AI can be simplified into three pillars – compute, algorithms, and data.

Compute refers to the hardware and infrastructure created by companies like Nvidia. 

Algorithms are the mathematical models or methods that AI systems use to analyze data. OpenAI fits into this bucket. 

The third pillar is data. It’s the lifeblood of AI. AI systems rely on vast amounts of data to learn patterns, make predictions, and continuously improve their performance.

The leader in AI data is a company you’ve likely never heard of. And even more surprising, they’ve got an army of 100,000 low-skill laborers around the globe performing the grunt work that’s powering today’s AI revolution.

Introducing Scale AI. 

Scale AI is a data-processing startup that provides services to OpenAI, Meta, the Department of Defense, and many more. 

Scale AI’s clients task them with requests ranging from composing haikus and summarizing news articles to writing funny jokes in various languages.

Here’s a few real-world examples:

Tesla and Cruise hired Scale AI to label hundreds of thousands of stop signs to help train their autonomous vehicles.

Meta hired Scale AI to create 27,000 question-and-answer pairs to help train its AI chatbots on Instagram and Facebook. 

The U.S. Army even hired Scale AI to help build datasets for its AI efforts. 

To complete these tasks, Scale AI uses the least-tech savvy process imaginable: it employs an army of 100,000 low-skill laborers around the globe to create and label data manually. 

Yes, the lifeblood of AI relies on 100,000 low-skill laborers in places like the Philippines, Vietnam, Kenya, Poland, and Brazil. 

Here’s an example of a typical workflow:

• A worker in the Philippines applies for a job to become an AI writing evaluator for Scale AI. 
• The worker is hired and assigned to write text that the model can use to learn how humans think. Sample assignments could include creating and answering such questions as: 'Explain the Cold War to a 5 year old,' or 'Write a story about dinosaurs,' or 'Give me some books about humanity’s first contact with aliens.'
• Scale AI provides the writing samples to customers, like Meta, to train their AI models, with the goal of teaching them how to converse in more helpful and natural ways.
• In the Philippines, the worker is paid on average $8 an hour.

It’s certainly not the sexiest AI-related business. Wang himself actually called his business “mundane and unsexy.”

The good news for Wang, however, is this “mundane and unsexy” business recently made him one of the youngest self-made billionaires in the world after raising a $1 billion funding round at a nearly $14 billion valuation.

According to the Wall Street Journal, the labor-intensive operation has become so in demand by businesses eager to enter the AI race that Scale AI’s revenue pace tripled last year.

This highlights the importance of data in today’s AI revolution.

Going back to the three pillars:

Compute will likely become more commoditized in the future as competition improves.

Algorithms will be constrained by the availability and quality of the data they are trained on. 

This leaves data as the key differentiator. Access to high-quality datasets will soon be the primary source of competitive advantage for companies developing AI.

Turns out that competitive advantage is in the hands of 100,000 low-skilled laborers around the globe.