What the Heck is Web3? A Primer
Crypto protocols, open data and the semantic web, open-source AI, 3d graphics and the metaverse, and open hardware with BMIs.
TL;DR
Web3 is the next stage of the internet, mixing decentralized crypto protocols, open datasets and semantic representations, open-source AI, 3d graphics and presence via metaverse worlds, and a ubiquitous mesh of hardware, including brain-machine interfaces. It’s already started and is going to be wild because it is a giant leap forward from the Web2 world we mostly live in - the only comparisons I can think of are going from 2d to 3d, or from a world with little or no electricity to one where everything around you is powered by it.
A lot of people have been associating Web3 just with crypto projects. While that is a key part of Web3, there’s a lot more. I expect Web3 will be much weirder, more unpredictable, and more consequential than current projections of it, so the list below is partial and comes from looking through much uncertainty.
Historical context on Web1 and Web2
Pre-web internet (1969 to 1990): from ARPAnet to the CSNET, this was the early period when the military and academic communities built many of the early protocols that made the internet a niche tool. It was limited to employees of a few large institutions and no commercial or frivolous activity was allowed.
Web 1.0 (1991 to 2002): With desktop ownership growing at home, it took the release of the HTTP protocol in 1991 and Mosaic as the first web browser in 1993, along with the commercial opening, to really launch the first web wave. Early Linux in the 1990s became the open-source operating system, and email, instant messaging, search, and VOIP (video/voice) led to the boom and the first tech craze. Note that most of this was done by startups or newer players/ISPs like Prodigy and AOL.
Web1 was mostly about translating old-school media in a skeuomorphic way to web browsers. Magazines and business marketing materials were translated to web pages, while music and pictures were illegally shared on distributed file-sharing services like Napster. There were few individual content creators, though many people had personal web pages, often on places like Tripod or Geocities. Web pages were static and communication was via HTML forms to send emails. Web1 lived on ISPs and physical servers that were a pain to set up and maintain.
Web 2.0 (2003 to 2015-ish): After the global tech crash, the world saw the rise of web apps (Facebook, Yelp, etc) to mobile apps with the iPhone and Android launches (Uber, Airbnb, etc). There was a massive shift from desktop to mobile and IoT devices, and lots of free, advertising-driven apps (Gmail, Facebook, search, etc). Linux and the Apache Foundation projects became the open-source powerhouses while AWS built cloud primitives to further re-centralize the web in cloud servers and empower large companies and new startups. There was another battle between closed versus open systems and the closed ones mostly won (the big tech consumer and enterprise apps). Also, national governments started to hive off the internet (esp. China and Europe) with the Great Firewall and the GDPR superstructure.
Web2 meant software apps were built on the web, and this enabled more interactive user-generated content (UGC) with sites like Wikipedia, Yelp, and Reddit. The other characteristics of Web 2.0 were a rich user experience, user participation, dynamic content, metadata, Web standards, scalability, APIs for devs to access, and Software as a service (SaaS). Finally, much of Web2 was built on AWS, and then later on Azure and GCP.
The sneaky, creeping nature of Web3
Web 3.0 (2016-ish to today): In hindsight, Bitcoin presaged the third wave, but Ethereum, along with open data and recently open AI, were the really magical technology that kicked it off. This continued all of the above waves, but added crypto and pervasive AI/ML systems (after AlexNet in 2012, Google replaced all their older systems with ML systems over 5 years). The goal was to personalize everything. Web3 is the next stage of the internet, mixing decentralized crypto protocols, open datasets and semantic representations, open-source AI, 3d graphics and presence via metaverse worlds, and a ubiquitous mesh of hardware, including brain-machine interfaces.
We are deep in this crypto, AI/ML, semantic, 3d/metaverse, ubiquitous hardware world today, but most people don’t realize it (not even most engineers, who haven’t upskilled to ML frameworks or the new crypto languages, and hence there are big dev shortages). As Web3 progresses, I expect massive crypto communities and DAOs will overtake AWS and power the core of the global internet.
Let’s go over all the elements of Web3.
Crypto protocols
As I wrote in a previous post on “Why Crypto Matters”, it all started with Bitcoin, with a $1TRN market cap in 2021, was the most successful tech project of the 2010s; it put crypto on the map, but isn’t even the most important coin or project. Bitcoin is an older, reserve currency alternative that competes with gold, USD/EUR/JPY reserves.
