Convergence of Artificial Intelligence and Web3
Current state of play, future use cases and possible challenges
Introduction
Over the past few months, we have seen how ChatGPT has taken the world by storm. It is estimated that ChatGPT reached 100 million monthly active users in Jan 2023, just two months after launch, making it the fastest-growing consumer application in history.
The popularity of ChatGPT has reignited conversations on the potential of artificial intelligence (AI). According to ChatGPT itself, AI is defined as the development of computer systems that can perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making and language translation.
The attention around AI has piqued investors’ interest in the space and they have placed big bets on AI-related projects. AI startups received over $50 billion in funding across around 3,400 venture capital funding deals in 2022. The sub-field of Generative AI - the technology that ChatGPT is based on - experienced a more than 70% increase in funding in 2022, compared to the previous year.
Here at Ocular, we believe in the power of technology and are excited to see how AI continues to shape our world. The full capabilities of AI have yet to be uncovered and web3 is a potential area that AI can play a big role in. In this article, we will highlight the significance of AI in web3 and how the two concepts have converged to transform many fields of work. Nonetheless, this only marks the start of the journey. In the future, we foresee many other use cases in which AI can be integrated with web3, and this will create new opportunities for businesses and consumers alike.
Current State of Play
AI Projects Leveraging Web3 Capabilities
We have observed the rise of AI-focused projects adopting web3 technologies to safely and securely store, trade and share information. These technologies can also be used to encourage buy-in and participation from users.
An example of this is DeepBrain Chain, which is an open-source, decentralized platform for AI computing. The platform aims to provide a cost-efficient and private way for companies and organizations to access the computing power they need to train and run their AI models. The network uses blockchain technology to create a decentralized marketplace for computing resources, where users can buy and sell computing power with the use of their token, $DBC.
Other decentralized AI marketplaces include SingularityNET - for anyone to create, share, and monetize AI services at scale; and Fetch.ai - for anyone to share or trade data to enable the creation of intelligent autonomous agents, which can interact with other agents and help perform tasks such as data analysis, prediction, and decision-making.
In finance, there is Numerai, which is an AI-based hedge fund built by a network of data scientists. The project uses its token, $NMR, to crowdsource new datasets and machine learning models that can help interpret and predict the stock market. Numerai then combines the best prediction models to form its in-house Stake-Weighted Meta Model and uses the Meta Model to construct and inform its portfolio.
Web3 Projects Incorporating AI Techniques
The converse is true as well - web3 projects are tapping on the power of AI techniques to optimize and improve existing processes. This is particularly prevalent in gaming and NFTs.
In gaming, Delysium is one of the first web3 open-world, player-owned games that use AI-generated content and characters as the foundation of its metaverse. In Delysium, AI-powered MetaBeing characters have a complete neural brain system that drives their lives in the virtual world. These MetaBeings are also able to hold assets and automatically participate in game modes to earn cryptocurrencies, as human players do.
In NFTs, Alethea AI proposes the concept of iNFT, which is an NFT that incorporates AI to give it unique personality characteristics and the ability to interact with users. Optic is also building an NFT verification protocol based on AI, focusing on NFT fraud analysis and NFT value discovery within the community. The Optic intelligent engine retrieves the collection of NFTs in the market by learning the real NFT series. Optic then returns a match score indicating how well the checked NFT matches the real NFT. This is aimed at helping the NFT market achieve higher authenticity and transparency.
Beyond gaming and NFTs, we are also seeing other infrastructure projects adopt AI. An example of this is Push Protocol, a web3 communications network that enables cross-chain notifications and messaging for dApps, wallets and services. It intends to integrate a ChatGPT-equivalent chatbot, to make messaging on the blockchain more user-friendly and accessible to a wider audience.
Future Use Cases
While the current use cases of AI and web3 are exciting, we believe that we are only just getting started in exploring the possible synergies between the two fields. We envision that AI could potentially be used to solve many web3-related challenges (or vice versa), and below are some potentially interesting areas:
Finance
AI can help with dynamic investment strategies, including the real-time optimization of on-chain trades. By reviewing past investment decisions, analyzing macroeconomic conditions and predicting price movements based on market sentiments, AI can provide investors with valuable insights that can help them make more informed decisions and achieve better outcomes in their investments. AI can also improve liquidation procedures on DeFi to provide better protection for debt positions.
In addition, AI could also assist with the efficient pricing of illiquid assets, such as NFTs. By making sense of past transaction data and predicting future demand for these assets, AI can help investors to better understand the value of these assets and construct their portfolios more efficiently.
Through analyzing on-chain activities, AI could also facilitate on-chain credit scoring to improve capital efficiency in DeFi lending.
Marketing & Recommendations
Web3 data comes from a variety of sources, and it is often unstructured and noisy. AI can be used to preprocess and analyse these on-chain data, to generate more detailed and complete profiles of web3 users. This can be combined with existing web2 data on the user’s behaviours and preferences, allowing marketers to better understand their audience. Armed with this information, marketers will be able to provide more personalized recommendations to users, e.g. an upcoming web3 game that is in line with a user’s interest or a DeFi project that matches the user’s risk-reward appetite.
Smart Contract Auditing & Tax
One of the biggest productivity gains from AI that we are already seeing (and will continue to see) is the shortening of time needed for programming. The process of writing and reviewing code will become more efficient with AI’s assistance. Smart contract auditing is a natural extension of this capability. By looking through past audits that were conducted, AI can be trained to assist with the auditing of smart contracts. This will reduce the time taken and manpower needed for smart contract audits (there is often a backlog of contracts to be audited due to capacity constraints), and may even be more accurate than existing methods.
