Web3 and AI have been celebrated as the herald angels of the new digital revolution. Both have promised to change the way we interact with the internet and offer huge potential financial gains to those lucky enough to ride the wave to the top.

In fact, by 2025, the market size of Web3 and AI is set to grow to $39.7 billion and $126 billion, respectively.  

What is AI?

AI is the new kid on the block to some people and seems to offer unlimited opportunities in almost every industry. We all know the basics—smart robots that can make decisions and do things themselves. But what does the term AI encompass, and what role can it play in Web3?

Artificial Intelligence (AI) is the simulation of human intelligence using machines programmed to think and act like humans. It can apply to any computer or tool that can mimic learning or thinking. It can be programmed to make independent decisions or provide information for decision-making.

How smart is the AI we see today?

Pretty much all AI that exists today is what's known as Artificial Narrow (or weak) Intelligence. While it can perform tasks with varying degrees of human-like standards, it can still only do what it's programmed to.  

From the first AI program, a checkers program made in 1951, to the complex tools like ChatGPT making headlines today, none of these programs can “think” like a human. They can only work within the parameters we give them. So thankfully, we’re still a while away from the first HAL 9000.  

The next step in the evolution of AI will be Artificial General Intelligence—AI that can understand and learn to do any task that a human can. And one day, we might even see Artificial Superintelligence—when AI capabilities far surpass that of even our brightest minds. For now, though, let’s focus on what our current AI is composed of and what it can offer.  

Subcategories of AI

There are many categories within AI, and we can combine them in different ways to build the AI tools we see today. The six main categories of AI are as follows:

Machine Learning (ML)

This uses data and algorithms to enable machines to continue learning. It is designed to imitate human learning by improving accuracy over time.  

For example, Netflix’s recommendations become more accurate as you watch. (For those of us who binge-watch, this can be a quick process!)

Neural Networks  

A subset of ML, this models the human brain with interconnected nodes to handle more complex data.

The facial recognition that unlocks your phone uses this to improve accuracy and recognize your face from different angles and expressions.

Deep Learning

Deep learning uses larger neural networks to handle complex data sets. The word “deep” refers to the added depth of layers in the neural network. It involves less human intervention than other neural networks and has been termed “scalable machine learning” by some.  

Self-driving cars, like Tesla, use this to recognize more and more objects. It’s how they tell a stop sign apart from a yield sign.

Cognitive Computing

This combines many elements of AI, but rather than focusing on independent decision-making, it aims to augment our intelligence by providing insights and deep analysis of data.  

Cognitive computing is often used to help determine the risk levels of investments. It can analyze huge amounts of data to give insights into the investment, but humans make the final decision.

Natural Language Processing (NLP)

This covers training programs to understand how we speak naturally. It combines AI, linguistics, and social sciences to give computers the ability to understand text and voice, similar to how we do.  

Siri, Alexa, and whoever else might be tuned in to your home use NLP to interpret our different requests.

Computer Vision

This is the process of training programs using machine learning to interpret meaningful data from videos and images.  

Google Lens uses this to recognize objects or find clothing online from a picture.  

These can be used to handle a range of tasks, from basic automation to complex problem-solving and communication. As you may have noticed, many AI tools combine one or all of these components.  

Use cases of AI in Web3

So how can AI be used in Web3?  

AI can be used to simplify Web3 for users and developers. With NLP, we can create tools that write smart contracts for users who don’t code. Or create chatbots that can help you use decentralized exchanges and protocols. AI can be a powerful tool in demystifying Web3 and helping bring about mainstream adoption.  

It is also used for improving security and detecting fraud—flagging suspicious activity, improving consensus mechanisms, or analyzing behavior. Blockchain projects or blockchain-as-a-service companies can use AI to create customizable blockchains and develop layer protocols. Personal AIs are being used to add further utility to DAOs (Decentralized Autonomous Organizations) and NFTs. With a world of opportunity, let's look at some examples where web3 and AI have made the perfect match.  

