Article contents:
- Introduction: A hidden race at the heart of the AI revolution
- What is “data labeling” and why is it so important?
- The Invisible edge: Focusing on complexity and expertise
- Rapid growth and new leadership
- Beyond labeling: Expansion toward integrated solutions
- Fierce competition in the data market
- Conclusion: From hard work to thoughtful partnerships
- Frequently asked questions
Introduction: A hidden race at the heart of the AI revolution
While all eyes are on major artificial intelligence companies like OpenAI and Google, another equally important race is happening behind the scenes: the race to provide the “building blocks” upon which this revolution is built. In this context, Invisible Technologies, a competitor to the well-known Scale AI, has raised $100 million in a new funding round, pushing its valuation to over $2 billion. This investment highlights the growing importance of the “data labeling” sector, the laborious yet vital work of organizing the information that feeds AI models.
What is “data labeling” and why is it so important?
Data labeling is the process of simplifying and organizing the massive amounts of information consumed by artificial intelligence models so that the technology can use them and learn from them. The simplest example is teaching a model the difference between a picture of a cat and a picture of a muffin. In reality, however, the tasks are much more complex. Without high-quality and accurately labeled data, AI models would be inaccurate and unreliable. This field gained widespread recognition in June when Meta acquired a 49% stake in Scale AI, boosting investor interest in competitors.
The Invisible edge: Focusing on complexity and expertise
Invisible, which played a role in training the original ChatGPT, has sought to differentiate itself from competitors by offering labeling services in more complex fields. While almost anyone can classify pictures of cats, some advanced labeling tasks require significant expertise. Invisible offers what it calls an “expert marketplace,” which allows AI companies working on large language models to select data labelers with relevant backgrounds, such as advanced degrees in mathematics or astrophysics. This ensures specialized data is handled accurately by people who understand its context, resulting in more robust models.
Rapid growth and new leadership
The company is experiencing rapid growth, with sales doubling in 2024 to reach $134 billion. The company appointed Matthew Fitzpatrick, the former head of McKinsey’s AI software development group, as its new CEO in January and doubled the size of its engineering teams this year. Its major clients include companies like Cohere, Microsoft, and Amazon Web Services, demonstrating the industry’s trust in the quality of its services.
Beyond labeling: Expansion toward integrated solutions
Invisible’s services are not limited to data labeling. The company also offers a range of other products, including tools for model fine-tuning and measuring model data diversity. It also has enterprise-focused businesses targeting industries such as food and beverage, insurance, and healthcare, with products for supply chain management and generating text for customer service agents. This shows the company is moving toward becoming a comprehensive AI partner rather than just a single-service provider.
Fierce competition in the data market
Invisible faces intense competition in its core business. In addition to Scale AI, which is valued at over $29 billion, there are other companies like Surge AI, which has been in talks to raise $1 billion at a valuation of at least $25 billion. However, Neel Kamath, general partner at Vanara Capital—which led the latest funding round—believes Invisible’s strength lies in its close and specialized work with clients. He says: “There are a lot of companies competing on the labeling side that just throw bodies at the problem. Invisible’s strength really lies in doing thoughtful design partnership work with its large language model partners.”
Conclusion: From hard work to thoughtful partnerships
As companies increasingly hope to see tangible results from their investments in AI tools, the skill in delivering outcomes that boost the bottom line will become even more crucial. CEO Fitzpatrick says: “There isn’t a well-understood concept of how to put these models into production, test them, and validate them. What we think is really exciting about our next decade is that we simply believe we can make AI work.” The success of companies like Invisible Technologies proves that the future of artificial intelligence depends not only on clever algorithms, but also on the meticulous and thoughtful human work that makes them smart in the first place.
Frequently asked questions
Q1: What is Scale AI?
A1: It is the leading and most well-known company in AI data labeling, founded by Alexandr Wang and valued at over $29 billion following a major investment from Meta.
Q2: What is model fine-tuning?
A2: It is the process of taking a pre-trained large language model and training it further on a smaller, more specialized dataset to make it excel at a specific task, such as answering legal questions or writing code in a certain programming language.
Q3: Why does AI need human data labelers?
A3: Because AI learns from examples. Humans need to provide it with correctly labeled examples (such as “this is a picture of a cat” or “this review is positive”) so it can recognize patterns and apply them to new data.
Q4: Will the data labeling job itself be automated?
A4: Artificial intelligence is already being used to help automate parts of the labeling process, but complex and specialized tasks still require human judgment and expertise that machines cannot currently provide.
