AI is Unlocking the Next Generation of Social Media Platforms
Recent advancements in artificial intelligence and machine learning, along with changes in online consumer behavior, have created opportunities for next-generation social media networks.
AI has the potential to be the next paradigm shift in how we search and consume information. In this post, I will explore:
The unbundling of existing social platforms
Opportunities to apply AI for new social platforms
The opportunity for vertical social platforms
It's hard to deny the technology shift we are seeing with foundation models. Rapid technological advances have been made over the past few years, including the recent launch of OpenAI's GPT 3.5, which was trained on 175B parameters, and the soon-to-be-launched GPT 4. While enterprise application startups with solutions for coders, content creators, marketers, and sales reps are exploding, an opportunity is emerging for applications to leverage AI in serving curated, personalized content to the consumer. The transformer, (the “t” in GPT) is now good enough to provide a personalized algorithm. When Instagram was first built in 2010, the iPhone 4 had just launched, and the camera was just good enough for Instagram to gain traction. Everyone had a camera in their pocket, allowing easy content creation and consumption. Advances in transformer models enable three key themes for new verticalized social media platforms:
Personalized content curated for each user
Enhanced content creation tools that allow everyone to become their own creative director
New distribution algorithms that help content creators reach users who will be most engaged with their content
New social companies may find it challenging to gain traction, and even harder to retain early users. Established social media "giants" such as Facebook, Instagram, and YouTube dominate most users' screen time, and their network effects make it difficult to convince users to post or consume content elsewhere. However, recent advancements in AI present an opportunity for significantly improved user experiences across verticalized social media platforms. For the purposes of this blog post, a social network is defined as a place where users post, consume, and engage with content and other users. This can include traditional social networks such as Instagram, but also extends to gaming platforms like Rec Room or vertical social commerce companies such as Flip.
As AI-curated content becomes more prevalent, social media companies will increasingly resemble traditional media companies, with less emphasis on likes and follows. TikTok, for example, allows users to have a great experience without any "friends" or "followers." The next generation of social media platforms is less about finding your friends and more about finding personalized content, which creates a more authentic user experience.
The unbundling of legacy social media platforms:
The demand for personalized, curated, algorithms is evident in consumer behavior and the advertising dollars spent across social media giants. In 2015, 71% of teens aged 13 to 17 used Facebook, but as of today, only 32% do so. Facebook's popularity has been surpassed by YouTube, TikTok, and Snapchat. In fact, 67% of teens report using TikTok. Advertising dollars are shifting to TikTok, which provides users with AI-curated content, ultimately increasing user engagement. Q4 '22, Meta, Snap, and Pinterest saw their user bases grow, but their revenue declined. However, TikTok saw its annual revenue increase by more than 170% in 2022. Social networks that lack AI-curated algorithms have struggled to innovate over the past two years. For instance, Instagram rebranded itself as a video platform, pushing its short-form video format Reels in an attempt to expand beyond its roots as a photo-sharing app and compete with TikTok. Creators, brands, and publishers adopted video strategies to remain relevant, but now the platform is shifting back to prioritize photos, as the company struggles to compete with incumbents.
Meta, Snap, Pinterest, and TikTok combined generated north of $135B in revenue in 2022, creating a huge market opportunity for verticalized social media platforms worth $1B+. Ten years ago, a gaming-only network being worth a billion dollars seemed impossible, but now we have multiple examples such as Twitch, Discord, Roblox, and Rec Room! We also have multiple beauty related vertical social networks that overlap with commerce: Flip, Supergreat, Trendio. As teens shift their attention away from Facebook, there is an opportunity to unbundle Facebook. Teens are the easiest customers to capture, and they prefer personalized algorithms similar to TikTok's. The "unbundling" of Facebook is already occurring across categories including events, groups, live, personal CRM, and social context as users shift to other social networks.
The emergence of AI-curate social media platforms:
In the past few weeks, two new vertically-focused social networks have launched, leveraging AI to provide users with curated content: Lemon8 and Artifact. ByteDance, the parent company of TikTok, launched Lemon8 in the US and UK markets last month. The app closely resembles the Chinese social e-commerce platform Xiaohongshu, which first launched in Japan back in 2020.
Lemon8 is a social media platform that combines aspects of TikTok and Pinterest, with a focus on e-commerce. Users are prompted to choose verticals that interest them, such as fashion, beauty, food, fitness, and travel, upon creating an account and selecting a username. The app features a TikTok-style feed that separates content into "For You" and "Following" sections, allowing users to enjoy personalized content without requiring their friends to be on the platform.
While Lemon8 taps into the desire to go viral on social media, it has yet to fully master the flywheel of inspiring users to post content and encouraging others to do the same. However, if the platform can create a unique use case for content, it may attract more users seeking a differentiated social media experience.
