Blockchain-Powered AI is an Overrated Hype Train

Artificial intelligence (AI) has been around for several decades. However, it was not until the launch of ChatGPT that AI became more than just an investor’s checklist for emerging technology. As users around the world engaged with the revolutionary chatbot, the appetite for AI-related technologies exploded. On the back of a $10B dollar investment by Microsoft into OpenAI, stock markets entered into AI mania. Anything touching AI on the NASDAQ, including ETFs, followed suit.

Naturally, this was not just limited to Web2. Crypto is primarily a narrative-driven economy and we saw similar levels of mania once ChatGPT proved its value proposition.

In light of recent developments, we thought that this was a perfect time to dive a little deeper and understand the potential of these two technologies together.

Before we start sifting through the noise, however, let's first broadly define what each of them is. AI refers to a set of algorithms and technologies that enable machines to perform tasks that typically require human intelligence, such as pattern recognition, decision-making, and language understanding. On the other hand, blockchain is a decentralized, distributed ledger technology that enables secure and transparent record-keeping.

In theory, there are a variety of synergies between the two technologies. We cover three (3) of the proposed value propositions below:

Accessing reliable data: Combining blockchain and AI enables the creation of decentralized AI applications and algorithms that have access to a reliable and shared data platform for storing knowledge, records, and decisions.  ability to trace and verify the data and decisions that are input and output by an AI model

Improving analysing capabilities: Decentralized AI systems also allow for processor independence and avoid the downsides of sharing aggregate data. Users can process information independently across different computing devices, leading to diverse findings and fresh solutions to problems that a centralized system may not be able to solve, all while recording their results in an accessible and transparent manner.

Autonomousness: AI can be used to empower any decision-making process. With the accessibility of blockchain data, decentralized autonomous AI agents can leverage that data to make trading decisions, manage DAOs, and even develop their own smart contracts.

However, there are many challenges to making that a reality. We highlight three (3) of the core issues below:

1. The scalability issue: Scalability is the bane of cohesive integration. AI and blockchain require large amounts of computing power, and when they are combined, the computational requirements increase even further. This can result in slow processing times, which could be a significant roadblock for industries that require real-time data processing.

2. The data issue: Blockchain technology is not yet designed to handle large amounts of data that are required for AI training and inference. Current blockchain systems have limited storage capacity and slow transaction processing times. This can make it difficult to store and retrieve large amounts of AI data, which can impact the performance of the AI models.

3. The trust issue: One of the primary benefits of blockchain is its ability to establish trust between parties without the need for intermediaries. However, AI algorithms are often black boxes, meaning that the decision-making process is not transparent. This lack of transparency could make it difficult to establish trust in the data and insights generated by AI algorithms.

   
   

The question then becomes: if AI and blockchain are so prohibitive in nature; why are there so many Web3-AI-based projects which garnered significant attention? For one, due to the current trend of “AI is good”, projects often leverage that market narrative to drive attention toward their protocol(s), even if there is no meaningful AI integration.

But perhaps more importantly, the level of integration between AI and blockchain runs on a spectrum - the methodology of integration with the blockchain, and how it utilizes blockchain data is highly varied. Moreover, AI is a broad and interdisciplinary field that encompasses various sub-fields and domains which are depicted below:

Some projects might use these technologies but the level of use and their usefulness are usually questionable. From our observations, AI adoption within Web3 falls into one of three (3) categories:

Improving the development of blockchain technology: One of the most practical use cases for AI is assisting developers with coding. ChatGPT is the best example of this and has proven itself in programming simple smart contracts and other Web3-related functions (e.g. programming trading bots). Smart contract auditors like Certik have even showcased how it can be a supplementary tool for auditing code.

Interpreting blockchain or blockchain-related data to generate insights: AI and blockchain are being used in various ways in the healthcare industry to improve patient outcomes, protect patient data privacy, and streamline administrative processes. One example is Nebula Genomics, which offers a platform for patients to securely store and share their genomic data with researchers. The platform uses blockchain to ensure data privacy and security, while AI algorithms are used to analyze the data and provide insights into genetic risks for diseases. More recently, an up-and-coming project, KaitoAI, is plugging into ChatGPT in order to index crypto-related data across different platforms like social media platforms (e.g. Discord and Twitter) and even on-chain data (e.g. data dashboards).

Improving or creating new experiences for blockchain users: On-chain products are getting creative by either incorporating AI elements or building an entire experience around AI. Altered State Machine, for example, uses AI to influence the attributes of their in-game characters through decentralized nodes that run machine learning algorithms. Another example is Botto, a social experiment where an on-chain community chooses which AI-generated images are minted, sold, and burnt as NFTs; ultimately creating a ‘final NFT masterpiece’.

The reality is that most AI-blockchain-related products today focus on extracting data from Web3 onto existing AI engines built outside of the network. Truly integrating AI and blockchain together, where transparent data seamlessly flows between both mediums, requires a considerable amount of computational capability, high-speed network connectivity, and sufficient storage capacity.

While AI is becoming more accessible, there are still many limitations that limit the integration and production of truly decentralized AIs. Even blockchains which prioritize scalability are unlikely to be able to efficiently train AI models. For example, most nodes run on GPUs or CPUs but there are AI algorithms that may require Tensor Processing Units (TPU), which are designed for high-volume, low-precision computation. We would likely need to build specially-designed blockchain infrastructure that caters to AI, none of which is readily available yet.

At this juncture, AI should primarily be treated like any other piece of supportive technology, whether it be HR software or the type of development engine used. Marketing the use of AI because it is used in some obscure context is vastly different from when it provides a meaningful impact on the product or service built. One thing that is clear, however, is that we are only seeing the tip of the iceberg. It is only a matter of time before the true potential of AI and Web3 materializes.