Turing, an artificial intelligence data startup, announced a threefold revenue increase, reaching $300 million last year, while also achieving profitability. The Palo Alto-based company, which counts OpenAI, Google, Anthropic, and Meta among its clients, specializes in providing human trainers to AI labs.
The Rise of AI Data Companies
With AI systems advancing rapidly, human expertise is becoming crucial in refining these models. Companies like Turing and Scale AI play a key role by supplying highly skilled human trainers, which reduces the operational burden on AI companies. This demand has elevated valuations, with Scale AI reaching $14 billion last year.
Turing, which was last valued at $1.1 billion in 2021, facilitates AI training by sourcing specialized human experts, including:
- – Software developers
- – Scientists with PhDs
- – Industry professionals with domain expertise
These experts label and annotate data to help AI models learn more efficiently.
The Cost of Human-Trained AI
AI training using human-labeled data is an expensive but essential process.
Turing revealed that:
– A single complex annotation can cost hundreds of dollars.
– Advanced AI models require millions of annotations for optimal performance.
For instance, Meta’s Llama 3 models were trained using over 10 million human-labeled annotations, according to Meta executive Joe Spisak.
The Growing Reliance on Human-Labeled Data
As AI labs face the “data wall”, a term describing performance plateaus due to limited high-quality internet training data, human-curated datasets become increasingly essential.
Turing’s CEO, Jonathan Siddharth, highlighted the company’s role in bridging this gap:
“Companies like Turing are helping the scaling laws keep going to make up for the data deficit that we have.”
Turing’s rapid growth underscores the critical role of human expertise in AI training. With rising demand for specialized data labeling, the company is well-positioned to continue scaling alongside AI’s evolution. As AI models require more curated data, human trainers will remain a cornerstone of AI advancements in the years ahead.
