Mark Zuckerberg gives props to DeepMind's CEO for using him to get a better price from Google

• Google purchased AI research lab DeepMind about a decade ago for more than $500 million.
• Before the acquisition, Facebook also was in talks to purchase the startup.
• Meta CEO Mark Zuckerberg said he wanted to buy DeepMind at the time.

Meta CEO Mark Zuckerberg tipped his hat to DeepMind CEO Demis Hassabis for using then-Facebook to leverage a better deal out of Google in 2014 when the search-engine giant purchased the AI research lab for more than half a billion dollars.

About a decade ago, DeepMind was a little-known London-based AI startup founded by Hassabis, Shane Legg, and Mustafa Suleyman.

In 2013, Facebook was in deep talks to pick up the startup as the AI race was heating up, but those conversations fell through, The Information reported in 2014. A year later, Google purchased the company for more than $500 million, according to the report.

Reports on the acquisition at the time did not make clear why the talks with Facebook fell apart.

But in a recent interview with South Park Commons, a San Francisco-based tech community, Zuckerberg seemed to suggest that Hassabis used the company to get a better deal out of Google.

"I did want to buy DeepMind, but they went to Google," he said. "Demis was good, by the way. He totally did a very good job of playing me off of Google to get a good price, which I respect. Power to him. Power to him."

Zuckerberg didn't elaborate on that point. Spokespersons for Alphabet and Meta did not immediately respond to a request for comment.

Since its acquisition by Google, DeepMind has been focused on research to achieve artificial general intelligence, or AGI, which is where computers will showcase human-like reasoning.

Instead of acquiring DeepMind, Facebook launched its own lab with Facebook AI Research (FAIR) which Zuckerberg said in the interview with South Park Commons helped provide a "open-source" approach to Meta's work on AI.

When Meta launched its Llama model last year, the company touted the technology as open source, which essentially means that developers, researchers, or companies can access the model's innards to modify Llama for their own use cases. However, some critics challenged Meta's labeling and argued that Llama isn't truly open source.