Web 10 in Under 10 Minutes

Organizing information to help people and AI think better and improve the world.

Inspiration

Last week, I visited The Interval, the Long Now Foundation’s cafe/museum/HQ along the northern shores of San Francisco, for the first time. I had a wonderful four-hour conversation where I explained my personal theory of change from scratch, of which Web 10 is a significant component. The conversation reminded me that while I’ve invested a lot in detailed documentation for Web 10 (my shortened vision article takes 15+ minutes to read!), I don’t have an elevator pitch yet, or a short writeup that explains it. This article is my first attempt at that. Let me know what you think!

As a social entrepreneur, my initial motivation for creating Web 10 was to greatly improve humanity’s ability to solve problems, thus allowing us to effectively address global challenges and bring about a brighter future. Douglas Engelbart, one of the early pioneers of computing and the internet, describes his related concept of “Collective IQ” as leveraging technology to improve “collective perception, memory, insight, vision, planning, reasoning, foresight, and experience” in order to more effectively pursue any type of goal, in particular valuable goals like “seeking a cure for cancer” and “improving conditions in underserved communities.” Sadly, Engelbart’s vision was never achieved, and the internet (whether Web 2.0, web3, or other variants) remains far from that ideal.

Web 10 aims to make this, and more, a reality in the immediate future. Here are some one-sentence ways to describe Web 10:

1. Organizing information to help people and AI think better and improve the world.

2. Revolutionizing knowledge and intelligence—on an individual, team, and planetary scale.

3. Web 10 an advanced, societally beneficial vision for the future of the internet, software, and knowledge work.

Let me know if you like any in particular!

Diving in further, there are three stages that explain how Web 10 works: data, knowledge, and reasoning.

Stage 1: Data

Data is information without context and meaning. Every document you have, or page on the internet. Every file on your computer. Every book. Every image. Every data point: temperature, location, time.

Web 10’s can represent essentially all data in the world in a unified manner, from how data is stored to how it’s represented to the user. Web 10 data storage is closest to the concept of multi-model databases, which elegantly unify different types of data. In both systems, every single data point about the same thing can be associated with that thing (text, images, datapoints, relationships with other related things) instead of being stored separately in specialized databases. In Web 10, any user can define any type of thing they want to store data about as a data model. An example data model is a thing of type “person” must have the data point “age” which must be a number greater than or equal to 0. Web 10 also stores data about that data, like who the data can be accessed by, and when it was created. Web 10’s data system enables all data in the world to be efficiently unified.

What can be done with this data? Generally, people will store data using predefined data models, or map custom data models they create themselves with predefined ones. This enables users to take advantage of Web 10’s ecosystem of free and openly available prebuilt components that work with this data. Data models, automations, and on-screen elements are the three basic parts of most apps. When put together, they can instantly replicate nearly any app currently in existence. A similar concept is itemized operating systems, which allows users to easily create their own custom apps. Moving a step further than that, if users combine public data in the Web 10 ecosystem with components, users aren’t just limited to making apps to manage personal data like to-dos and budgets. Anyone can instantly make their own version of multi-user apps like LinkedIn or Yelp if they’d like for their own use or everyone’s use.

While apps are useful because they have specialized interfaces, such interfaces aren’t needed for many use cases. Users will interact with most data via a single unified interface that is compatible with many types of data. We’re already seeing this trend where users use nocode apps like Google Sheets and Notion to try to replace their own apps, although those systems have data models and interfaces that are insufficiently flexible for most use cases, so they only work some of the time.

Web 10 takes this trend to its fullest extent. Imagine an app/OS with all of the information you need, and all of the possibilities you’ve ever wanted. This could be as granular as being able to immediately add the one feature you’ve been missing to your favorite app, or much broader, like doing everything twice as quickly by no longer needing to switch between and copy data across hundreds of apps, files, and websites.

Stage 2: Knowledge

Next, let’s move to knowledge. Knowledge is data that has meaning and is true. I’m going to use meaning in a specific way; it must be unambiguous to the extent that no one can misunderstand it, even a computer. Let’s take the sentence “the child is one.” Ignoring AI, computers cannot understand words and sentences at all, so this fails the test. It’s not just computers; this sentence is not perfectly human understandable either. For example, which child? And what does “one” mean? Is the child one day old, one year old, one whole object, the number one person in the world, one with the force, or something else entirely?

