By Constantin Kichinsky in The Future Of... —

🔮 Knowledge Collaboration Tools

🔮 Knowledge Collaboration Tools

TL; TR;

The knowledge management tools reached a culmination point in their lifetime. Their creators might use approaches developed for modern collaborative tools and information management and analysis tools to inspire new life into them.

Collaboration & BI tools are on their path to finishing the mainstream digitalization process in 2-3 years, and their creators will look for new transformational ways to evolve such products.

These two streams will merge into Knowledge Collaboration Machines (γσ-machines).

γσ-machine (pronounced as "gnosio-machine", but also as "YO-machines") is a knowledge app that unpacks a specific knowledge domain or scheme into analytical and exploring experiences. There could be different runtimes to run such apps (centralized or decentralized, open or closed, targeting personal or collective usage). I believe the app format and data must be standardized.

The end of the Knowledge Management Systems Era

There was no significant development in KM systems for the common (e.g., corporate) domain since the invention of wikis in the second half of the 90s. Wiki systems introduced collaborative editing, tags and tag clouds, hierarchical trees, taxonomies, and other features.

  1. Ontology editing software is practically dead or frozen in its development since the early 2010s when the semantic web paradigm died. Mind mapping software is stuck in development since the late 2010s. Diagramming tooling like Microsoft Visio is stagnating as well.
  2. Visual representation of knowledge is practically stuck in graph and network visualizations, and BI systems are focused on structured data rather than flexible knowledge.
  3. The latest notable reshake of the status quo was the launch of Notion in 2018 and Coda.io in 2019 and their continuous growth among startups and project managers. It is a fresh take on visual and textual emoji-friendly knowledge management but becomes complicated at large-scale libraries.
  4. Domain-specific computational notebooks like Wolfram notebooks for math-related research, Jupyter notebooks for Python and AI/ML-related "live" documents, and Benchlinq notebooks for biotech are unnoticed in the corporate sphere. But RPAs might bring the light.
  5. Another fresh line in knowledge management is the Zettelkasten-method driven by tools like Roam Research, Obsidian, Foam, Hypernotes, and many others. Some of these tools are positioned as personal KM systems, and some are tailored for collaboration in teams.
  6. None of the "challengers" got into the corporate or educational mainstream, but all three types give us insights into the future of the KM tools.

The midpoint of the Collaboration Tools Era

Most modern collaboration tools launched in the post-AJAX/post-Flash/HTML5 era when new APIs got browser support and the cloud paradigm had already begun its march.

  1. Over time they evolved to absorb teamwork/collaboration, enterprise ID, real-time chatting and calls, plugins/extensions, and other popular features.
  2. For the companies behind those apps, to continue the expansion of the user base means going enterprise with extra focus on security, sales & marketing (including growing field teams), and other integration features that will drain focus from the core product. Companies like Microsoft have an enterprise sales force, but lack product vision (Teams is a rare exception).
  3. None of these tools is pure digital in their nature (meaning they all implement some analog process in a digital form); some "just" shifted the medium from previous desktop runtime to a browser one. Their destiny is to stick with their root concepts (e.g., a sticker-based app).
  4. By the early 20s, such apps finished their mainstream development (the core feature set is almost complete) and now focus on building developer platforms and communities to expand integrations and usage scenarios.
  5. All collaboration tools look for cross-integration fulfilling the gaps of each other meanwhile working on independency towards a full-cycle software to gain the maximum of the user wallet, tired of maintaining separately charged subscriptions.
  6. All the statements above mean that the core value propositions of these tools are fixed by now, and there are no strong reasons to forecast their evolution in the coming years. Accepting it should disappoint the teams building these apps, but it doesn't mean there is no growth or incremental development. To get the idea of what I'm saying here – look at the MediaWiki release notes over the last ten years.

The γσ-machine

γσ-machine (pronounced as "gnosio-machine") is a knowledge app that unpacks a specific knowledge domain or scheme into analytical and exploring experiences. As of now, it is a future-looking statement requiring more exploration, research, and design, but few things we already can predict:

  1. The core element of the γσ-machine is a domain scheme that describes how to markup knowledge items. Such schema defines attributes, tags, and other properties that should be attached to knowledge chunks and probably allows for automating such tasks by training a neural network for knowledge extraction from external sources.
  2. Another element is one or more visual representations (layouts) for that domain serving as a contour map than can be fulfilled and colored based on a specific set of data, marked up using the corresponding scheme. Such representations serve dual purposes: data representation and data generation. One might use an empty domain map in a brainstorming session to fulfill it with some ideas (solo or collectively). The γσ-machine should be able to convert a populated map into initial knowledge chunks with some row data (e.g., headers) using templates.
  3. The labeled knowledge can be linked, augmented, and edited in a row format with algorithmic assistance. By default, by "row" I mean linear text. It is important to note that linear formats are becoming interchangeable as text is becoming a useful and must-have interface for editing audio and video. Services like Descript use audio recognition to generate a textual representation of the user file and use smart and AI-based technics to update media content based on text edits. In the future, it should be cheap enough to generate audio and video from scratch using your voice sample and avatar.
  4. Thus we can create a cycle between analytical and explorational modes of knowledge creation. And the γσ-machine defines this cycle and interfaces for external tools (runtimes). Runtimes, in their turn, allow multiple γσ-machines to coexist over overlapping knowledge chunks. So a knowledge chunk might combine multiple domain schemes and be presented in various ways.
  5. It is important to note that a γσ-machine could be compact and purpose-driven, representing a subdomain of some large knowledge domain. One might design such a package with some ontology in mind, and another will start from a specific visual representation scheme. Probably we should end with an open-ended system where γσ-machines can be infinitely extended, refactored, redesigned, reassembled, etc. I imagine it could be an analog for open source code or container repositories, but focused on knowledge.

The Future of Knowledge Collaboration Tools

Now let's briefly discuss the overall future of knowledge collaboration tools and some specific features:

  1. As I mentioned above they will adapt to any input method preferable by the author: voice or text, manual or automated. As linear textual representation is the default editable form it might be automatically converted into an audio or animated video in any language and vice versa.
  2. The γσ-machine serves as a bridge between analytical and exploration modes and as an M2M (machine-to-machine) interface for "collaboration" with external recommendation systems to augment the knowledge.
  3. Both textual and visual representations might be designed for collaboration from the beginning using best practices from such tools as Dropbox Paper and Miro. But the design language, format, and interface to define the γσ-machine itself are still an open question. Templates in the Miroverse are a huge step forward in bringing the community into play, but lack domain knowledge.
  4. We should note that the design language for collaboration between a human and a machine is also a great area for research. We might take first insights from the Github Copilot. Copilot conveys domain knowledge of programming languages and serves as a proxy for external expertise preserved in the open-source code. One might think of it as a manifestation of the noosphere idea.
  5. The biggest challenge of all the knowledge systems was always to create more business impact than just serving as a corporate or personal archive. So such systems should be redesigned for exploration, insights, highlighting unexpected links, etc. In some sense, they should bring the same value on top of knowledge, as BI systems do on top of data.
  6. I think this new generation of KMs will play a big story in the Antiverse (a dual mirror twin of the Metaverse), but it is a story for another article. Stay tuned.