Where Web3 is embracing AI

NFTs

The most obvious use case for AIs in the NFT (Non-Fungible Token) space is the generation of unique art through tools like MidJourney and Stable Diffusion. However, AI has much more to offer the world of digital assets.  

CharacterGPT is an AI NFT project that will give you the ability to turn text into original interactive characters. These characters will have an inbuilt personal AI that can give them, the project claims, a personality. Holders can then train the AI, customize the personality, and use it in dApps within the Alethea ecosystem. This has the potential to bring real utility to the NFT space.

Security  

With $3.9 billion lost in crypto through hacks and fraud, the web3 sector is scrambling for tools to protect its assets. By reducing the chance of human error and monitoring suspicious behavior, AI can help protect on-chain and off-chain activity. It can also analyze huge data sets and train models to recognize different activities, including fraudulent behavior, 51% attacks, and money laundering.  

Many security companies are developing AI-powered tools that can help reduce the number of hacks and fraud in the crypto ecosystem. Chainalysis currently uses AI to analyze huge amounts of transaction data to flag suspicious activity and identify patterns and trends in the crypto market.

AnChain.ai analyzes transactions, smart contracts, and risk profiles and has recently signed a deal with the SEC to help monitor and regulate the US DeFi (Decentralized Finance) industry. AnChain.AI uses a predictive model to identify unknown users and suspicious transactions as a “preventative” measure rather than only reacting to incidents.  

Metaverse

One of the main criticisms of the metaverse is the lack of a truly immersive experience. The graphics are often clunky or basic, and the technology used to interact—such as headsets or controllers—with the metaverse doesn’t provide the deeply intuitive experience of real life. This lack of that perfect balance between complexity and simplicity that makes up the human experience can hopefully be improved by AI.  

In fact, Meta (née Facebook), the metaverse's most fervent supporter, is banking on it.  They recently announced they are developing an “AI Supercomputer” called the AI Research SuperCluster (RSC), set to be the world's fastest AI computer. It will be used to build better AI models and training models in NLP, deep learning, and computer vision. Meta envisions RSC being able to provide real-time translations, augmented reality tools, content moderation at scale, and much more to create an immersive and safe metaverse experience.  

One tool they are developing is BuilderBot, a metaverse experience where users can generate objects and backgrounds by simply using natural voice commands. Although it is at an early stage—with the (embarrassingly) basic graphics we’ve come to expect from Meta—the AI tools behind it show impressive capabilities and potential. Meta is playing the long game; time will see if their gamble pays off.

Gaming

AI has the potential to add extra depth to gaming in general with the ability to create NPCs you can have an unscripted conversation with that can respond to actions and act intelligently. It can also provide “endless” landscapes that generate as you travel or unique settings that depend on your actions. Rather than having a limited scope for decision-making, with AI, we could see a more detailed “cause-and-effect” in the world of gaming. We may also see innovative monetization schemes that tailor experiences based on complex data mining and player behavior.  

In Web3, we’re seeing the advent of AI-powered games that include NFT characters, such as the previously mentioned CharacterGPT. Delysium is a play-to-earn MMO game with a generated open-world format, customizable characters, and “AI Metabeings.” These AI Metabeings (NPCs with an AI flair) are designed to interact with the game similarly to humans and can even earn tokens. Each is powered by its own unique AI tool to create distinct personalities.

Web3 and AI—the future of the internet?

Web3 and AI have met at the intersection of a technological revolution. Both have changed our day-to-day lives. Web3 has redefined how we consider finance, centralization, and the growing tech monopolies. AI has become a part of many of our daily lives (often without us even noticing), from facial ID unlocking our phones to self-driving cars, investment tools, Google Lens, Siri, Alexa, Maps, ChatGPT, Midjourney, and much more. Both Web3 and AI are powerful tools that can enhance and complement each other.  

Over the coming years, we’ll see how these technologies can unlock the true potential of the internet. Used properly, they can help create a better internet like that envisioned by the early pioneers of the world wide web—free information, open-source, decentralized, for the people. Used incorrectly, they may only widen the divide and inequality between people. One thing is for sure; it’s going to be a wild ride.

‍