Meta Expands its Tests of Generative AI Features as it Looks to Play a Bigger Role in the Space


While more and more generative AI features are being added to social media apps every other week, I’m still yet to see a killer use case for AI in a social context.

Sure, AI tools can help you create unreal images of yourself, or of totally imagined scenes, and that’s interesting as a novelty. But beyond that, is there really a value to, say, enabling users to generate entire posts in-stream, which are not actually created by a human?

Like, isn’t that the antithesis of “social” media?

As such, it feels like we’re still waiting for new forms of generative AI that will complement the core use case of social apps, as opposed to diluting the authenticity of what you see, with authenticity still being the foundation of social interaction. And while many people are already using generative AI to come up with full posts on their behalf, in order to portray themselves as something that they’re not, it still feels like further enabling this is a misstep for social apps.

Which is why it’s interesting to see how Meta is looking to integrate generative AI, in varying forms, across WhatsApp, Instagram and Messenger.

As reported by TechCrunch, late last week, Meta confirmed that it’s expanding its test of its AI assistant tool to users in India and parts of Africa, as it looks to become a leading AI player in these markets.

Meta first launched its AI assistant with selected users in the U.S. last September, before expanding it to Ray Ban Stories users a few months later.

Meta AI assistant

The chatbot, powered by Meta’s Llama LLM, provides what’s now become a basic AI assistant offering, able to provide responses to conversational queries, and generate images in-stream. It’s essentially Meta’s answer to ChatGPT, though given the ubiquity of ChatGPT as the best-known conversational AI engine in Western markets, it’s going to be difficult for Meta to drive significant take-up of its in-stream AI tools.

But maybe, in other regions, Meta will see more success.

ChatGPT is the most popular AI chatbot in India as well, but WhatsApp is the most dominant app, with over 500 million users. Instagram also has 360 million Indian users, and Meta will be hoping that its expanded local market presence will make its chatbot tools a more popular offering, helping it to stake its claim as an AI leader in the region.

And in Africa, ChatGPT is still not widely available, which means that Meta could have an opportunity to become the clear AI chatbot leader in the region.

Though it remains to be seen whether users will actually be interested, because Meta has tried Messenger bots once before, back in 2016, and they didn’t resonate with users.

Generative AI has changed the game on this front, and the current wave of AI chatbots is much more advanced. But again, do chatbots like this really add significant, valuable functionality to these apps?

Meta’s also exploring other ways to utilize generative AI, which may provide more utility.

Instagram chief Adam Mosseri has noted that creators should be looking to use generative AI “to negotiate deals, or write contracts, or come up with ideas”. Which are less flashy uses of these tools, but are more aligned with the technical processes that drive their systematic approach, and it’ll be interesting to see if and how Meta looks to build in tools around these types of elements in future.

Meta’s also testing an AI chatbot in search to help you find content related to any topic:

Meta AI example

As you can see in this example, as a complement within search, Meta’s AI tools could help you find more relevant content matches in-stream. Though you would also assume that Meta would need to set some pretty strict parameters around how it can be used in this way, in order to avoid potential misuse.

Because that’s where Meta has run into trouble with similar tools in the past.

Back in 2013, Meta launched “Graph Search”, which was a more rudimentary form of conversational discovery, built into Facebook.

Facebook Graph Search

As you can see in this example, with Graph Search, you could enter queries like “People who like things that I like”, and the system would display users with similar interests, as per the detail on their Facebook profile.

Which was an amazing tool for finding people with related interests. But it also highlighted more questionable elements.

Facebook Graph Search

The capacity to search for virtually any interest or topic led to more headaches for Facebook, as people used the tool to highlight how the app is facilitating community around such interests. So, eventually, Facebook shut Graph Search down, and with this in mind, it’ll be interesting to see if or how it can avoid similar concerns with a generative AI powered search process in-stream.

But essentially, social platforms still haven’t cracked the AI code as yet.

While generative AI is amazing in its capacity to produce human-like responses to conversational queries, the actual use of that, in a social media context, is still questionable, as the system is more geared towards simplifying repetitive, technical tasks, than replicating actual human interaction.

Which is how it should be. We don’t want bot profiles taking over social apps, and really that’s the biggest concern, that AI bot posts will soon overwhelm actual human, “social” interaction, and as a result, make these platforms less engaging overall.

In that context, social apps should tread very carefully with the AI tools that they implement. Otherwise, they could be feeding through the rope that eventually strangles their offerings.