Internet protocols have been evolving for ~50 years now and we’re seeing a surge of new ones in the crypto space that will have a massive technological, economic, social, and political impact. You have to understand the history of the web to grok this.
The crypto portion of Web3 likely started with the launch of Ethereum in 2015.
One of the key questions in how computing and Web3 affect society, politics, economics, etc: Who has root access and control? Public blockchains are massively multiclient databases, where every user is a root user. You need the decentralized database aspect of blockchains, along with coins/tokens as a form of resource ownership to fund the projects and let users gain access to the tech, and a robust, global community of devs, users, and apps. Crypto is all three: blockchains with a governance/incentive structure of tokens/coins and an active community.
The crypto economy is almost like a new alternative economy that’s being built, where people are not just trading bitcoin. They’re earning a living. They’re launching new start-ups — crypto start-ups. They’re borrowing and lending. They’re doing all different types of economic activity, buying products and goods and services.
After Bitcoin, Ethereum is the most important project of Web3 - it’s a decentralized, global universal computer and is a lot more than just smart contracts. There are many new and important protocols and apps coming out beyond these two, despite the 95%+ of fraudulent, hyped, or incompetent projects (like Dogecoin). Ethereum may ultimately get out-executed by projects like Solana or Cardano (or something else), but it shows the power of decentralized compute, consensus, mini-programs, an almost Turing-complete blockchain, and much more. The earliest use case was Decentralized Finance (DeFi), which is currently challenging the global financial system and regulators in a big way as a dominant financial protocol layer with many banks, insurance companies, securities, non-fungible tokens, and more built on it.
The crypto adherents to Web3 are pushing the following principles to varying degrees: open, decentralized, censorship-resistant, immutable, trustless, and permissionless. Zero-knowledge proofs and Merkle trees are crucial underlying tech. Initially, many of the protocols will have to create decentralized versions of web primitives like compute (Ethereum, Solana, Cardano), storage (IPFS, Arweave), identity, domain names, and all the services AWS offers.
The biggest problem with crypto in 2021 is that many of the protocols are inefficient, messy, and expensive, while the UX sucks. Setting up online wallets or hardware wallets is a pain, transactions are annoyingly technical and expensive ($50-100 gas fees), and there’s been too much focus on the protocol layer and not enough on easy-to-use and useful decentralized apps (dapps).
However, despite the curtailed descriptions and even hype, Web3 involves a lot more than just crypto.
Open-data and semantic representations
The second part of Web3 is open data on the semantic web, which improves web technologies in order to generate, share and connect content through search and analysis based on the ability to understand the meaning of words, pictures, and videos, rather than on keywords, numbers, or simpler metadata.
This part of Web3 means lots of free datasets on Google Dataset search (ImageNet, Fashion MNIST, etc), Kaggle, or via APIs like what Chainlink offers. Closed data that users created, but platforms owned, was key to Web2. Web3 allows creators and the public to have much more control over data and makes it harder for data to be suppressed because it can exist in non-cancellable, permanent places like IPFS or Arweave.
Web3 includes a wide range of semantic web tools, including open-source knowledge graphs like Wikidata/Freebase, KBPedia, Schema.org, etc and open-source internet-scale datasets like Common Crawl, WebText2, Books1/2, Wikipedia, & Reddit that large AI models train on. Finally, blockchain tools like Chainanalysis, Elliptic, Etherscan, and so on are key to helping humans understand the data in the crypto world.
Open-source AI for Web3
The third part of Web3 is open-source AI. While most of the cutting-edge, Web2 AI models and tools are used and controlled by large tech companies (esp. Google, Facebook, Microsoft/OpenAI, Apple, Amazon, Salesforce, Adobe, Uber, etc), a host of startups like HuggingFace and EleutherAI (which started on a whim) are democratizing these and making them open to anyone, for free.
This is the least discussed part of Web3: how much will a few, well-resourced corporate titans and unicorns control these models, versus how much will be released by universities with the help of a National Research Cloud and some scrappy upstarts trying to provide open models and tools?
Web3 will be hobbled if it’s just crypto protocols - it needs the open-data/semantic representations and the open-source AI to be really useful. Otherwise, Web2 will continue to dominate with rich, low-friction experiences that provide more intelligent applications to users.
The biggest under-funders of open AI are governments, crypto projects, and large research universities with big endowments. Instead of complaining about large tech companies sucking up talent, these institutions need to start spending money to develop and deploy open AI themselves (otherwise they will be far behind the Web2 winners).
