Social media users are getting a powerful tool with PostIntelligence, an artificial intelligence (AI) app that launched its desktop and Android beta (iOS will follow soon) today. The AI, which currently supports Twitter, Facebook and Pinterest, promises to help users become much better at posting engaging content and grow their followers.
PostIntelligence CEO, Bindu Reddy, has a great deal of experience with social networks. She was the former head of product for Google’s social apps, and she launched the anonymous social platform, Candid last year. With Candid, AI was used to moderate hate speech and make sure that posts were relevant to the topics. That’s what helped spark an interest in artificial intelligence and its application toward social media.
“There are aspects to artificial intelligence that are very interesting, especially when it comes to social media,” Reddy told [a]listdaily. “What I mean by that is that—to date—digital assistants like Siri and Alexa let you perform easy tasks, but in an automated fashion. These include things like turning on the lights, which is fairly easy for humans to do. But what fascinates me, and what I’ve been working on, is seeing how AI can help humans do tasks that are difficult. What are the salient things that machines can do that humans can’t? That way, they can be complementary to humans, as opposed to replacing them.”
Reddy also worked on a social media advertising company called MyLikes, which led her to discover a major problem with social media.
“The most difficult thing about social media is that it’s very hard to grow a presence and get engagement, especially for normal users,” she said. “If you want to share things with ten friends, that’s fine, but if you want to have a social media presence and grow your followers by sharing engaging stuff, that’s far more difficult.”
Reddy then worked with social media influencers, who talked about how hard it was to maintain a strong social media presence. That’s where PostIntelligence arose from.
“We’re dubbing this the world’s first AI-based social media assistant,” said Reddy. “It helps you get engagement with great content suggestions and will help predict how your posts will do. In time, it will help you grow your following.”
According to Reddy, the AI assists users by creating “a deep learning model for every person that signs up by looking at all the things that they’ve posted before. The PostIntelligence AI then learns about the kinds of things you like to post about and understands what your audience wants. It then uses this model to make posting recommendations that are likely to have high engagement.”
Reddy continued by explaining how PostIntelligence works, saying “for a normal consumer, it helps make Twitter better by helping them to tweet better. One of the things that we realized is that tweeting is difficult for normal individuals. It takes a lot of time and energy to be up-to-date, understand what to tweet about, and then come up with something. This is a supplementary app that sits alongside Facebook or Twitter and helps you figure out what to say and how to engage people. It helps you discover yourself in some ways. It’s not supposed replace your original voice; it makes it better.”
Instead of having to scour the web and Twitter for interesting news and trending stories, PostIntelligence does all the work for you and lists recommendations based on your interests. As proof, Reddy stated that her Twitter engagement levels have already gone up significantly because of the AI. “Part of it is that adding good media makes a tweet more interesting, and the other part is that it keeps users up-to-date with what’s going on,” she said. In fact, the recommendations could become a source for the latest news for many users.
After users select a recommendation for posting, PostIntelligence can also analyze the written message, predicting the level of interest and engagement they might get from audiences based on the wording. It also helps users schedule posts for specific times to help get highest impact, and it supports multiple account management. In just a few minutes, users can set up a series of highly optimized tweets.
Additionally, the AI detects when you’re sharing media, such as animated GIFs, images, emoji and videos, then tells you what type of media does better. However, it cannot yet detect what type of image it is, meaning that PostIntelligence can’t predict whether that picture of Michelle Obama will resonate well with your audience, but Reddy assures us that it’s being worked on. What it can do well is work with links by extracting descriptions and titles from them to figure out what keywords and sources work best.
Reddy also talked about how her work with the social media influencer platform, MyLikes, impacted the development of PostIntelligence’s sponsored recommendations.
“We mostly hooked up social media celebrities—people with 10,000 or more followers—with brands and sponsorships. What we learned over time was that these sponsorship opportunities tend to be from brands like Uber or Airbnb. But when people posted sponsored stuff on Twitter or Facebook, most people reacted negatively. So, what we realized was that instead of having them talk about a brand, it’s better to give them relevant content recommendations, which is something their audiences would enjoy.”
She then went deeper into explaining. “If I were an influencer, it would be very weird for me to suddenly say, ‘go sign up for Airbnb.’ Instead, it would be very relevant if I tweeted a tech article. What we’ve done is partner with around 30 media companies, including Simplemost and Huffington Post, to create content and PostIntelligence recommends what should be shared on your newsfeed. When you share it, you can be paid for it. It’s a sponsored post, but it’s something that’s very relevant.
“This is social advertising done right. It’s not about posting on your feed to tell people to go buy this or download that. It’s about posting something that’s interesting and relevant to your audience, but at the same time, is very valuable to the company that you’re driving traffic to. Brands want to write compelling stories, which people can share and have it potentially go viral on social media, and this is a really good platform for them to amplify that branded content.”
The AI also makes suggestions based on trending topics. For example, if Apple or Airbnb are trending in the news, PostIntelligence will suggest posting about them. Reddy explained that it’s very different from Twitter Trends because it’s specific to your interests. “It’s basically saying that these are the things that you should start conversations about, which is relevant to your audience and to your account,” she said.
“We are hoping that the analytics will help you find out what your profile looks like,” Reddy continued. “This will also be useful to brands and people who manage multiple accounts. We’re starting with individuals, and we’re hoping that, in the future, brands engage in this. It’s much better than the tools that they currently use to manage multiple accounts. This is not just an ability for you to post or schedule stuff, it’s also the ability to help you get better at what you’re doing.”
Perhaps the biggest challenge for a social media AI assistant is discerning genuine interest from sarcasm, irony and humor.
“This is something that I get asked about a lot,” Reddy said. “The answer is not easy, but it’s very difficult to convey humor and sarcasm in general when you’re online. You could do it if you’re on a livestream or video, but just using words is more difficult.”
People who know you might know when you’re joking, but there are plenty of others who would not. By analyzing language and detecting things like double negatives, the AI will try to figure sophisticated human communication. “Are we at the stage where it gets that right all the time? No, we are not,” said Reddy. “Almost all the stuff that you’re seeing today, and the buzz about AI, is about baby steps. People are excited by these baby steps.”
Many experts believe that AI may someday fully understand human communication, but there are just as many skeptics who say that tone and body language will be extremely difficult for a machine to interpret. In either case, it looks like interpreting meaning and intent will still be a human job, and Reddy is content with that.
“To me, it’s about how AI can help and complement you with doing your job as opposed to having it replace you,” said Reddy. “PostIntelligence is not meant to take away your voice. We want you to continue saying what you want to say. I don’t know that, as humans, we want AI to replace us. As it is, PostIntelligence gets me the data I want and it’s still me commenting on the link or topic.”