MENLO PARK, Calif. - Meta on Wednesday launched a conversational chatbot that will allow users to search its social media networks using artificial intelligence - an attempt to compete with OpenAI's popular ChatGPT amid an industry-wide boom in generative AI.
Meta said the conversational assistant, Meta AI - which will populate WhatsApp, Messenger and Instagram - relies on one of its large language models and a partnership with Microsoft's search engine Bing. The company is also launching AI-backed photo creation tools and 28 AI-powered chatbots, played by celebrities and cultural icons such as Snoop Dogg, Tom Brady, Kendall Jenner and Naomi Osaka.
"Meta AI is your basic assistant that you can talk to like a person," Meta CEO Mark Zuckerberg said in a keynote address at the company's Connect developers conference.
Meta has long been a powerhouse of artificial intelligence research. Behind the scenes, the technology powers countless features: from the algorithms that surface viral content on its social networks to the systems that flag toxic content. But as Google and OpenAI have pushed out advanced chatbots and other stand-alone AI-products, Meta appeared to be falling behind.
"They haven't really gotten credit as being seen as an AI winner," Bernstein senior internet analyst Mark Shmulik said this week. Though the company has excelled at "building foundational models … it's just unclear what their commercial strategy there looks like," he added.
Now, Meta is seeking to change that perception, introducing a fleet of AI tools throughout its products.
Zuckerberg emphasized that Meta's strategy involves creating different AI products for different uses, as opposed to a single flagship chatbot. He added that Meta's conversational bots aren't intended to just convey information - they're meant to be entertaining.
"I think this is going to transform the way that people use all of our products," Zuckerberg said.
The new assistant will have limited availability in the United States. Meta also expects to bring the new assistant to Ray-Ban Meta, the newest version of its smart glasses and its newly released Quest 3 headsets.
The company said Meta AI will be able to help users answer questions in real time. For instance, users who are on a hike in Santa Cruz, Calif., can ask the tool for other locations to explore, according to the company. Meta AI will also be able to drum up creative images from text-based prompts.
New features on Instagram will let users create AI-generated images. One of these tools, Restyle, enables people to transform their photos into new styles, including making them look like they were painted with watercolors. Another, Backdrop, can change the scene or background of an image based on a text prompt such as "Put me in front of a sublime aurora borealis," the company said. Meta said that any images that are created or edited by Meta AI, Restyle and Backdrop would be labeled.
Celebrity-driven chatbots are intended to give users the impression they are talking to familiar people about their interests on Instagram, Messenger and WhatsApp.
Meta's announcements arrive in a crowded arena, as other tech giants such as Google, Microsoft and OpenAI battle for dominance in generative AI - a technology that has captivated Silicon Valley. On Monday, OpenAI announced it was enabling users to talk with its ChatGPT chatbot. That same day, Amazon said it had signed a deal to invest up to $4 billion in the AI start-up Anthropic.
Meta follows a cohort of start-ups and major companies creating personality-driven chatbots designed to offer companionship and entertainment along with information. Earlier this year, the ephemeral message-sharing app Snapchat launched My AI, a chatbot designed to give users recommendations on everything from their cars to planning a wedding. The company has started experimenting with integrating sponsored links in the chatbot's conversations with users.
Last year, two former Google employees launched Character. AI - a chatbot that allows users to start conversations with impersonations of Sherlock Holmes or Albert Einstein. In its first five months, the company said users have sent more than 2 billion messages and spend on average more than two hours daily, according to the company.
Shmulik said the action makes "perfect sense" as a way to make Meta's social media networks more engaging.
"When you look at some of the engagement metrics around it - people are really spending a lot of time, you know, talking to, like, Einstein," said Shmulik, who cited Character.AI and Snapchat as models. "There clearly is some appetite to kind of talk to characters."
Over the past year, Meta's AI investments have had an uneven reception. In August 2022, Meta released BlenderBot 3, an AI chatbot that was quickly ridiculed for spewing bigoted tirades and conspiracy theories. Just three months later, ChatGPT's debut gained mainstream appeal for its sophisticated conversational skills.
This public image was on display in May, when the White House summoned the CEOs of four top AI companies to talk about policy with Vice President Harris. Notably missing: Zuckerberg. Later that month, the Senate Judiciary Committee held a hearing on AI oversight that included leaders from OpenAI and IBM - but not Meta.
Shortly after the White House meeting, Meta announced it was launching AI Sandbox, which enables marketers to use AI to create a range of text and visual options for their social media ads. The company has also been testing an internal productivity assistant known as Metamates and a generative-AI-based tool to enable users to modify photos via a text prompt and share them on their Instagram Stories.
And the new AI-powered personas are likely to prompt debates among regulators and activists about how the company plans to protect users from chatbots that spew toxic content at vulnerable users. Privacy hawks may also question how the scandal-plagued company will protect users' data and how much information it will collect from its social networks to train the underlying models powering the personas.
Sen. Edward J. Markey (D-Mass.) sent a letter to Zuckerberg on Wednesday urging him to pause plans to release AI chatbots until the company understand the effect they will have on young users.
"These chatbots could create new privacy harms and exacerbate those already prevalent on your platforms, including invasive data collection, algorithmic discrimination, and manipulative advertisements," Markey wrote.
Meta's open-source approach, allowing companies and researchers to alter the source code, has also come under fire from regulators and activists who argue giving away technology opens the door for misuse. Last month, Tristan Harris, the co-founder of the Center for Humane Technology, told members of Congress that his team had easily stripped Meta's large language model of its safety controls, allowing it to tell users how to develop a biological weapon. (Zuckerberg retorted that anyone can find that information on the internet, The Washington Post has reported.)
On Wednesday, Meta said all AI models have the power to spread inaccurate or inappropriate content but the company would continue to improve the safety and accuracy of its chatbots.
Meta has also been pushing back against the doomsday scenarios from some prominent AI leaders such as self-described decision theorist Eliezer Yudkowsky, who say the tech is advancing so quickly that it might surpass human intelligence. President of Global Affairs Nick Clegg and others have urged regulators to move cautiously, arguing that those existential threats are hypothetical.
Even if Meta overcomes regulatory hurdles, the company will still have to figure out how to make money off its AI research. The company has been recovering from a tough post-pandemic economic slump.
But since the beginning of the year, Meta's stock has risen more than 140 percent, thanks to improvements in the digital advertising market, the company's job cuts and the growing popularity of its TikTok competitor Reels.
"When your core business is doing well you're afforded more leeway to do some of these longer-term, more out-there type of initiatives," Shmulik said. "Whereas if the core business is challenged, you know, you don't get a lot of leash or runway."
Bhaskar Chakravorti, dean of global business at the Fletcher School at Tufts University, said Meta's personality-driven chatbots could help the company attract younger users. Meta has long worried about younger people flocking to apps such as Snapchat and TikTok.
"If these chatbots get believable enough and credible enough, they could actually draw Gen Z," Chakravorti said. "I mean, right now they are a little corny, but I'm the wrong generation to ask that question."