As Easy as 'Inhyeong Nungal Butigi'... A 20-Year-Old Who Made 1 Trillion Won Starting a Company [Tech Talk]

Scale AI Co-founder Alexander Wang
Joins Data Labeling Startup
Achieves Personal Wealth of '1 Trillion' in Mid-20s

The most important factor when training artificial intelligence (AI) is data. However, not just any data can become 'training data.' It requires filtering and refining the data to make it useful for training. This process is commonly called 'data labeling.'

Data labeling can be easily performed even without specialized AI-related job skills. However, because an enormous volume of data must be processed, a large workforce is needed.

This is why it has earned the nickname "putting doll eyes" in the computer industry. Yet, there is even a young person in their 20s who entered this "putting doll eyes" business early and grew their personal assets to 1 trillion won.

Alexander Wang, who started a data company at 19

Alexander Wang, founder of Scale AI [Photo by YouTube]

Last May, Silicon Valley startup Scale AI raised $1 billion in investment. Although Scale AI is still a private company, its estimated corporate value is currently $13.8 billion, and the personal assets of co-founder and CEO Alexander Wang are reported to be close to 1 trillion won.

Wang founded Scale AI in 2016 when he was only 19 years old. After eight years, he has become a billionaire in his 20s leading a global business. He has also earned the honorable title of a first-generation tech entrepreneur who built wealth through AI.

Developers from developed countries dominate labeling business using part-time workers from developing countries

As explained earlier, Scale AI is a company that handles data labeling tasks. Leading global AI companies such as Meta, Google, DeepMind, and OpenAI have signed memorandums of understanding (MOUs) with Scale AI for business cooperation.

In particular, Scale AI focuses on data labeling related to computer vision. This is partly because when the company first established itself in the industry, it mainly provided services to autonomous driving AI development companies.

An example of data labeling. To enable a computer vision system to recognize cars, manually marking the cars within images makes the process easier.
[Image source=Scale AI homepage]

So, what exactly is data labeling? Modern AI operates by learning patterns recognized in training data. To utilize these patterns more clearly, it is necessary to highlight important parts of the data so that AI can better recognize them. For example, to help a computer vision AI more clearly recognize an object like an apple, you would draw an outline around the apple in the photo. This kind of work is data labeling.

Data 'putting doll eyes' is an ongoing process

Data labeling itself can be done by anyone, not necessarily a highly knowledgeable computer scientist. However, since labeling must be done for thousands or tens of thousands of data items, an enormous amount of labor is required.

Scale AI was the first company to platformize data labeling work and actively utilize inexpensive labor from developing countries to implement cost-competitive labeling services.

Today, Scale AI has core development headquarters in Silicon Valley, the cradle of the international IT industry, and London, UK, but the actual labeling work is performed by part-time workers in developing countries in Asia and Africa. Thanks to this, they have been able to offer labeling at much lower prices than competitors and have now established themselves as a global subcontractor for labeling services.

Automation is just the tip of the AI iceberg... the workforce demand keeps growing

Scale AI's business model relies on inexpensive labor in developing countries. The photo is not related to any specific expression in the article. [Image source=AP Yonhap News]

The secret to Scale AI's success resembles the traditional manufacturing industries such as toys and textiles. The headquarters in the US and UK focus only on design and technology development, while the actual manufacturing workforce is employed in low-cost developing countries. In that sense, the nickname "putting doll eyes" for data labeling is quite an apt metaphor.

At the same time, it is a reminder that AI still cannot be properly implemented without human hands. Automation using AI is only the very tip of the massive iceberg called the "AI industry."

In reality, countless labelers involved in training tasks, personnel to expand and maintain data centers, IT engineers to manage and integrate systems?all these "human collaborators" are needed. Moreover, as AI investments continue to increase, the demand for such personnel will only grow higher.