Variant: Interpretation of Sapien, a decentralized data annotation network

Author: Jesse Walden, co-founder of Variant Fund; Translated by: 0xjs@黄金财经

"Data is the oil of the new era." This analogy is apt because data is the resource that powers the most transformative AI models, and like crude oil, data must be processed and refined to be useful. This largely human-dominated process, data annotation, is already a $2.2 billion industry in 2022 and is expected to grow by orders of magnitude in the coming years.

However, with the rapid and continuous development of AI, the demand for annotation has exceeded the supply. Many customers cannot get the data they need because annotation companies lack the skilled labor to provide professional services. Others cannot afford it because the price is too high. This high-growth but supply-constrained industry seems to have reached a stage where innovation is urgently needed - we think encryption technology can be a breakthrough.

Variant is pleased to announce that Variant has led Sapien's $10.5 million seed round to expand its decentralized data annotation network.

About Sapien

The leading idea behind Sapien is that cryptography can be used to build a larger annotation network faster while maintaining high data quality. The team is building a protocol to organize and compensate human annotators (or "AI workers") around the world more efficiently than centralized companies - including part-time workers, skilled workers, and even small businesses. Here are a few ways we think cryptography can give Sapien the edge it needs in this space:

1. Global network: Data annotation is a global industry, and large companies (mainly in the San Francisco Bay Area) often set up branches in Asia, recruit local workers, and then sell the labeled data to customers around the world - this creates friction when switching between different banking, legal, and labor systems. An Internet-native protocol is the smoothest platform for these different players to interact. Sapien can easily open up any new geography for customers or annotators, including convening skilled workers around the world to complete niche and specialized tasks.

2. Optimize cost structure: Sapien will use crypto rails and stablecoins to handle all payments, and rely on protocols for coordination as much as possible rather than manual management. These innovative measures enable it to automate some of the operational redundancies that are currently plaguing the industry, which will reduce costs.

3. Token incentives: Many new annotation networks have struggled in the early stages of development, but as a tokenized market, Sapien can use tokens to incentivize supply on a larger scale. Sapien can also use rewards and its on-chain reputation system to design better incentives to improve the accuracy of annotation data - so far, they have reported an accuracy rate of 99%. Finally, distributing network value and ownership should incentivize more skilled workers to join, as this allows them to gain potential benefits as the network grows.

With these innovations, we believe Sapien can grow faster than its competitors while providing more cost-effective services. This model is still in the experimental stage, but if executed well, it could revolutionize the entire industry.

Although Sapien is still a young startup, it has made great progress and is growing rapidly. Its early customers include some of the world’s largest enterprises, such as Alibaba, Baidu, and TAL, as well as well-known AI research institutions such as Midjourney. Its network includes participants from more than 70 countries, made possible by partnerships with large organizations such as Worldcoin and Yield Guild Games. The network has annotated more than 2 million data points, has just passed 10,000 active annotators, and is growing at a rate of 50% per month.

The Sapien Team

Sapien’s progress is a testament to the strength of its team – a lean, elite group of industry veterans.

They are led by CEO Rowan Stone, a founding team member of Base who helped launch Layer 2, and Trevor Koverko, a veteran member of the Ethereum community since 2017 and founder of Polymath, one of the largest real-world asset (RWA) protocols at the time.

The team also includes other members who previously ran business development at Kraken, brought the Flow blockchain to market, and built traditional AI businesses.

In our opinion, they have the ideal combination of crypto-native go-to-market (GTM) and enterprise backgrounds for this space.