Meta is taking a different approach to the AI race

Meta is forging its own path in the AI race, as Microsoft and Google chase proprietary search and enterprise offerings.

Amid an overall rosy earnings report this week, CEO Mark Zuckerberg offered more details on how the company sees its role in the race around generative AI. The quarterly call comes on the heels of Meta’s release last week of the latest version of its large language model (LLM), Llama 2, as an open-source alternative to ChatGPT.

Meta posted solid Q2 numbers for its core ad business, with overall revenue growth of 11% YoY, giving Zuckerberg more leeway to continue pursuing longer-term goals, like integrating generative AI, as well as money-losing metaverse ambitions.

Zuckerberg once again touched on his vision of building generative AI-powered agents through which businesses can interact with users on Facebook, perhaps a more high-tech redux of the push the company made around more rudimentary customer service chatbots in 2016. Like Google CEO Sundar Pichai, Zuckerberg also discussed how generative AI is beginning to change the way ads are created on the platform.

“One of the things that’s different about how we think about this compared to some others in the industry is we don’t think that there’s going to be one single AI that people interact with,” Zuckerberg said in the call. “You can imagine a world…where over time, every business has an AI agent that basically people can message and interact with.”

Forced to open-source: While Zuckerberg cited Meta’s “long history of open-sourcing our infrastructure and AI work,” the company found itself in this position more or less by happenstance. Meta’s original Llama model was originally only available to researchers by invitation, until it was leaked on 4chan a week later.

This time around, the company chose to make the model open to anybody from the start, despite worries that open models might wind up in the hands of bad actors. Zuckerberg said in the earnings call that he expects the decision to lead to more crowd-sourced improvements.

“These are often improvements in safety and security, since open-source software is more scrutinized and more people can find and identify fixes for issues,” he said.

The open-source debate: Whether or not big companies should open-source the LLMs at the heart of the current frenzy around AI has been the subject of much debate in the tech world. While OpenAI originally took a more transparent approach to documenting its GPT series, it’s become more tight-lipped about releasing details of its models as it’s transitioned from a nonprofit research group to a commercial enterprise, citing fears of misuse.

But others in the industry say smaller, more open models will be needed to supercharge innovation in the AI space. Sierra Ventures managing partner Tim Guleri said in a previous interview with Tech Brew that open-source development could help startups find more uses for generative AI.

“There’s a big open-source wave coming,” Guleri said. “Then you’ll see the innovation cycle even accelerate when it comes to verticalized applications.”