Social media is indisputably a fundamental part of modern life. People can still get by without it, but almost every facet—from posting about one’s day to watching seconds-long videos—occurs on platforms like Facebook, Twitter, and TikTok. The latest numbers show that close to three out of five people spend around two and a half hours browsing social media every day. (1)
To put those numbers into perspective, the number of people on social media is over three times greater than the population of China. No doubt that’ll entail billions more in posts and videos every day, too many for human administrators to screen. As a result, social media companies have recently invested in developing artificial intelligence (AI), specifically machine learning (ML).
AI is taking a transformative role in social media. Here’s a closer look at how it operates and how platforms harness it to improve user experience.
Manual Then, Digital Now
One paper published in the International Journal of Business Analytics & Intelligence suggests that AI development is a sign that the world’s transitioning from manual to digital. While some tasks are still best done by humans, repeating the task could result in a different result. Having to review every piece of content before approving it for public viewing would take a lot of time and resources.
Such an issue isn’t lost on social media platforms, all of which are dealing with data that’s constantly getting bigger and more complex. Back then, simple data like the number of active users at a given time sufficed. Today, the data includes more specific parameters such as the type of posts users want to see in their feed and the type of words they use when posting or leaving comments.
Given this, the focus on AI in social media involves developing ML models that continuously learn while going through complex data. When the AI knows that a particular user fancies cat videos, it’ll suggest similar ones on their feed. The more suggested videos the user watches, the better the AI can generate further recommendations.
As autonomous as such models may appear, they still require codes inputted with human hands. The heart of any ML model entails using a data labeling platform to correctly identify elements in any form of media and reiterating those labels multiple times. The quality of the data it feeds on also matters; giving a model the wrong data can cause the entire system to fail.
Social media platforms harness AI a bit differently from one another, given that they cater to a specific type of media more than others. Despite this, the reality is that AI forms the cornerstone of the platforms’ functions.
A recent Statista report found that link and image posts made up nearly 80% of all posts on Facebook in 2020. It makes sense for the platform to develop its ML models around labeling and iterating elements in images and text on links or captions. (2)
Two models come to mind: DeepText and DeepFace. These two AI use multiple neural networks to identify keywords in post captions or personal messages and specific objects in photos. They’re impressively accurate, too; DeepFace achieved a 97.25% accuracy rate, beating facial recognition tech used by law enforcement by a wide margin.
Despite Meta announcing its plans to shut down its facial recognition feature last year, its decade-long existence is a testament to the wonders of AI. Perhaps the move is part of its long-term goal of establishing the AI-centric “metaverse.”
Twitter’s new owner, Elon Musk, is bullish about AI and its role in the platform. After the USD$44-billion buyout, one of the first plans he announced included developing a policing algorithm to root out the endemic proliferation of hate speech. Whether or not it’ll work, it’s still too early to tell. (3)
The core of Twitter’s AI is Watson, a system that works on natural language processing. By answering any questions asked as a human would, Watson can detect variables in a Tweet like its tone and determine if it’s offensive to readers. (4)
TikTok has the distinction of being one of the latest platforms to be nearly fully managed by AI. Its model hinges on two separate systems: one for the content creator and another for the viewer. Combining the data from these systems culminate in human intelligence mining, studying user behavior instead of their preferences for recommending videos. (5)
This unique setup, among other things, fueled TikTok’s rapid ascent in the social media industry. The app has been downloaded more than three billion times in app stores and has logged over one billion users monthly within six years.
These days, social media without a working AI/ML model is almost unheard of. Data is as complex as it gets and will only grow more complicated in the future, so platforms old and new would do well to harness and refine AI to get a better glimpse of their user base and improve their services.
- “Global social media statistics research summary 2022”, Source: https://www.smartinsights.com/social-media-marketing/social-media-strategy/new-global-social-media-research/
- “Leading types of branded posts on Facebook worldwide from 2014 to 2020”, Source: https://www.statista.com/statistics/296628/facebook-brand-post-interaction/
- “Elon Musk’s plan for an open-source algorithm won’t solve Twitter’s problems”, Source: https://techmonitor.ai/technology/ai-and-automation/open-source-twitter-algorithm-elon-musk
- “How are social media platforms using AI?” Source: https://www.marketingaiinstitute.com/blog/ai-for-social-media
- “What is TikTok and How is AI Making it Tick?” Source: https://www.analyticssteps.com/blogs/how-artificial-intelligence-ai-making-tiktok-tick