Human Role in AI Data Creation to Persist for Decades, Claims Startup CEO

Despite the swift advancements in synthetic data, Invisible Technologies' CEO contends that human involvement in AI training remains indispensable.

On a recent episode of the '20VC' podcast, Matt Fitzpatrick, CEO of Invisible Technologies, challenged a common industry belief that AI will soon manage its own training without human intervention. Fitzpatrick, who took the helm at the company last year, recalled encountering persistent skepticism that came with the prediction that synthetic data would obviate the need for human input within a few years. He dismissed this notion as inherently flawed.

Synthetic data, while useful in situations where real data is limited or confidentiality is paramount, is essentially manufactured information aimed at supporting AI and machine learning development. Conversely, human feedback introduces a critical level of curation—real individuals are responsible for evaluating, organizing, and honing AI outputs.

Fitzpatrick elaborated that AI faces a multitude of complex challenges across various domains where understanding nuanced language and cultural subtleties is crucial, such as within the legal sector, which is replete with confidential information.

He emphasized that even as generative AI technologies evolve, the necessity of human oversight in these processes will persist for many years. This perspective is gaining traction throughout the industry, according to Fitzpatrick.

Before his current role, Fitzpatrick was integral to the operations of QuantumBlack Labs at McKinsey, focusing on AI research and development.

Invisible Technologies, which secured a $100 million funding round in September, positioning the company at a $2 billion valuation, operates alongside other key players like Scale AI and Surge AI. As the competition among tech firms for superior training data escalates, these businesses employ vast numbers of contractors dedicated to refining AI abilities in areas such as mathematical reasoning, scientific knowledge, coding expertise, and even social attributes like humor and empathy.

Continued Emphasis on Human Expertise

Joining voices with peers such as the leaders of Mercor and Handshake, Fitzpatrick underscores the sustained demand for human input in data curation.

Mercor's CEO, Brendan Foody, underscored the business's commitment to data excellence, attributing success to the employment and management of exceptional personnel.

Similarly, Garrett Lord of Handshake highlighted a shift in the data labeling sector, moving from generalist roles to specialized positions requiring profound expertise in fields like mathematics and science.

Lord pointed out that AI systems have ingested an extensive variety of content available online, yet their sophistication does not negate the need for experienced human specialists who can enhance AI's educational processes.