Meta adds a human element to AI, while others warn it all could be too ‘human like’

Not to be outdone by the other tech giants launching new advances in generative AI products and tools over the last few weeks, Meta unveiled a number of major hardware and software updates related to its ambitions for virtual reality, mixed reality and generative AI at the annual Meta Connect developer conference this week.

Between debuting the new Quest 3 mixed reality headset and new Ray-Ban smart glasses, Meta CEO Mark Zuckerberg previewed the numerous new genAI tools it’s creating for users, developers and businesses.

To compete with OpenAI’s ChatGPT and Google’s Bard, Meta debuted its own new chatbot called Meta AI, which will be available on various Meta apps, and also will power the new Ray-Ban Meta smart glasses. (Just last month, Amazon announced a new version of its own smart glasses, which will be powered by a newly updated version of Alexa.)

Meta also announced a new “AI Studio,” a platform for building AI chatbot that will include an API as well as a new “sandbox” for people who don’t code but who still want to develop chatbots. Other plans include ways to integrate AI bots with the metaverse, new tools for creators and also new tools for businesses, which will all be released at a later date.

“This isn’t just going to be about answering queries,” Zuckerberg said during his Wednesday keynote. “This is about entertainment and about helping you do things to connect with the people around you and helping you accomplish the things that you want across whatever your different goals are.”

Meta is also developing ways to make its AI bots more engaging, which Zuckerberg said will help make the chatbots feel “fun” and “familiar.” For starters, it’s partnering with more than two dozen celebrities that will star as AI characters to chat with users about a range of interests. Some of the celebrities lending their likeness include Tom Brady, Snoop Dogg, Paris Hilton, Dwyane Wade, Kendall Jenner and Naomi Osaka. Other celebs involved include chef Roy Choi and social media stars like MrBeast and Charli D’Amelio.

Rather than limit Meta AI to the social media giant’s apps, the company is also partnering with Microsoft to let users search via Bing. (Just last week, Microsoft announced a new partnership with Snapchat to integrate Bing’s chat ads into Snap’s My AI chatbot.) Other new AI features include a tool for making stickers with text-based prompts that will be released next month, followed by new AI tools for creating and editing images coming “soon” — all powered by Emu, a new foundation model that powers Meta’s image generator.

This isn’t the first time Meta’s made an effort to make chatbots go mainstream. Back in 2016 and 2017, then-Facebook built out new tools for branded chatbots — and even a new chatbot store — while other startups created bot versions of celebrities like Lindsay Lohan, Cardi B and 50 Cent. However until, the generative AI boom began, the bots were often limited in what they could say, which limited how they could engage with people using them.

Risks of designing “human like” AI 

Despite all the star power, consumer advocates warn about the dangers of designing AI to appear “human like.” Earlier this week, the nonprofit Public Citizen published a new report about how anthropomorphic AI systems potentially pose threats to people that chat with them.

“Counterfeit people” could more easily manipulate humans in ways that make them unwittingly susceptible to marketing, as well as to more nefarious actors. For example, researchers have found that users are more likely to divulge personal information about themselves when talking with “human-like” chatbots, creating new risks for data privacy as well as psychological and emotional abuse.

Although people might not believe a chatbot that looks or sounds like a generic person, turning them into real people, especially celebrities, could be a “shortcut” for getting users to suspend disbeliefs, said Rick Claypool, research director in Public Citizen’s president’s office and the author of the report. In an interview with Digiday, he said the potential for abuse is “alarming,” adding that it’ll be “hard to stop businesses and bad actors from exploring [AI’s] worst capabilities.”

“It’s very risky territory in terms of hijacking these sort of ingrained human qualities, weaknesses even, for profit,” Claypool said. “It’s almost analogous to how millions of years ago we evolved to taste things that are sweet and fatty to help us survive. Millions of years later we like bacon double cheeseburgers.”

With Meta’s history of misinformation and data privacy, there’s also the risk of whether AI-generated content will further exacerbate concerns. Although Meta’s keynote focused on the potential benefits of genAI, Zuckerberg noted there’s “undoubtedly going to be some new challenges.” He added that Meta plans to roll out new AI capabilities “a bit more slowly than we normally would,” so the company can address problems before they scale. To improve transparency — a key concern many have raised about large language models in general — Meta also plans to publicly share system cards detailing how AI models work.

“We’ve tried to think about this in advance and brainstorm as to what these would be and build as many safeguards in as we can … and fine tuning the models to fit our safety and responsibility guidelines,” Zuckerberg said. “We’ve done a lot of red teaming with experts on this and we’ve been building guardrails around inappropriate conversations.”

Generating new potential for marketing

Some analysts said they perceived Meta’s genAI announcements as a “shot across the bow” of OpenAI’s ChatGPT and Google’s Bard. In a research note published on Wednesday by the equity research firm William Blair, the new tools could provide “an extended use case” across Meta’s family of apps and keep users engaged longer than they would’ve otherwise. Longer term, the analysts think Meta’s genAI tools give small- and medium-sized businesses more opportunities for targeted advertising, and for Meta to monetize it.

“Meta also showcased what the technology is capable of now, while also demonstrating where it thinks the technology can go in the future,” wrote William Blair analyst Ralph Schackart, adding the genAI tools will help Meta diversify its product portfolio and reach a broader consumer base. “We also believe AI is allowing Meta to develop and roll out new products (at scale) in significantly shorter time periods.”

Beyond generative AI, others see new potential for Meta’s ambitions across virtual and mixed reality, especially with new integrations with Xbox, more VR games — including a VR version of Roblox — and various other new features. And at $3,000 less than Apple’s Vision Pro, the Quest 3’s lower price point of $499 could help drive adoption, noted Sarah Morris, director of digital strategy and investment at independent advertising agency PPK. She thinks the potential for gaming, entertainment and social interaction are all uses that could appeal to mainstream consumers.

“There are a lot of cool opportunities out there with this portal into digital reality,” she said. “Virtual product placements fit so well as they will help make the experience feel even more like reality, allowing advertisers to be in an immersive environment where the consumer is visually and emotionally connected.”

Some agencies already have teams testing Llama 2, which Meta debuted in July. Some early results are “quite good,” said Elav Horwitz, head of applied innovation at McCann Worldgroup, who added that it’s easy to pilot because it’s free to use and also open-source. She thinks integration of Meta AI into the new smart frames could be interesting for live-streaming, but a big question will be if the new features help solve real human problems. 

“It always feels like they’re coming from a tech place ‘look at this cool thing we’re doing here,’” Horwitz said. “Why do we need 28 chatbots of celebrities What are the real use cases? Show me how the glasses can help disabled people. There could be amazing applications there.”

https://digiday.com/?p=520045

More in Media

Media Briefing: Publishers’ Q4 earnings paint a gloomy picture of 2023

An analysis of four publishers’ Q4 and full-year 2023 earnings.

The Rundown: The Trade Desk’s take on the next year in ad tech

Sharing a stage with leading media executives from PepsiCo, Samsung Mobile, and Unilever, leading execs at the DSP shared their vision for the year ahead.

How much can states regulate social media? The Supreme Court hears cases for and against

The U.S. Supreme Court addressed separate cases about a similar question: Can states limit social media companies’ moderation?