The Potential of AI Data Track and the Rise of Web3 DataFi

Recently, the most eye-catching event in the AI field is the powerful strength displayed by Meta. Zuckerberg is recruiting talent everywhere to form a luxurious AI team mainly composed of Chinese research talents. Among them, 28-year-old Alexander Wang has become the team leader. Wang previously founded Scale AI, which is valued at 29 billion USD, providing data services for several AI giants including the U.S. military, OpenAI, Anthropic, and Meta. The core business of Scale AI is to provide a large amount of accurate labeled data.

Scale AI stands out among numerous unicorns primarily because it recognized the importance of data in the AI industry early on. Computing power, models, and data are the three pillars of AI models. If we compare a large model to a person, then the model is the body, computing power is the food, and data is the knowledge and information.

In the rapid development of large language models, the industry's focus has shifted from models to computing power. Today, most models use transformers as their framework, occasionally incorporating innovations like MoE or MoRe. Major companies either build their own supercomputing clusters or sign long-term agreements with cloud service providers like AWS. After addressing the foundational issue of computing power, the importance of data has gradually become more prominent.

Unlike traditional B2B big data companies, Scale AI is dedicated to building a solid data foundation for AI models. Its business not only includes mining existing data but also involves long-term data generation efforts. The company has also assembled an AI training team composed of human experts from various fields to provide higher quality training data for AI models.

Model training is divided into two stages: pre-training and fine-tuning. The pre-training stage is similar to the process of a baby learning to speak, requiring a large amount of text, code, and other information crawled from the internet to be input into the AI model. The fine-tuning stage is akin to school education, usually having clear rights and wrongs, answers, and directions. Through specific datasets, we can cultivate the model with the required capabilities.

Therefore, the required data is also divided into two categories: one category is a large amount of data that does not require much processing, usually sourced from crawled data from UGC platforms such as Reddit, Twitter, Github, public literature databases, and corporate private databases; the other category requires careful design and selection to cultivate specific excellent qualities of the model, which involves data cleaning, filtering, labeling, and manual feedback.

As the capabilities of models continue to improve, various more refined and specialized training data will become key influencing factors for model performance. In the long run, AI data is also a long-term strategy with snowball potential. With the accumulation of preliminary work, data assets will have compounding capabilities, becoming more valuable over time.

In the Web3 space, the concept of DataFi has emerged. Web3 DataFi has multiple advantages:

1. Smart contracts ensure data sovereignty, security, and privacy.
2. Natural geographical arbitrage advantages attract the most suitable globally distributed workforce.
3. Clear Incentives and Settlement Advantages of Blockchain
4. Conducive to building a more efficient and open "one-stop" data market

For ordinary users, DataFi is the most beneficial decentralized AI project to participate in. Users do not need to sign data factory contracts; they can participate by simply logging into their wallets and completing various simple tasks, such as providing data, evaluating models, using AI tools for simple creations, or participating in data trading.

Currently, a number of potential projects have emerged in the Web3 DataFi field, such as Sahara AI, Yupp, Vana, Chainbase, Sapien, Prisma X, Masa, Irys, ORO, and Gata. These projects focus on different data service areas, including data collection, model evaluation, personal data monetization, on-chain data indexing, and more.

Although the current barriers to these projects are generally not high, once users and ecosystem stickiness are accumulated, the platform advantages will rapidly accumulate. Therefore, early-stage projects should focus on incentives and user experience. At the same time, these data platforms also need to consider how to manage manpower and ensure the quality of data output, avoiding the situation where bad money drives out good.

In addition, improving transparency is also an important challenge faced by current on-chain projects. An increasing number of projects need to demonstrate a public and transparent long-term commitment to promote the healthy development of Web3 DataFi.

The large-scale application path of DataFi includes two aspects: first, attracting a sufficient number of ordinary users to participate in data collection and generation, forming a consumer group for the AI economy; second, gaining recognition from mainstream large companies, as they are still the main source of large data orders in the short term.

Overall, DataFi represents the process of long-term cultivation of machine intelligence by human intelligence, while ensuring the benefits of human labor through smart contracts, ultimately achieving a feedback loop from machine intelligence to humanity. For those who are filled with uncertainty about the AI era, joining DataFi is undoubtedly a wise choice in line with the trend.