AI can also help with populating and auditing tax filings based on the user’s on-chain transactions. This will minimize the risk of errors and streamline the entire process.
Tokenomics & Airdrop Optimization
Tokenomics is the key element of determining a project’s incentive scheme, both internally and externally. Once set, it is difficult to adjust later on after token generation. For games, the in-game tokenomics is akin to monetary policy, which has to be carefully designed and managed. Currently, a lot of web3 startups’ tokenomics appear to be designed arbitrarily. Allocation and vesting schedules are often based on prior examples, which may not have stood the test of time and have subsequently turned out to be flawed. AI can potentially help simulate token dynamics during the design and planning stage, and optimize for various scenarios and interest alignment across teams, investors, users and the wider ecosystem.
Airdrops are an important one-off event to incentivize the right users and grow the community. It is a marketing expense and is often dilutive to the existing token holders. By leveraging AI, teams can better allocate airdrop to the true supporters of the project that will continue to help in its growth.
Gameplay Simulation
As many reputable web2 gaming studios are now entering the web3 space, we expect more advanced and complex web3 games, combining web2 strategies and web3 digital ownership. With such complexity, it will be greatly beneficial to have AI-powered gameplay simulation, to achieve an optimal balance between fun and monetary incentives.
Security
AI can be used for fraud detection, e.g. identifying malicious entities on-chain by tracking their geographical and behavioural data. It can also be trained to bypass aliases and draw links between multiple wallets/accounts.
The other aspect of security that AI can help with is content moderation and monitoring. AI can review and flag false/abusive content on web3 platforms.
Authentication
AI’s ability to generate high-quality content is growing exponentially - from text to images to audio to video. It is already quite difficult to tell whether something is generated by AI or humans.
The famous deepfake video of Morgan Freeman generated by AI still confuses a lot of people:
This is a concerning trend given how much Internet content could sway people’s perceptions and views. Web3 could be the potential solution to solve this issue - by implementing digital on-chain signatures or using zero-knowledge proofs to authenticate the content. Similar to users using their private keys to sign off on transactions, creators can use their private keys as well to authenticate the content created. Several on-chain projects have already enabled linking cryptographic public addresses to signal the authenticity of content, e.g. Lens Protocol and Mirror.
Below is a summary of the current and possible future use cases for AI and web3, and how they can address both web2 and web3 issues:
Possible Challenges
The road to a full convergence of web3 and AI is not easy and does come with its fair share of challenges.
Technical Difficulties
The first is technical difficulties, as it is not always easy to integrate the two concepts. At this juncture, we have yet to see large-scale AI models being built onto the blockchain, as issues of interoperability, data privacy, and security must be addressed and resolved before a successful integration can be achieved.
In addition, existing AI models might require massive amounts of parameters, which could make it difficult to run on mobile devices or laptops that are mainly used for recreational purposes. To overcome these challenges, researchers are developing new algorithms and models that are more efficient and require fewer parameters. These new models could help with the seamless integration and operation of AI technologies on the blockchain, thus enhancing the user experience and expanding the reach of these technologies to a wider audience.
Regulatory Landscape
Italy previously imposed a temporary ban on ChatGPT, given concerns raised by data protection regulators. The ban was later lifted after the company made changes to comply with age verification requirements. However, other countries might follow suit and introduce similar restrictions.
Elon Musk and Steve Wozniak also recently signed an open letter with more than 2,600 industry leaders and researchers to halt further AI development for 6 months.
Given that this is a relatively new space, regulatory frameworks are not yet fully in place. As a result, it is difficult to predict how regulators will view the use of AI technologies in the future. This uncertainty can potentially affect the pace of developments in the AI <> web3 space.
AI-related Concerns
One of the key issues surrounding AI-generated content is the uncertainty and ambiguity around ownership and copyright structure. This includes the challenge of determining who is allowed to profit from such content. There are many factors to consider when it comes to ownership and copyright of AI-generated content. For example, who created the AI technology that generated the content? Was it an individual or a company? If it was a company, who owns the rights to the technology? Additionally, who provided the data that the AI technology used to generate the content? Was it a single individual or a group of individuals? There is currently no consensus to address these difficult questions, and this may deter projects from adopting AI-related content due to the risks it may pose to their business model.
Even if AI-generated content is adopted, there is a separate issue of bias in the AI models that analyse and process data that may affect the accuracy and fairness of the end results. AI algorithms are not immune to reflecting human biases. In fact, they can even amplify and reinforce them. This is because AI algorithms are trained on historical data that may contain inherent biases, such as gender or ethnicity. As a result, AI algorithms may create or reinforce gender biases, such as giving better credit scores to men than women, or discriminating against certain ethnicities. While there are ways to mitigate this issue, such as training AI algorithms on a diverse and representative dataset and using techniques including adversarial training or counterfactual analysis to detect and reduce biases in AI algorithms, it is important to recognise that AI models are not completely impartial and cannot be fully relied upon yet.
Conclusion
The convergence of AI and web3 is a rapidly evolving field with vast potential. At this point, we are just scratching the surface of what's possible. As technology continues to improve and regulations become better defined, the potential use cases for AI and web3 will only continue to grow. From the way we conduct business to the way we live our everyday lives, the impact of this technology could be significant.
At Ocular, we are committed to staying at the forefront of these developments and keeping a close eye on this space. If you're keen on delving deeper into these concepts or are currently working on a project related to these areas, we'd be thrilled to hear from you. Feel free to reach out and let's discuss the exciting possibilities of AI and web3!