Artifact is an app launched on February 22nd that uses AI to personalize news and written content for users. It groups articles by topic, providing users with a personalized and diverse feed. Although users cannot see what their friends are reading, they can upload their contacts to see which articles are popular within their network. The app is similar to China's Toutiao, which uses algorithms and machine learning to curate news for each user. Unlike other social media platforms, Artifact does not rely on creators to make content because it can pull articles from across the web.
Artifact, Lemon8, and TikTok provide value to users through personalized, AI recommendations that ensure the user immediately finds value in the experience. This is different from social media apps like Instagram and Facebook, which rely on the user having friends and possibly "friending" people they do not really know for more diverse content.
Artifact uses a transformer, which is the "T" in GPT-3 or 3.5, a deep learning model that adopts the mechanism of self-attention, differentially weighing the significance of each part of the input data. This technology creates a market for new applications to emerge that may, in some way, resemble their predecessors but differentiate themselves by how they leverage this technology. In this regard, some might say Artifact is similar to Twitter. As people use Artifact, users may be able to follow what others are reading, have their own curated news (similar to a custom version of The Skim or Morning Brew), and, if there’s an eventual option to post journalism content on Artifact, it may help journalists reach their audience more directly.
Diem is an example of a next-generation social media platform leveraging AI for a niche group of users. Building a social search engine for women using LLMs, it is a reimagined version of Reddit designed for the way women communicate. They trained a model that can help answer questions about birth control, career changes, money, relationships, and other personal topics. The more Diem is used, the smarter the model becomes, making it curated for a group of users.
Evaluating next-generation social media platforms:
Lemon8, Artifact, and Diem are all examples of next-generation social media platforms that provide personalized experiences with an emphasis on niche interest areas. Launching a vertically-focused social media application that prioritizes personalization is an exciting opportunity with increasing opportunity due to advances in machine learning. Here are a few keys to success for launching a new social media network:
Content needs to feel personalized: Users need to find content that is tailored to them without relying on their friends already being on the application. This can be delivered through a machine learning algorithm and a "for you" page.
The platform should provide a differentiated use case: New social media companies need a unique use case that can generate virality. Post things that users won't post anywhere else. If a user decides to post instead of just consuming, the platform needs to make the user feel special or like they can't post their content anywhere else. For example, people post professional content on LinkedIn. Nextdoor became the default platform to post content to share with neighbors. As Artifact serves users content curated for their interests, it may become the best place for journalists to post content so that it's distributed to their target readers.
A core utility should be provided or solved: To differentiate from existing "giants," vertical social networks need to provide a core utility. Strava utilized GPS data and wearables to automate logging workouts, creating a way to meaningfully connect fitness communities through a social application.
The platform should create “stars” or enhance users’ status: Being a YouTube creator is one of the most sought-after careers for kids today, and it's increasingly difficult on existing platforms. During the pandemic, TikTok gave rise to a new group of social media creators. A new platform needs to help its earliest adopters stand out and become key creators for the platform. Lemon8 is currently doing this by paying creators that are early adopters.
Creators should have a path to monetization: YouTube is easily the best monetization platform for creators. It is straightforward and creators receive direct ad payments. To maintain a supply of content, the platform should have a way to provide monetization for creators.
The platform should serve as a system of record, creating a sticky product: The platform needs to become a system of record within a vertical. For example, I check my connections on LinkedIn to see who I know in my professional network in a specific city. Queue allows users to track their movies and TV shows, serving as a system of record for the next show I choose to watch. Facebook started as an online directory for college students. Today, it remains a system of record for birthdays.
Founders should have a unique GTM and distribution strategy: New social media platforms need to find unique avenues to reach users. Partiful reaches new users when a user plans an event and invites their friends to RSVP. Even if the content is sticky and the platform solves a core utility, how will consumers hear about it? Why will they make the first decision to come?
It needs to create joy for users: Most importantly, new social media companies need to bring joy and happiness to their users. For example, users find TikTok funny, and its videos make them laugh. BeReal is light-hearted and feels "authentic."
There are increasing opportunities for vertical social platforms that combine commerce or unique content creation with a social graph. Twitch built a vertical social platform around gaming. Similar, highly curated social media platforms could be built around other trends such as comedy shows, sustainability, clean beauty, travel, DIY, career tips, and education, or any hashtag trending on TikTok today.
If you’re building a next generation social media company, I’d love to chat with you. You can reach me at maria@madrona.com. Thanks to my colleagues at Madrona for their feedback on this post!
Additional resources:
https://www.nbcnews.com/tech/tech-news/gen-z-facebook-pew-research-center-finds-rcna42429
meta/https://www.adweek.com/media/views-on-instagram-reels-fall-as-platform-course-corrects-to-images/?
https://www.vice.com/en/article/y3pa5b/tiktok-parent-launches-xiaohongshu-competitor-lemon8-in-us
https://www.collectivemedia.info/
https://futureparty.com/stories/Paid-Twitch-Streamers-Money-Burnout