Mark Pesce: How will we navigate the sea of knowledge?

Recently, something quite lovely popped up in my social media feed: the Radio Garden.

It presents itself as a three-dimensional model of the globe – with one significant addition: little blue dots against the green of the Earth’s surface. Each dot represents a radio station, in its exact location. Position yourself directly over one of those blue dots, and that station begins streaming into your web browser. It’s a lovely, obvious and joyful interface to the full breadth of the different voices that broadcast across the planet. Move the world, and change your tune. 

Music of the Spheres

Radio Garden feels like a 21st-century version of something that I tried to hack into being way back in 1995 – in the very early days of the still very new World Wide Web. For the ‘Doors of Perception’ conference in 1995, I worked with two musician friends to prototype something we called WORLDSONG. 

The idea behind WORLDSONG seemed simple enough: collect all of the range of voices – pre-recorded and real-time, human, animal and machine – into a single interface (a globe) that presented a place to ‘listen to the entire planet’ in a way that anyone could understand and access. 

Broadly speaking, the web of 1995 had very little to say for itself, with just a few rare examples of ‘streaming’ media, and even one or two streaming ‘radio stations’. On the whole, slow connection speeds meant that this very-early web didn’t allow for sound as part of its experience. WORLDSONG imagined a future where every street had an open microphone, capturing the sounds of nature and people, where anyone could create their own ‘streaming’ radio station, and cities, nations – even the whole planet – could create a joyful noise, or tune in just to listen. 

Our ambition exceeded our capacities by twenty years…

Today, we have streaming on tap from hundreds of thousands of sources – video and audio – but we don’t have an easy time finding any of that. It’s what makes Radio Garden (and WORLDSONG before it) unique and valuable, reminding us that having a wealth of wonderful sources means little if you can’t organise and present those sources in a way that makes it obvious and joyful for anyone using them. 

We have so many channels from so many available sources and through so many mechanisms that we quickly and unwittingly overwhelmed our own capacities to bring some sort of organisation to all these voices.

We could use a search engine, but text search presupposes you know more-or-less what you’re looking for. If you don’t – or if you’re exploring, to see what you can find by serendipity – text search simply fails.

That keeps us restricted to small islands that reflect our own experiences, expectations and recommendations. Anything outside of these has become increasingly invisible to us. 

Although it has allowed us to find our way through an unknowable vast web of knowledge, text search (which almost always means Google) looks as though it has met its match – and found its limit – due to recent advances in ‘Generative AI’ tools like GPT-3. These tools have ‘read’ the entire internet, and can return text that seems to make sense based on any other occurrence of similar text they’ve come across.

Notorious for spitting out statements that make semantic sense but fail on logic or factual grounds, these systems have already been put to work generating new content for the web, generally the kind of simple, pre-digested facts that, with a bit of optimisation, rise to the top of a search engine query. Sprinkle in a few programmatic ads and that single top-of-category answer, generated by a machine, earns a few pennies a day. Get a machine to generate a few million such answers and you have a real business.

Increasingly in 2023, this ‘dark forest’ Web will see human-generated knowledge swamped by algorithmically generated and search-engine-optimised results. As that happens, the value of text search – already in decline over the last several years – will fall precipitously. We may well return to the state of nature we confronted twenty-five years ago: a web too big to know, and impossible to search. 

So what happens now?

We already have several other ways of searching for knowledge: Wikipedia for factual information, Google and Apple maps for information about places, plus more constrained searches within social media systems such as Reddit.

The web will not become utterly unknowable – but the way we approach the internet and its dark forest of AI-generated mendacities has already begun to change. We can no longer freely ‘just Google it’, the way we did a decade ago. We became too dependent on text search, and that attracted anyone looking to grab a share of our attention for their own purposes. 

Increasingly in 2023, this ‘dark forest’ Web will see human-generated knowledge swamped by algorithmically generated and search-engine-optimised results. As that happens, the value of text search – already in decline over the last several years – will fall precipitously.

We have difficulty apprehending the web largely because it presents itself to us as a single point, a bit like a singularity of all knowledge. There’s no way to make any sense of that point – so we beam messages into it, and hope that the reply makes some sense to us.

Yet for hundreds of thousands of years we have had no trouble managing the incredibly information-dense world around us. Hunter-gatherers still maintain an understanding of the natural world, dense with exquisite detail. Something feels completely natural to them, precisely because their knowledge can be situated within the world.

We have always been good at making sense of the world (it’s why we’re still here) and don’t seem to have any natural limits around where it concerns the information richness of our environments. Our abilities appear to adapt naturally to the wealth of the world. 

Tying our wealth of information to the world – as the Radio Garden does, and as WORLDSONG attempted – looks to be an interesting way forward for us. At just the moment that text search surrenders to an AI invasion, we find ourselves with augmented reality systems – today on our smartphones, tomorrow on spectacle-like always-on devices – which will be consistently adding ‘digital depth’ to our experience of the world.  

Everything in the world has a story to tell. To this point we either knew those stories because of the culture we belong to, or because we searched them out. We’re approaching a moment when those two paths for stories about our world merge into a single stream, where what we know from our culture’s sense of place and what we know from or collective, shared, digitised wealth of knowledge marry, giving birth to a single, unified view of the world and our place within it.   

Already in place, the elements of this transformation look ready to meet our need for a future of knowledge that truly can be called ‘navigable’.