Superintelligence At Meta And Elsewhere
Forbes contributors publish independent expert analyses and insights.
In recent weeks, people have been hearing the term "superintelligence," tied somewhat loosely to artificial general intelligence or AGI, and Ray Kurzweil's early theory of a singularity event. There was an essay penned by Zuckerberg himself, (covered here at Popular Mechanics and elsewhere) published in August where Meta's chief talked about practical steps on that journey. There was also the announcement that Meta will be releasing smart glasses that will revolutionize the interface - so that your AI will be able to see what you see in real time.
In our IIA event at Stanford, Clara Shih, the head of Business AI at Meta, was interviewed on the Luminary stage by Shirin Ghaffary, a Bloomberg reporter. She talked about this and other projects going on at Meta, and the overall work of teams on building solutions around superintelligence.
Shih noted her familiarity with the venue as a Stanford alum, and then described how Meta is maintaining labs to work on AI models. There's also the Meta.ai platform, which she estimated as having 100 billion monthly users, and over a billion downloads of Llama models to date.
She also mentioned how Meta is developing "business AIs" which she characterized as "turnkey customer chatbots" that companies can deploy across WhatsApp, Facebook and Instagram.
She talked about how this can work for businesses in emerging markets, outside of the U.S.
"Many of (these businesses) are small," she said. "They're running their business on Messenger, or on WhatsApp. So that's where all their customers are. They have pictures of their product catalogs and menu items that they upload, and send and message to their customers all the time. Their customers are messaging them, asking for products, asking questions."
The AI chatbot, she suggested, will help to automate these processes.
"What we've been able to do is: we've realized that these small businesses in particular, they can't get back to everybody," she added. "It's hard just to scale as they grow in their business. And so we offer them ... turnkey business agents, for them to delegate which topics they want the AI to answer for them. They're able to turn that on, and it'll respond on the business's behalf."
In America, she noted, things are different, and the service is set up differently.
"In the U.S., businesses are more sophisticated," she said. "It's a more complex enterprise software landscape. So we are integrating with the systems that these businesses are using, and building a digital sales concierge that lives on both meta surfaces. Think Facebook and Instagram ads, as well as on the merchant website."
Shih spoke to the overall goals of the project this way:
"What we want to do is deliver personalized shopping experiences on behalf of each brand, and do it in a way that's really accessible, so that even if you're a small business and you don't have a big IT team or you don't have any developers, you can turn this on within a few minutes," she said.
The pair talked about CRM, and AI agents, and how an estimated 95% of AI enterprise projects fail.
"It's hard to get AI projects, especially autonomous agents, working," Shih said. "There's so much context that's required. That's why everyone talks about context engineering. And so I think what we're seeing is that domain-specific, narrow agents are working the best, because you can really understand it, you can feed it the data and the context that it needs, to be able to automate the planning and the reasoning steps that you want."
Other topics included vibe coding, and how to keep younger users safe in social media spaces.
"I know that Meta takes teen safety and child safety very seriously," Shih said. Guardrails, she added, are important, and engineers have to think proactively about hallucinations, user trust, and other key issues.
Later, there was this exchange, where Ghaffary asked Shih to define superintelligence and talk about applications.
"Super intelligence ... is AI that surpasses human ability across domains," Shih replied. "And it's a really powerful concept ... AI is already transforming everything, from drug discovery to physics, chemistry, agriculture, education. I really think we're on the cusp of just this incredible moment - Meta's vision is to empower everyone, not just a few companies, but every individual and even the smallest of businesses."
She talked about businesses using a Meta tool called "advantage plus," which consists of AI-powered automation for ad placement, and the AI assistance that they will get from the chatbots.
At the end of the interview, Ghaffary asked about risks. Shih had this to say:
"I think it's up to us, and it goes back to the importance of trust and safety, and knowing there (are) risks that we can talk about and anticipate in advance, and there's a lot of work being done to do that, and then there (are) risks that we can't anticipate, and being very vigilant that as those emerge, that we take action," she said. "I would say I'm more optimistic than pessimistic, because of the very brilliant people that I work with every day, who are thinking about this, working on it, ethically."
All of this suggests that we are very near some critical mass for AI in business. Not only in business, but in our personal lives as well. Business leaders are thinking carefully about just how to apply these powers, and what the fruits of LLM research will look like as we move forward. I thought this provides some good illustrations of what's going on at Meta after Zuckerberg's recent promotions of superintelligence as an imminent phenomenon.