Anytype: Pioneering the future of trustful collaboration through knowledge graphs
Of end to end encrypted knowledge graphs, AI's context problem, Graph Neural Networks and proofs of identity, data provenance and compute.
We live in an era where the management and utilization of information have become the cornerstones of progress. This holds particular significance in the context of AI, where information and its intricate interconnections can greatly enhance the capabilities of machine intelligence. This post delves into the concept of knowledge graphs and their role in AI model training as pioneered by our portfolio company Anytype.io.
Anytype is a company community of 100,000 people built on the foundations of freedom of expression and data sovereignty, putting the user at the center stage of a next generation super-app. The early version of the product is a powerful knowledge management tool which leverages bi-directional links, new types of data structures and aesthetic, freely composable interfaces. As a heavy user I can attest that Anytype is not just incredibly fast and reliable software - it is a piece of art.
Having been early backers since 2019 and living through the pandemic and geopolitical turmoil alongside the Berlin based founding team we are deeply grateful and proud of them to announce their $13.4M Series A round led by Balderton with participation from Inflection.xyz, Square One, Script Capital, Protocol Labs, Connect Ventures, New Forge, and Foreword VC and prominent angels from Heroku, Muse, Centrifuge, Polkadot, Ocean and others.
From Caves to Screens: A Brief History of Knowledge Management Tools
The evolution of knowledge management tools is a captivating journey that mirrors the progress of human civilization itself. From cave paintings to the vast troves of papyrus scrolls housed in the ancient Library of Alexandria all the way to the boundless digital databases of the modern world, our tools for managing knowledge have continuously evolved. Beyond information storage they shaped how we access, retrieve, share and perceive knowledge across time and space. While the monumental shift from physical to digital mediums is in full swing, significant challenges need to be overcome: graph contruction, information provenance and user sovereignty are some of them.
Solving AI's context problem through knowledge graph construction
Against this backdrop emerges the concept of knowledge graphs. Knowledge graphs can be visualized as expansive networks where nodes of diverse information interlink, akin to cities on a map connected by an intricate web of roads.
Each node in a knowledge graph represents a piece of information, while the connecting lines, or edges, denote the relationships between these pieces. Graph structures can be compared to the human brain’s synapses (edges) and neurons (nodes). As opposed to relational databases which refer tables to tables, graphs consider the links between different data points. This is an incredibly powerful concept as the graph is both a place to organise and store data, and to reason what it is about and to derive new insights. It is also flexible and extensive in terms of the types of data and schemas it can support. Graphs evolve to reflect changes in the domain and new data is added to the graph as it becomes available. They are a powerful structure to support advances in GNN (Graph Neural Networks) you can explore further in Introduction to Graph Machine Learning, Must-read GNN papers or Graph ML in 2023: The State of Affairs.
Anytype approaches graph construction bottom up: users are at the center of organising their bookmarks, thoughts and content of all types (images, videos, music) - you can think of it as a second brain. Next stops will be access control, collaboration- and publishing features.
Open Source Signing Of Everything
The second profound challenge Anytype addresses is that of content provenance and user authentication - particularly in the age of generative AI which which allows for the creation of deep fake content distributed through fake social media accounts and bots. Crypto systems have been leveraged to prove ownership ("not your keys not your coins") and solvency of exchanges since 2015. Now, innovators are leveraging them as a tool to verify human identities, information provenance as well as compute in order to let humans and machines interact more reliably over the internet. By embedding signature schemes into every interaction seamlessly (soon to come) Anytype creates a new type of trustful collaboration environment for individuals, enterprises and communities of all sorts.
In order to be trusted such a crypto system needs to be open sourced - and that is exactly what Anytype just did with its open beta version.
Rather than relying on conventional, proprietary systems, Anytype offers a departure from the norm. Users are not bound by typical software limitations, nor are they required to consistently be online to access or share their data. They are not subject to the whims of any developers, including the ones working with the Anytype Association. The platform's Anysync protocol ensures that data synchronization occurs seamlessly across a peer-to-peer network, all while maintaining stringent encryption standards. Each user's space is managed locally, ensuring that they remain the sole custodian of their data. This emphasis on individual control, combined with the platform's collaborative ethos gives us a glimpse into how global scale collaboration between humans and (AI) algorithms over the internet might look like in the future.
Onwards & Upwards