Social media and AI have recently been all over the news. Executives from Twitter and Facebook were hauled into...
the U.S. Congress to testify about foreign influence on their networks caused by bots and other malicious users. Recently, California passed a bill banning bots for telemarketing and election-related calls. Increasingly, people are starting to worry and wonder about how bots are infiltrating social media networks and using them maliciously.
However, there are many additional and less sensational ways in which AI and machine learning are already applied to social media. Many of these applications are proving to be helpful and are making these networks more valuable for end users, customers and technology partners.
AI-enhanced content optimization and moderation
Social media is filled with lots and lots of content, and most of it is unstructured, meaning that it doesn't fit into databases or rows and columns on a spreadsheet. In particular, lots of text, images, video and audio content is posted daily. According to Domo, over 2.5 quintillion bytes of data are created daily. There are more than 456,000 new tweets created per minute and over 1.47 billion logins a day on the Facebook platform, with millions of new posts created every minute. That's an awful lot of content to manage, enhance, connect and moderate. Fortunately, machine learning is being applied to assist.
Machine learning is being actively used at Facebook, Twitter, LinkedIn, Google and all the other major platforms to provide information such as which content is most relevant to display in feeds or recommend to users. Machine learning approaches are also being used to make content recommendations and help users filter through the millions of posts on the platform daily. These machine learning-based systems are able to determine patterns of content that users find most valuable and promote advertising, posts, images and other content that would be of most interest to them.
In addition, AI is being utilized to add value to audio, images, video and text data. Machine learning is able to automatically identify faces in images and tag people using facial recognition technology. Images are enhanced with automatic captions and other information, utilizing machine learning-based summarization technology. The same technology is being used to automatically identify objects and people in videos and provide automatic text captioning using natural language processing (NLP) capabilities.
In another area of social media and AI, the platforms are using natural language technology to automatically transcribe audio files, messages and voicemail. In addition, these platforms are making use of machine learning to provide advanced machine translation, enabling users in one language to understand posts in another. Social media platforms are making use of a wide range of machine learning-enabled capabilities to handle all sorts of unstructured data requirements.
Most of the platforms are using machine learning-enabled content filtering and moderation to make sure content that violates terms of service or other guidelines are not shared on the platform. In most parts of the world, this content moderation is mainly focused on acceptable content. However, in China, this automatic moderation is applied to content that runs afoul of censorship guidelines. The power to moderate is therefore immensely potent, and the use of machine learning is being applied in a wide range of contexts from harmless to blunt.
Improved marketing automation
End-user companies and advertisers also benefit from the AI and machine learning on these platforms. A number of marketing automation tools are using AI to better identify and profile users that could make an ideal audience for the offerings of the advertisers. AI systems are good at finding relationships among users that could identify affinity for a brand, product or candidate, and these relationships can then be used for improved advertising targeting.
AI systems are also being used to create and schedule social media content. Advanced marketing automation tools can automatically identify the best locations and times to post and even suggest content. Some tools are able to generate text, image or video content automatically from blog posts or other sources to be repurposed for social media.
These tools are also being used to gather additional information on customers and visitors to customer sites. Rather than only using the information provided by the social media platform or browser, AI tools are enhancing data by using machine learning to gain insights from social and customer data. This includes finding posts from the users on other platforms, identifying similar posts on the same platform or merging social media data with information from third parties. Similarly, these platforms are being used to find influencers that have the most direct impact on a company's brand. Then, the marketing automation tools can better target and focus on influencing these influencers.
Unexpected ways social media and AI combine
In addition to these bread-and-butter applications of AI and machine learning on social media platforms, the technology is being used to help solve larger problems in society. The University of Helsinki is using AI and social media to help catch poachers and stem the tide of illegal wildlife trade. Researchers noted that wildlife dealers were active on social media, posting photos and information about their products for their customers. The university designed a system that uses computer vision and NLP to automatically comb through social media posts to identify images, text descriptions and other content associated with illegal wildlife trade. The image detection software can identify specific products, such as elephant tusk ivory or rhinoceros horns. The audio can identify endangered bird calls, and the NLP systems can identify animals, as well as locations where poaching might be happening.
In another helpful use case, Facebook is using NLP and machine language understanding to help detect when its users might be having suicidal thoughts. Added in 2017, the system automatically identifies and flags posts that exhibit expressions of suicidal thoughts. Previously, Facebook would depend on users flagging suspected posts, but now the automated system is detecting at least 20 times more cases of suicidal thoughts than human notifications. These posts are then analyzed by human reviewers from the company's compassion team. In just its first month of operation, the company has already seen over 100 wellness checks on its users performed by the system.
It's clear that social media is making increasing use of AI, and users are making increasing use of AI to augment their social media activities. There's no doubt that the combination of social media and AI will grow more intelligent in the future.