AI Weekly: Meta’s flashy, AI-powered vision of the metaverse brushes over concerns

This week, overshadowed by Russia’s invasion of Ukraine, Meta (formerly Facebook) unveiled its ambitious plan for a “metaverse” powered by AI. During a presentation, CEO Mark Zuckerberg spotlighted several of the company’s related efforts, including an AI system that lets users build virtual worlds by describing them with text and internal projects to build a universal speech translator and a voice assistant that responds to voice, eye movements, and body language.

Few, if any, of the AI technologies that Zuckerberg showcased are fundamentally new. Systems that can generate interactive 3D models from text have existed for some time, as have systems that can directly translate speech from one language to another — one of the projects that Meta announced during its press conference. Meta’s push toward systems that can learn to translate language using only a few examples is duplicative of the work being done by startups including Hugging Face. And Amazon has long been exploring AI that can read body language, along with acoustic and linguistic cues, to better anticipate what users want.

It might be argued that Meta is positioned to surpass what’s been achieved technologically to date, of course, given its well-financed R&D division and troves of user data. (Last fall, Meta estimated that it would spend at least $10 billion on Facebook Reality Labs, the lab tasked with creating its augmented reality and virtual reality products.) But this week’s briefing served less than a demonstrating of cutting-edge technology than a pitch of the future that Meta envisions for itself, enabled by AI.

In this future, AI eliminates the language barrier, and intelligent assistants cater to users’ whims — transporting them to tropical beaches with a voice command. If only.

It’s undoubtedly a rosier platform than Facebook, which lost users for the first time ever during Meta’s most recent fiscal quarter. Countless reports including Meta’s own found that its platforms have enabled disinformation and misinformation to flourish, leading in some cases to riots and violence. According to internal documents shared by Facebook employee-turned whistleblower Frances Haugen and her legal team, more than 95% of hate speech shared on Facebook stays on Facebook — despite Meta’s repeated public claims to the contrary.

Meta’s optimistic version of the metaverse is less problematic. But it brushes over problems that have already begun to crop up in the virtual world, including abuse and cyberbullying. In a recent BBC investigation where a reporter posed as a 13-year-old girl, the digital avatar was harassed, catcalled, and chased. Meta was recently forced to implement a “personal boundary” in its virtual reality platform to combat lewdness.

Meta has maintained — and continues to maintain — that its moderation policies, backed by AI, will ensure that its metaverse remains a safe place to work and play. But given that the company has failed to achieve this on a more basic level, the prospect of a harassment-free, Meta-run metaverse is difficult to fathom.

AI looks unlikely to be Meta’s salvation. On the contrary, it’s often exacerbated or concealed the extent of the company’s moderation problems. In 2019, engineers at Meta reportedly discovered that a moderation algorithm at Instagram (which Meta owns) was 50% more likely to ban Black users than white users. More recent reporting revealed that, at one point, the hate speech detection systems Meta used on Facebook aggressively detected comments denigrating white people more than attacks on other demographic groups. And a mistranslation by Meta’s AI software led to the arrest of a Palestinian man in 2017 by Israeli police.

Even assuming that Meta overcomes the countless technical — and political — issues plaguing its AI systems, many groups, particularly speakers of uncommon languages, fear losing hold of their culture if it’s controlled solely by a corporation. While Meta may see an AI-powered metaverse as the solution to its largest platform’s woes — which, in addition to hostility and misinformation, includes declining ad revenue — it’s far from a sure bet that the public will embrace it. Nor should they; it’ll take more than aspirational, flashy tech demos to allay fears that Meta will fumble in the ways that it long has.

For AI coverage, send news tips to Kyle Wiggers — and be sure to subscribe to the AI ​​Weekly newsletter and bookmark our AI channel, The Machine.

Thank you for reading,

Kyle Wiggers

AI Senior Staff Writer