AI Euphoria May Fizzle Out Over Lack of Data

Tech giants are running out of data used to train AI systems. To overcome the problem, some companies are exploring new methods powered by 'synthetic data', write Cade Metz & Tripp Mickle

- Cade Metz & Tripp Mickle

Cassabis, one of the influential artificial intelligence (AI) experts in the world, has a warning for the industry: Don't continue to rely as they have over time on fairly simple concept systems. The more data from the internet that they feed into large language models behind chatbots - the better systems performed. Hassabis, who oversees Google's AI systems, says that method is a problem simply because they are running out of data. The industry is seeing the limits of this approach.

"The industry is seeing the limits of this approach," Dr. Hassabis said in an interview with The New York Times as he prepared to finalize his work on the project.

He's not the only AI expert to note a slowdown. More than 20 executives and leaders in the AI industry have observed a widespread sense that the industry is running out of data, a situation that to many was unimaginable just a few years ago. They are running out of the digital text and data from the internet.

The limits of this approach are starting to surface as billions of dollars continue to be poured into AI development. OpenAI, a data company, has already spent more than $10 billion in funding, making it the largest-ever private investment in a start-up. And the signals from investors in tech are signaling that they have no plans to slow down spending on the giant data centers that power AI systems.