Meta proposes a solution: why not remain inside its apps and converse with a bot? Mark Zuckerberg boasted last week that Meta’s AI models had been upgraded to the point where they were competitive with the best in the business. He shared the goals of his company’s Artificial General Intelligence (AGI) efforts. He offered some more concrete forecasts: “I believe that by the end of the decade, many individuals will engage in conversation with AIs on a daily basis through smart glasses such as the one we’re developing with Ray-Ban Meta.”
The business, however, is currently planning something else. Facebook, Instagram, WhatsApp, and Messenger are among the most used platforms that Meta is rolling out its chatbot to. In social feeds, the chatbot might make tips, comment on Facebook posts, or chime in whenever tagged in group messages. Like ChatGPT, you can have direct conversations with it. Meta has introduced assistants into many social settings; similar to how Google and Microsoft have included AI assistants into their work tools, it will create photos and compose messages on your behalf. It will be fascinating to observe how and if individuals utilize them in these situations, and Meta will certainly discover the answer pretty soon.
People have said that this is a desperate and clever approach. Is Meta just trying to keep up, investing in a passing trend, and releasing unfinished products to consumers? Or does Meta currently have more users than any other AI company, an important hardware business, and a suitable model? It isn’t easy to compare claims like this to actual AI models because everyone in the AI industry is heading in the same general direction, hoping to reach an undefined endpoint where they (or their investors) will find enormous wealth.
However, when put into practice, Meta’s AI reveals a less impressive side of its ideals. Users are most likely to interact with Meta’s chatbots while conducting a search. Facebook, Instagram, WhatsApp, and Messenger all have Meta AI search options. Without switching between apps, you can get up-to-the-minute data from anywhere on the web. Consider a Messenger group discussion where you and your friends are organizing a ski vacation. Ask Meta AI to locate flights from New York to Colorado and determine the least crowded weekends to go—all without exiting the Messenger app—using search in Messenger.
In reality, this deployment is both extensive and noticeable. The search bar has evolved into a hybrid of a chatbot and a search engine, allowing users to look for more than just individuals, pages, groups, places, or topics. Here is how it appears:
It works similarly to ChatGPT in that you may ask it anything, and it will come up with a response. When you ask Meta’s AI a question, it will typically give you a summary with links to relevant web pages, similar to what you would see in an AI-powered search engine like Google’s Search Generative Experience or Perplexity. This is different from other chatbots. When implemented properly, the goal is obvious: these features aim to make users never want to leave Facebook or Instagram rather than provide additional content inside. You can get your query answered right in Meta’s search box instead of leaving Instagram to go to Google or fumble around the web.
Though it certainly plays a role, this isn’t only about Meta increasing engagement. This kind of AI is costly to train and consumes a lot of computer power to operate; OpenAI charges customers for comparable products because it is likely costing Meta a tremendous amount of money to operate at this scale. Additionally, it lays out a strategy for a future when most people’s online experiences will be less centered in the web, which refers to publicly accessible websites that are not part of walled gardens like Meta’s. People nowadays use their smartphones not just to access the web but also to switch among apps and web browsers. Most apps make some kind of reference to the web, and links allow programs that don’t normally communicate with each other to do so. Meta provides a glimpse into a future when machines do much of the web browsing for you, reducing it to a library for summaries and references rather than an object you interact with directly.
This is bad news for the web and all the people and organizations that work on it. Artificial intelligence companies are using all the data they can get their hands on, which could make data even harder to come by and discourage its producers from sharing it. For example, Meta’s AI is currently built on top of search results from Google and Bing. But one can digress. When we initially used this Instagram feature, we typed “Washington DC” It brought up a list of profiles and a handful of suggested searches, one of which was, strangely, “Best restaurants in Washington DC?” So we went ahead and checked it out:
At least we didn’t ask for food, which is good. On the other hand, Meta’s AI is posing as a parent on Facebook, offering guidance to parents of gifted and impaired children attending public schools in New York City.
By the end of this decade, maybe Zuckerberg will be correct, and we will be able to have everyday discussions with AIs while wearing our Ray-Bans. At this time, though, Meta is insisting that we carry on with these discussions regardless of whether we want, need, or comprehend the responses we receive. It is us who are stranded here, testing the AI.
