Photo Credit: shutterstock
In May, tech giant Microsoft made headlines with its decision to get rid of human journalists in favor of automated bots in order to curate content for its MSN website. Quite simply, Microsoft determined bots might be better than humans when it comes to curating content, and they’re also a heck of a lot cheaper. You can force a bot to work overtime and not hear any complaints or demands for extra pay. And now comes news that bots may actually be advancing to the point where they can not only curate content – they can also create content. An AI-powered language program known as GPT-3 recently was used to create a super-viral blog post that shot up to the #1 spot on Hacker News. So is it really any surprise that the future of social media content is the bot?
Early signs of a bot content revolution
Some media organizations have already played around with bots for content creation purposes. For example, sports is one area where bots can churn out nearly the same content as their human journalist counterparts. Consider the average baseball box score. It’s easy to see how a reasonably smart AI program could churn out something like the following, just by parsing the numbers in a baseball box score: “The Philadelphia Phillies won today, by a score of 4 to 2. Bryce Harper hit a homerun for the Phillies in the bottom of the fifth inning. Aaron Nola pitched 7 innings and was the winning pitcher. The following hitters all had two hits in the game…”
And it’s the same story with other forms of content creation. Think about your average daily weather forecast. “Today in the Philadelphia area, it will be cloudy with a 20% chance of rain. The high will be 80. The low will be 65. Sunset tonight will be at 8:02 pm.” Do you really need a human to tell you that? All you really need is access to the underlying meteorological data, and a bot can do the rest. If you think about it, most of the mindless dreck you find on websites BuzzFeed these days can probably be auto-generated by a bot. Couldn’t a bot generate a post about “The Top 10 Most Popular Items on Amazon You Need to Buy Right Now”? As long as you have access to all the back-end data (customer reviews, sales rankings, etc.), this seems easy enough to do.
The GPT-3 blog writing program
The AI-powered language program GPT-3 represents the next stage in this bot content revolution. Simply put, GPT-3 uses Deep Learning techniques to look for patterns in human-created text. If the program can train on a sufficiently large body of text (such as all the text in the U.S. Library of Congress), then it can look for statistical patterns in this text. It can start to figure out which words are likely to follow other words, and how humans put together words into sentences and paragraphs. It’s like the old joke about monkeys and typewriters: put enough monkeys on enough typewriters, and one of them will eventually bang out Shakespeare.
At this point, GPT-3 is able to complete your text based on a few prompts from a human writer. With just a headline prompt and maybe a sample first sentence or two, it can auto-generate a full blog post. This is exactly what happened with the viral blog post for Hacker News. The headline prompt was simple: “Feeling Unproductive? Maybe You Should Stop Overthinking.” With that as a prompt, GPT-3 produced a fairly convincing article about boosting productivity. In just a few hours, the post went viral (after all, who doesn’t want to become more productive?), with only a handful of those readers even guessing that the post was auto-generated by a bot.
Pros and cons of bot content
Even better, GPT-3 can even generate a few versions of the same post, and you can choose which one you like best. Imagine telling a blogger – “Hey, write five versions of the same article for me, and I’ll choose which one I like best…” This would be a goldmine for content marketers, who are forever A/B testing various headlines and story concepts to see what goes viral and gets engagement.
The downside to all this, of course, is that the content that gets generated can be pretty banal. The AI-powered post was pretty impressive from a grammatical, logical and syntactical perspective – but as one reader complained, it provided “zero substantive content” and was “pure regurgitation” of some other article found online. If you’re running a marketing campaign, just how effective are five plagiarized versions of the same boring article really going to be? And do people really want to read mindless, non-substantive content all day and all night?
The bot future
The future is clearly bot-writing machines or robots. All the early signs are here. In order to reduce the cost of creating content, media organizations are going to use them. And marketers, always looking for a way to squeeze out a little extra ROI, are going to experiment with them. So it’s natural to suppose that after all the initial hubbub about journalists and marketers losing their white-collar jobs dies down, bots will be everywhere. Anything you read online will either be generated by a bot or curated by a bot, and then delivered to your social media news feed by a bot of your choice.