Personalized discovery engines for knowledge
The next massive search engine will be a personalized discovery engine. As we experiment with new platforms like AR/VR and explore new models of engagement in web3, I imagine that the next wave of human-computer interaction will focus more on “pushing” us the right, relevant information versus the user “pulling” for information.
The debate between search and discovery (and sometimes browse) is a tension between two desires: speed to information versus exploration of information. Both aim for knowledge gathering; but while Google search can tell you in .43 seconds that most of you have until April 18th to file your taxes, discovery introduces a serendipity that sends the user down a rabbit hole to find new ideas and the links between them.
This Internet of the future will feel a lot more like a collaborative library — a place to browse and unearth ideas, publish your thoughts, encounter new opportunities, and make connections across bodies of knowledge. In the early 2000s, I spent most of my days online on Tumblr, Wordpress, Reddit, StumbleUpon, Blogger, Blogspot, not always knowing precisely what I was looking for yet coming upon ideas and topics of interest. The Internet was more unruly then, and in the 2010s, we optimized search and layered in social features to support discovery via curation. Nuzzel used your Twitter following to recommend top shared readings while Medium encouraged following new writers by topic. Today, we have no shortage of writers on the web via platforms like Substack, Mirror, Foster, collaborative Roam graphs, and more. They enable us to contribute to our own public and private web, creating unprecedented volumes of observations and thoughts that we can index.
In a world where the Internet will be more decentralized and with more contributors than ever, search results that feed us information will be insufficient. Knowledge is information with context, and “language is humanity’s longest running program.” The latest advances in machine learning — namely transformers and self-supervised learning in natural language processing — will also make it possible to build personalized discovery engines that organize and surface information that is timely, relevant, and impactful. We won’t always have to know what we’re looking for, but akin to meandering around a library, our fellow readers, authors, and knowledge seekers in this new paradigm will point us in the right direction.
Special thanks to @smerity, @jamescham for our many conversations on all things language and discovery
Thank you Kim for the write up, bringing out the difference between push and pull techniques.