Key Takeaways
When Facebook became Meta in 2021, the rebrand was positioned as a means for the company to “bring the metaverse to life.”
In the years since, the firm’s metaverse division Reality Labs has lost billions of dollars. But although critics have frequently dismissed CEO Mark Zuckerberg’s bold vision for the metaverse as an expensive vanity project, Meta’s Q2 earnings week presentation reveals that he shows no intention of giving up.
Four years after it was created, Reality Labs is still hemorrhaging money. Year-over-year, Q2 losses increased by 20% to $4.488 billion in the last quarter, Meta’s financial results revealed on Wednesday, July 31.
Due to ongoing product development efforts and investments to scale the ecosystem, Meta CFO Susan Li said the company expects the division’s losses to ramp up further in the coming quarters.
Nevertheless, she insisted that “Reality Labs remains [a] long-term initiative that we continue to invest meaningfully in.”
From the outset, Reality Labs was a vessel for Zuckerberg’s metaverse ambitions—a kind of Willy Wonka laboratory where Meta could cook up all manner of fantastical virtual and augmented reality projects without being constrained by budget.
But with the metaverse having now lost its status as the tech industry’s foremost buzzword in favor of unbridled hype over generative AI, some started to question Meta’s forays into virtual reality.
Against this criticism, Zuckerberg used his address to shareholders on Wednesday to expand on some of Reality Labs’ successes and explore “how AI is shaping our metaverse work.”
Attributing strong sales of Ray-Ban Meta Glasses in part to their AI features, he said the augmented reality specs have been “a bigger hit sooner than we expected.”
Last week, Meta also revealed that its Llama-powered AI platform Meta AI will be integrated with Quest later this summer, enabling headset users to interact with a smart assistant that can process live video inputs.
Finally, Meta’s new AI Character Studio deploys the firm’s latest open-source AI model Llama 3.1 to let users design their own chatbots.
Launched in the US on Monday, July 29, Meta’s AI Studio emphasizes customizability, letting users adjust their chatbots’ name, personality, tone, avatar and tagline. Custom chatbots can then be shared via Instagram, Messenger or WhatsApp.
The initiative chimes with Zuckerberg’s recent comment that “there should be a lot of different AIs out there […] not just one.”
“A lot of what we’re focused on,” he added, “is giving every creator and every small business the ability to create AI agents for themselves.”
Meta’s latest product aligns with general trends in the space, which has seen the monolithic chatbot model increasingly displaced by plural AI “agents”. But it could also serve to further Zuckerberg’s metaverse ambitions.
To date, gaming has seen some of the most effective integrations of AI-generated characters into metaverse applications.
While game developers have used decision tree algorithms to program non-player characters (NPCs) for years, the advance of highly sophisticated large language models (LLMs) in recent years has allowed for much more dynamic, responsive NPCs.
Today, platforms like Convai and NPC Playground make it easier than ever for gamers and developers to build and their own LLM-powered NPCs and deploy them in various metaverse settings.
In a hint that Meta is looking to explore the concept further, earlier this month, the company posted a job listing for an AI researcher to help develop “ immersive and interactive content, especially for games, across the Metaverse.”
The role will work mostly on Horizon, Meta’s flagship virtual world, and will involve researching and prototyping: “new consumer experiences where entirely new types of gameplay could be created, such as games that are non-deterministic, personalized, and change every time you play them.”