In order to satisfy the meaning requirement, data must be represented in a precise specification known as an ontology. Going back to data models, an ontology is a data model that everyone agrees to agree on (e.g. we agree to define that people must have an age that starts at zero and increments by one each year following their date of birth—interestingly, this is not shared among all cultures). If ontologies are unambiguous enough, computers can “understand them” in the sense they can perform logical operations on them. For example, if you paste “my child was born on January 1st, 2023” and “it is now January 1st, 2024,” Excel would have no idea what to do. But if you have a numerical column called birthdate and a function that fetches the current date, the computer can now calculate that the child just turned one.

Let’s move to the truth requirement. Web 10 can create an internet filled with content that is true, valuable, and kind; in others words, the exact opposite of the internet today haha. One key system that enables this is reputation-weighted voting. This is similar to upvotes on websites like Reddit and StackExchange, except users can increase their own voting power by making reliable contributions and voting reliably on the value of others’ contributions. This makes the internet meritocratic, not just democratic.

Reputation-weighted votes determine the likelihood of information being true. The Consensus Knowledge specification has a similar design. Web 10 can be extended to more subjective assessments as well, like the quality of content in general or along different dimensions. A powerful feature on top of reputation-weighted voting is to be able to personalize reported votes, such that a user can see what content a particular group of users voted highly on, rather than the output of an arbitrary group of users (such as demographics that the entirety of Web 10 skews towards, which may not be a desirable basis for gauging truth or quality for particular needs).

Imagine an internet where you can instantly find nearly any knowledge or information you need in a fully personalized and automated fashion. For instance, you could find all of the recipes for a particular dish you are craving that (1) meet your sugar and sodium requirements, (2) work with the exact ingredients you have at home, (3) take less than 15 minutes to prepare, (4) utilize culinary techniques you are familiar with, and (5) have a median rating above 8/10 from people that have similar tastes to you. Imagine recommendations at that level of specificity for anything you’ve ever wanted. What food to eat, which places to go, who to meet. And imagine what could be possible if such a system could be used not only for personal goals, but also for societal goals, like identifying the easiest ways to help people live longer, be happier, and gain financial security.

Stage 3: Reasoning

Let’s cover the final stage, reasoning. Reasoning is a process conducted by humans, computers, or AIs that uses existing knowledge to arrive at new knowledge. In Web 10, the reasoning that leads to new knowledge must itself be modeled as knowledge inside Web 10 so that the reasoning is transparent and verifiable. Like the requirement for knowledge in Web 10, reasoning must be expressed in a manner that is so unambiguous everyone can understand it, even a computer.

Let’s use a classic example: (1) all humans are mortal, (2) Socrates is a human, and (3) therefore, Socrates is mortal. If properly structured, points 1 and 2 can be fed into a computer, and the computer can calculate point 3. This is the field of formal logic, which kicked off the development of automated reasoning, a key facet of early artificial intelligence.

Web 10’s objective is to (safely and responsibly) increase intelligence (of which a major component is reasoning) and apply it to dramatically improve the world. Modeling simple forms of reasoning, which is the extent of current systems, is not sufficient. Web 10’s reasoning model must capture essentially all forms of reasoning, including causal and temporal reasoning, to reach the intelligence necessary to solve major challenges. Reasoning in Web 10 incorporates probabilities and explanatory models in order to determine which approaches are most likely to achieve goals. Web 10’s reasoning system is similar to probabilistic logic networks, a framework that was originally developed to help create an AI that could reason about anything.

Stepping beyond reasoning, other aspects of intelligence are also expressed as knowledge in Web 10, forming the basis of a cognitive architecture. A cognitive architecture is composed of systems like perception, memory, and goals/values that come together to form intelligence. For instance, sensors can act as perception, Web 10’s data storage can act as memory, and goals/values can be aligned on by users. Web 10’s cognitive architecture is far vaster than the capabilities of any single human or AI system, and is designed to behave superintelligently, but it would not be considered intelligent itself. It simply orchestrates information transfer across the systems that compose a cognitive architecture, and the humans, AI, and computers that power the architecture’s cognition. For example, a computer could scan a study, several AI models could convert the information inside that web page as knowledge in Web 10, and a person (or a thousand people) could share their probability and associated rationale for how strongly the study impacts the anticipated outcome of a policy to reduce poverty.

It’s difficult to imagine the impact that superintelligence could have. However, everything humanity has achieved to date could be seen as the result of a global superintelligence, with all of the knowledge that humanity has amassed over thousands of years distributed across countless specialized individuals and organizations, all coordinated by the invisible hand of the market. Imagine a world where we can bring people together and align on a positive shared goal, like maximizing well-being, and leverage a superintelligent system like Web 10 to discover the best way to get there.