3d graphics and presence via metaverse worlds
Glimmers of the metaverse have existed since massive, multiplayer online (MMO) worlds like Ultima Online in 1997, Second Life in 2002, or World of Warcraft in 2008. Today, large game worlds like Fornite, Final Fantasy, PUBG, Roblox games, and so on are one strain of the metaverse.
Ultimately, as the fourth part of Web3, widely adopted versions of the metaverse will need to offer presence with 3d graphics, full integration with AR/VR, interoperability across platforms, permanently owned digital goods, and a bunch of other properties that Matthew Ball and Tony Parisi have written about. I would also pay attention to what Zuck, Boz, and John Carmack are planning with the best-funded metaverse team in the world, Meta’s Reality Labs, as discussed in this long video that is as grand as Douglas Engelbart’s Mother of All Demos in 1968 or Steve Job’s demo of the iPhone in 2008. Tim Sweeney of Epic Games has a similar, detailed take on the metaverse, that he first discussed in SIGGRAPH 2019 and that he recently discussed with the NYT.
The intersection of the metaverse and crypto happens via Decentraland, and the intersection of the metaverse and AI will happen via startups like InworldAI or large language models like Blenderbot 2.0. Glimmers of the metaverse are already here, but a ubiquitous, set of interoperable platforms with 1bn+ users will likely take 8 to 10 years to arrive (closer to 2030).
A ubiquitous mesh of hardware, including brain-machine interfaces for Web3
Fifth and last, but not least, the hardware core of Web3 isn’t discussed nearly enough. ASICs were crucial to Bitcoin mining becoming a global phenomenon, and products link Antminer, Whatsminer, and Avalonminer were crucial. In a similar way, GPU compute growing faster than Moore’s Law was key to AI/ML in Web2. Central to Web3 hardware is RISC-V, an open-standard instruction set architecture (ISA) based on established reduced instruction set computer (RISC) principles. Unlike most other ISA designs which underlay every modern chip (from Intel’s x86 to Apple’s ARM M1 chips), the RISC-V ISA is provided under open source licenses that do not require fees to use. Numerous types of open hardware can be built on RISC-V, and this is crucial for Web3 to flourish.
Some important hardware projects for Web3 to succeed include:
Arduino’s open-source microcontroller board
Raspberry Pi - a series of small single-board computers
New hardware-enabling operating systems like Zephyr as an RTOS for connected resource-constrained devices, or Urbit as a clean-slate OS and network. As long as iOS and Android, and Windows to a lesser extent, dominate as the main operating systems for computers, Web3 will be stifled.
Hardware Wallets — such as Ledger and Trezor
The OpenCores intellectual property (IP) core enables connection and processing capabilities for sensors and other devices. Embedded sensors are created by using IP available on OpenCores to integrate into a field-programmable gate array (FPGA) to construct a System on Chip (SoC).
OpenCV enables the rapid development of complex image sensor-related processing.
Blue by BerkeleyOpenArms to test robotic systems
LibraryBox’s open-source, portable digital file distribution tool (a DIY anonymous offline file-sharing and communications system built with free software and inexpensive off-the-shelf hardware).
Defense Distributed for open-source guns and weapons.
The Global Village Construction Set’s (GVCS) modular, DIY, low-cost, high-performance platform that allows for the easy fabrication of the 50 different Industrial Machines
Just as open-source software projects like React, Angular, PyTorch, Tensorflow, and the many Apache projects (Hadoop, Hive, Spark, Beam, Cassandra, etc) were key to the flowering of Web2, Web3 will need more hardware.
The really wild part, where Web3 companies will battle with Web2, is over brain-machine interfaces like what Neuralink, Kernel, and Meta’s Reality Labs wrist interaction team are building. So far, OpenBCI and MindAffectBCI are the only projects I know offering that competition, and this area is severely overlooked and underfunded by Web3 builders and investors.
In sum, Web3 is a lot more than just crypto. Besides the topics above, I also think computational biology and ubiquitous robotics may be part of Web3, though it’s possible those fields lag and come later as Web4.
Web3 is the next stage of the grandest project of the 21st century, the internet, and it combines decentralized crypto protocols, open datasets and semantic representations, open-source AI, 3d graphics and presence via metaverse worlds, and a ubiquitous mesh of hardware, including brain-machine interfaces.
References
Tim Sweeney and the Metaverse (SIGGRAPH 2019)
Zuck and the Metaverse (FB Connect 2021)
Scaruffi on Web 1 to 3, the Metaverse and on Crypto History