The Cambridge Analytica scandal showed us how Russian AI-powered fake news had the power to steer the U.S. 2016 presidential campaign. It is now a matter of fact that intelligent machines aren't the future of media and publishing, but the present. Although that last sentence may sound ominous, our future is not necessarily linked to a nightmare of fake news and social media managers stealing our private information, though. Artificial Intelligence, automation, machine learning and all the latest technology trends of the last few years are going to revolutionize our current scenario, and maybe, even in a better way.
Mainstream Robotic Reporters
Believe it or not, you've probably read news articles written entirely by a machine. Mainstream publishers have started using AI to write some of their stories for them. In fact, the Washington Post’s automated reporter published a whopping 850 articles in its first year using Heliograf. During the presidential elections, the robot reporter was smart enough to ping the newsroom every time results started trending in an unexpected direction, effectively assisting the human reporters in their jobs. Other AI applications have been successfully used by the New York Times, Reuters and other media giants to automate mundane tasks, streamline media workflows and crunch a lot of data. (Read about this and other AI uses in 5 Ways Companies May Want to Consider Using AI.)
Fake News and Manipulation of Information (AKA – "The Bad Stuff")
Did you know that a study from Stanford University showed that some AI is so smart at understanding human beings, that it can detect a person's sexual orientation with an 81 percent chance of success just by looking at one picture? And this deep neural network is so advanced that, when the number of pictures increases to five, the percentage of success becomes 91 percent. And sexuality is not the only parameter that this breathtaking AI could guess just by looking at some random Instagram photo. Emotions, IQ, and even political preferences can be understood by this machine which is able to detect things that no human could even imagine.
Once again, if you think that this technology may be the future of facial recognition, well, you're wrong: This amazing discovery actually is a thing of the past – albeit recent. The first thing that comes to mind is, "If this amazing stuff can make such accurate guesses from just a couple of pictures, how much data can be extracted from people by accessing their social media accounts?" A lot, apparently – so much that it looks like other similar AIs may have been used extensively for political reasons. They may very well be some of the reasons why Donald Trump is now the President of the United States and the Britons left the European Union through the Brexit.
AI-powered psychometric profiling is used to extract data from social media profiles, and use this info to show potential voters a specific subset of targeted fake news or political ads. The idea is to manipulate information to a degree where humans cannot understand what's true and what's not anymore. To put things in perspective, this technique is so effective that some allege that it has been used again in Italy as well, and with much less subtlety.
What's even worse is that AI not only helps to find the right target for the fake news, but it can actually generate fake news at a speed that no human writer could ever hope to achieve. It can automate the entire process of writing and spam millions of articles in just a few seconds.
AI can create absolutely believable fake videos and even alter what a person said, for example, during an interview. Or it can generate realistic, lifelike photographs from scratch that are absolutely indistinguishable from a real human being. And it's quite hard to understand what the truth is when you can't believe even your own eyes.
The Battle Against Fake News – The Other Side of The Coin
Don't despair, not all is lost. Some of the most powerful machine learning software is ready to be deployed to scour the web and detect all those blatant lies – starting with Google, whose News platform will now be able to filter out all that information that is determined to be misleading or just false. According to Google's spokespersons, the AI will draw data about the believability of the info from a certain range of trusted sources, and will also organize and separate content into news, opinion and analysis to help people know the difference between a fact and an opinion.
Other software is also either available or currently being developed, to estimate if the headline of an article accurately reflects the body of the article itself. This is incredibly helpful in weeding out all those scaremongering news articles that use misleading headlines to prey on the laziness of people who don't even open the article to read its content. In a nutshell, the idea is to push people away from extreme, deceitful content, and direct them to more reliable and unbiased articles. The goal is to stop driving people into making emotional rather than rational choices.
The Introduction of AI in Broadcast and Media
It may be argued that broadcast is one of those technologies that still survives only thanks to its wide popularity in previous decades, even though it is now becoming obsolete in a lot of ways. AI adoption may help in regenerating this sector, although the process is still at an early stage. Up to 56 percent of media technology buyers said they were likely to adopt it in the next 2-3 years.
Netflix, for example, is among those which have already employed the efficiency of AI in reducing the routine workload through automation. And the results are right in front of everyone's eyes (pun intended). The rapidly growing company claims it saved almost $1 billion per year, thanks also to the AI's ability to reduce customer churn. Machine-learning algorithms can draw data from social media and use it to establish a more personal relationship with viewers, which is particularly effective since we're talking about how the customer is going to spend his or her leisure time.
AI can also help with efficiently managing and organizing content, which has traditionally been a serious issue due to the unstructured nature of video and audio data. All the recent advances in speech and emotion recognition, as well as computer vision, empowered the most recent AI tools which can now easily classify archives that were previously thought of as inaccessible. Algorithms and automation can also be deployed to optimize and improve the efficiency of networks, which is a great boon for pay-TV operators who want to reduce their bandwidth issues in streaming services. (If AI continues to be implemented in various industries, how will humans earn a living? Check out Is the AI Revolution Going to Make Universal Income a Necessity?)
The Impact of AI on Academic Publishing
The academic world is, in many ways, a closed world. Siloed into a handful of ivory towers, the modern scholarly publishing ecosystem relies on the ability of the scholar to be able to conduct manual web searches just as if it were still 2001. Many of the advancements that have improved and refined search algorithms in the commercial world haven't reached the world of academic literature, which is also devoid of many of the minor discoveries that get disseminated through blogs, press releases and social media.
Take, for example, the "related work" of a common academic paper. The entire cross-section of the alleged contemporary developments of a given discipline is often minuscule and limited to an artificially circumscribed set of references of that specific subfield. Citations are all but comprehensive, and frequently scholars fail to understand how many other similar studies and work have been already published that, in truth, described the same thing (possibly using an even better method).
AI can, once again, help expand the reach of these searches, and include all these data subsets that humans simply have no hope of monitoring and digesting. Scientific figures can be "read" and described by machines with metadata structures that enable them to eventually sort, analyze and search them. Natural language processing (NLP) helps the AI understand the true nature of the paper and integrate data coming from external sources (company blogs, tech magazines, etc.) to compare it with other relevant studies, including those outside the original discipline.
Machine learning can employ automated statistical analyses to improve the peer review process, showing human reviewers sources they may have otherwise missed. The process of verifying citations is also streamlined as the AI can quickly help flag a quote that was incorrectly attributed to another article, or scour through an entire document in minutes to spot miscited quotes or plagiarized content. Even better, modern image assessment algorithms could easily detect any sign of image manipulation in biomedical journals.
For those wondering whether this article has been written by an AI, well, the answer is no. At least for now, human beings are still necessary. And with due probability, humans will never be substituted by robots in publishing and media since creativity and art are a fundamental and irreplaceable part of writing. In truth, as we human writers will be assisted by AI, our jobs will become easier and the average quality of our products even better.