Photo Credit: pexels
Heading into 2020, it looked like artificial intelligence (AI) was going to be the future of social media. Everywhere you looked, AI was being integrated into the everyday workings of social media. Most notably, Facebook was using AI not just for image recognition in photo posts, but also for helping to determine what items appeared in your newsfeed in the first place. And AI was being quickly integrated into sentiment analysis tools, helping brands track what consumers were saying about them in real-time on social media. But then came the great coronavirus pandemic of 2020.
Problems with image recognition and visual AI
Until the pandemic hit in full force in March 2020, photo recognition tools were becoming remarkably advanced. Facebook, for example, could easily identify your friends and family members in posts, making it much easier to boost engagement and attention on the platform. If you’ve been “tagged” in a photo, aren’t you going to share it with others or comment on the post itself?
But all of this assumed that your friends, family members and neighbors weren’t going to start wearing masks any time they dared to venture out into public. Suddenly, visual AI tools trained on years of photos were having a hard time figuring out who was who. After all, the reason why protesters and criminals have traditionally worn masks was to obscure their identities. But now even your aunts and uncles were wearing masks, if not full-on balaclavas and elaborate headgear designed to protect every inch of their head from a droplet of coronavirus-infected water.
As a result, the whole coronavirus pandemic has made it harder to “train” AI systems on photos. Until this year, if you saw a masked driver in a car, you’d assume one of two things: this was a getaway driver for a bank robbery or it was someone headed to a ski resort for a day on the slopes. But now AI systems need to be trained that images of masked drivers might be your neighbors headed to the local supermarket. And just think about the “tags” used to identify photos in social media posts. Suddenly, the coronavirus pandemic has completely altered our social and cultural patterns. Previously, if you saw a photo of a person sanitizing every square inch of a perfectly clean and uncluttered room, you’d probably think the person had some form of OCD. Now, however, you’d have to re-think what that image means. It’s the new normal, folks.
Newsfeed algorithms thrown for a loop
And don’t forget about the sophisticated AI-powered algorithms used to streamline your social media newsfeeds. These algorithms need to be re-trained to look out for disinformation and misinformation about the coronavirus, and to prioritize content that people need to make sense of the rapidly changing world around them. With the sudden turn of events, human editors might actually be better than computer editors. Posts that might have earned a lot of engagement in the past – such as posts about new movie releases or stories about hot new vacation destinations – all of a sudden have little to no engagement value if people aren’t going to movie cinemas or boarding airplanes any longer.
Not to mention the fact that AI-powered algorithms now need to make sense of a whole new lexicon – such as “social distancing” and “lockdown” – in order to churn out the right content for newsfeeds. Just think about the potential for machines to misinterpret a term like “social distancing.” If machines have been trained for years to respect the social aspect of humans, and to do everything possible to boost social engagement and close social ties, doesn’t the whole concept of “social distancing” sound very foreign and even alien? Humans are social creatures, after all. In the past, images of “social” humans congregating in stadiums, churches, gyms or schools would have seemed perfectly normal. Now, these same images would need to be re-interpreted as “anti-social” humans trying to flout official CDC and WHO guidelines and bringing unimaginable risk to their local communities.
A small setback or a long-term trend?
Before this summer, the consensus opinion was that the pandemic would probably run its course within a few months. But now, it looks like the pandemic is here to stay for the rest of the year. Businesses are only slowly re-opening (and then closing, and then re-opening), the situation around schools and universities is still very much in a state of flux, and just about everyone these days realizes that nothing is going to change until 2021. That has huge implications for social media companies, which have invested so much time and money into building AI-powered algorithms. AI is struggling to adapt to 2020, and as a result, companies are working to scrub their datasets and re-train their algorithms. AI is still the future of social media, but it is a very different future than originally envisioned.