Opportunities in DeAI marketplace

Why is web3 more accessible?

According to the statistics of the World Bank in 2021, credit card ownership seemed to align with a nation's development. Canada led with 82.7%, followed closely by developed countries like Israel, Iceland, and Japan. The USA stood at 66.7%. Many European nations reported over 50% ownership. In contrast, many African and South Asian countries, such as Nigeria and Pakistan, recorded less than 2%. The trend suggests that developed countries have a higher percentage of credit card holders compared to less economically advanced nations. As a result, people in generally less-developed countries have limited access to paid AI services due to a lack of payment methods. Furthermore, teenagers under the age of 15 typically have limited access to credit cards. Consequently, they have restricted access to AI services and support. If they wish to use these services, they often have to rely on their parents' cards, creating transactional friction and making AI less accessible to the general public.

Web3 and crypto, however, offer a much simpler payment process. Taking Ethereum as an example, since anyone can create an Ethereum wallet easily by setting up everything on their mobile with a few clicks, it's more user-friendly than traditional banking and centralized payment methods. And having an Ethereum wallet and ether means you have access to the web3 world, so almost everyone can access web3 if they buy any type of crypto. Moreover, people are starting to accept crypto as salaries since they don't have to use an international bank to receive it. Famous comments, such as one from Vitalik in 2020, mention workers in Africa accepting ETH as payment. A broader and simpler access through web3 could make AI services more available and exciting as an industry.

Web3 infrastructure as an advantage

Any public chain requires a consensus mechanism to update the global states in a distributed network system, with Proof of Work (PoW) being the most commonly used. The consensus mechanism is often referred to as crypto mining, which involves creating and adding new blocks to a blockchain network using various consensus methods based on different resources (like mining rigs, staked tokens, etc.). In the PoW consensus mechanism, miners compete to produce the next valid block by being the first to solve a cryptographic puzzle, thereby earning a reward for their efforts. In the marketplace, as a public chain gains popularity, its crypto miners receive increased rewards. This attracts more miners, or in other words, more computing power, to join the chain. Consequently, the total hash power of the chain continues to rise over time. Notably, prominent blockchain projects in the crypto industry, such as Bitcoin (BTC) and Ethereum (ETH), have used the PoW consensus mechanism for years. The annual electricity consumption of Bitcoin mining surpassed that of the United Arab Emirates in 2021 and Sweden in 2022. Most of this energy is dedicated to solving cryptographic puzzles.

While this process achieves trustless consensus, it doesn't offer practical benefits beyond producing block hashes that, in Bitcoin's case, have a certain number of zeros at the beginning. Consequently, the absence of a theoretical limit on energy consumption for the PoW mechanism has raised global concerns. This led to the exploration of alternative consensus mechanisms, like Proof of Stake (PoS), and changes in institutional policies. For example, in 2021, Tesla announced it would no longer accept BTC due to climate concerns. In 2022, Ethereum transitioned from the energy-intensive Proof of Work (PoW) mechanism to the more efficient Proof of Stake (PoS) in response to environmental concerns. This shift resulted in a significant reduction in energy demand, with decreases ranging from 99.84% to 99.9996%. This reduction in Ethereum's energy consumption is comparable to the electricity needs of countries like Ireland or even Austria, marking a notable step towards environmental sustainability. However, this change also left a large amount of unused hashrate without a specific use. The advancement of computing resources in crypto mining isn't the only type of resource in web3. In cloud storage, the capacity of decentralized storage has surged, as per reports and statistics. Furthermore, the cost of decentralized storage is on average much cheaper than centralized solutions such as Dropbox.

Meanwhile, as artificial intelligence (AI) becomes integrated into various sectors of the economy, there's a rapidly growing demand for computational resources to power this machine intelligence. Training a model like ChatGPT can cost over $5 million, and the initial operation of the ChatGPT demo ran OpenAI an approximate $100,000 daily, even before its current usage surged. Midjourney, a service that provides high-quality images, operates with more than 9,000 GPU cards, contributing to its operational costs. Given the vast number of neural parameters and extensive GPU hours involved, the high computational demands of model optimization pose significant challenges for academic researchers and small-scale enterprises. This limits the broader adoption and use of artificial intelligence technologies.

It is, therefore, unsurprising that an increasing number of crypto miners are exploring ways to use their existing computational infrastructures to advance AI. They are redirecting computational resources, which were previously focused on mining, toward machine learning and other high-performance computing (HPC) applications, such as the Internet of Things (IoT) and data services. Another example is provided by Hive Blockchain, which is shifting its long-term HPC strategy from Ethereum mining to applications like artificial intelligence, rendering, and video transcoding, contributing to their total annual revenue generation.