Meta AI is a Generative AI Assistant But the AI Characters Shift Focus to Entertainment
Beyond productivity, consumption features are on the rise
“This isn’t just going to be about answering queries. 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 to across whatever your goals are.” Mark Zuckerberg, CEO, Meta
The past two weeks marked a shift in the focus of the generative AI chat applications market. Amazon's SVP of Devices, Dave Limp, distinguished between generative AI creation and consumption experiences during the debut of a large language model (LLM) enabled Alexa. That included an Alexa-native experience as well as integration with Character AI’s interactive personas.
Amazon’s approach differs from productivity-focused tools like ChatGPT, Google Bard, Bing Chat, and Perplexity. It also differs from the typical user experience of Alexa, Google Assistant, or Siri. Whether you consider it consumption or companions, you may comfortably deconstruct it into entertainment. That is precisely how Mark Zuckerberg distinguished between the new Meta AI assistant and several new AI-based interactive characters during the Meta Connect keynote.
Meta AI is your basic assistant that you can talk to like a person. You can message Meta AI in any of the messaging apps…and it’s going to help you answer your basic questions or requests. It is based on Llama 2. In addition, it also has access to real-time information and broader knowledge through a partnership we have done with Microsoft and Bing Search.
…
Our new image generation model EMU is also built directly into Meta AI, so you can just message it and tag anyting with “Imagine” and it will create high quality, photorealistic images right there in chat for free. Another neat feature of this is you [can] invoke Meta AI in any chat.
…
But in addition to this, we have also been experimenting with creating some AI’s that are a bit more fun; that have some more personality, opinions, interests, just a little more interactive and fun to play with…
This isn’t just going to be about answering queries. 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 to across whatever your goals are. Because we thought that this should feel fun and it should feel familiar, so we did something a little bit different for us. We partnered with some pretty awesome people to basically embody these and play them.
You can see the demonstration, including Snoop Dogg as a dungeon master, in the video above. The personalities don’t have real-time internet access, but Zuckerberg said that would arrive “soon.” Although the solutions are exclusively chat-based today, voice will come “in a couple of months…probably” early next year.
He also introduced AI Studio, enabling anyone to create their AI-based personalities via API. This will include a no-code environment in the future, though it will be a developer tool in the current alpha release. APIs for businesses that want to create AI assistants in Messenger and WhatsApp will be available in the “coming weeks.”
A Focus on Entertainment
The product information is pretty straightforward. This will look a lot like Character AI to many observers with the notable addition of moving avatars (or are they short video loops?). However, take note of what Zuckerberg is excited about in the video.
Meta is releasing its first true AI assistant, but Zuckerberg glosses over its productivity features. The assistant represents just two minutes of the video. This is Meta’s ChatGPT, Alexa, Siri, Google Assistant, and Bard competitor. It only merited 21% of the total keynote segment on the AI assistants!
In the other seven-and-a-half minutes, Zuckerberg focuses on how Meta AI can enhance messaging conversations by adding images or adjudicating debates over points of fact. However, his greatest enthusiasm is reserved for the AI characters based on popular celebrities ranging from Paris Hilton and Kendall Jenner to Snoop Dogg and Tom Brady.
These are all predominantly entertainment use cases. Even the “how-to” segments featured a celebrity chef and a YouTube DIY influencer.
Meta has tremendous distribution. Threads, Meta’s Twitter competitor, rocketed to 49 million daily Android users within two days, followed by a 79% decline over the next month. Will the AI personalities will gain this much traction? Will the use be sustained? It is hard to say how popular these AI personalities will be. Character AI has 18 million monthly active users, according to a16z. Maybe Meta can capture a similar level of interest.
If the AI characters do induce trial, they may serve as an onramp to getting users to try the more productivity-focused Meta AI, which would also be a good outcome for Meta. There are so many AI assistants already available that it seems unlikely many users would even try Meta’s new AI assistant without an interesting hook. Of course, the AI personalities, like with Character AI, may be compelling enough on their own to keep users engaged in entertaining chit-chat and fantasy role-playing games. Meta knows a thing or two about maintaining user attention.
Digital Twins on the Horizon
Meta is also planning to roll out a service enabling influencers to create digital twins. Zuckerberg said these features are still undergoing testing because there is higher concern about brand safety that could undermine an influencer’s image. That said, the wait is not expected to last much into 2024.
We also are excited about getting this in creators’ hands so you can build AIs that represent you to help you grow your community and engage with your community. But, one of the things as we are building that we think will come sometime probably in the first half of next year, because we think that getting the brand safety on that is just really critical. We want to make sure that if you are a creator that you can train these AIs that way they are going to say things that reflect just the way you want to reflect your brand and your community. That is still going to require some work to get there.
Creation vs Consumption
ChatGPT, Microsoft Copilot, Google Bard, Perplexity AI, and Anthropic’s Claude 2 all announced new features over the past three weeks. The productivity crowd regularly maneuvers for advantage or to maintain parity with their rivals. There is far less activity in the entertainment category.
It may be that the productivity-focused assistants add entertainment features, and users willingly adopt them. However, this seems unlikely. Focusing on one or the other will likely be required to remain competitive. Character AI, SK Telekom’s A. (pronounced “A dot”), and Meta’s new AI characters are creating a new category of AI for interactive entertainment, consumption, and companionship. This may turn out to be an even larger market than productivity in terms of users.
What is less clear is whether the interactive AI bot category will be as successful in generating revenue. If it does, you may find the entertainment segment is even more concentrated than in the digital productivity assistant segment.
What do you think? Is there merit to the thesis around productivity versus entertainment? Will users adopt Meta’s AI characters in any meaningful numbers? Let me know in the comments below.
Sure, but I could imagine entertainment taking a fair chunk of the market and the productivity claims undermining the proponents, ie, their own arguments.
"Is there merit to the thesis around productivity versus entertainment?" Yes, it's an empirically testable hypothesis: if Meta and others who go down the entertainment route succeed, that supports the hypothesis. If productivity is found to remain little changed then that also supports the entertainment side. It is a potentially disruptive strategy, because the productivity champions can't easily change course. I'm sure the productivity side won't lose everything, but entertainment may well turn out to be the better bet.