into how AI, Large Language Models (LLMs) in this case, seem set to alter the experience of the internet. The key theme for me after going through this process was that LLMs enabled an increase in quantity and diversity in the type of relationality for online experiences and information. Simply put, LLMs offer an additional infrastructure for navigating and understanding the internet.
LLMs often act in a professorial manner, giving a synthesised answer that provides a rich ‘vibe’ with connections between sources perhaps never made before. More specific details can then be found through additional questions or reference requests. However, if wanting to create anything original, these responses must be seen as partial answers or prompts rather than anything decisive for your specific line of questioning. There must be a recognition that the answers and connections, which may be very enriching, are idiosyncratic distillations by ‘the professor’ based on their experience.
This is in addition (and contrast) to the search engine which acts more like a librarian - upon a specific question, a rich index (or reading list) of media relating to the topic is produced, but usually with no thematic connection between references. Both approaches are uniquely valuable for finding and understanding new media but often provide the greatest benefit when experienced in conjunction.
visible context
One part of this project is an LLM & web scraping enabled app, 'Visible Context', that gives access to previously only implied or 'invisible' context of any twitter/X user. The context mentioned is the type of information that only becomes available to someone who has spent weeks/months/years noticing the themes, characters and connections in the orbit of any other twitter user or topic. For other Gen-Z readers (or hip members of older generations), AI is very effective at picking out ‘vibes’ from text-based content. It does this in a reliable (but quickly familiar) way which is useful as a first step but doesn’t yet come close to a practiced ‘vibe connoisseur’.
Unfortunately, due to the new inhibitingly high twitter API costs, this app is no longer live and is only a demo ... I'm very open to reviving/mutating it if you want to collaborate (DMs open), or you could just steal it.
The functionality of Visible Context is still easily experienceable by using ChatGPT or Bard - copy and paste text/tweets, then ask for themes and associated references (authors, artists, topics etc.) until satiated.
digital dead ends
There is already an overabundance of relevant content available to stimulate and develop our interests but the current limitation seems to be how we can access it (a function of at least technical ability and motivation).
LLMs provide connections that overcome what used to be ‘digital dead ends’ - internet journeys that could only be partially unblocked if a content creator had included suitable hyperlinks or references (but would usually be limited to the sources they used). For now, the price for this new infrastructure is regularly evaluating the results against your standards for truth, which is a measure I’ll leave you to determine for yourself…