But will we ever? That’s a philosophical and scientific question with no easy answer: in the meantime, programmers and businesses alike are working tirelessly to find the limits of AI by automating high-level tasks like content writing and customer support. So far, they’ve had some success.
AI-Generated Content
It will come as a surprise to many that they have been reading content generated by AI for years. For example, many of the popular “explainer” videos shared across social media are written and produced with little human input by commercial software like Lumen5.
Thanks to breakthroughs in programming and research, AI has come a long way in the last few years, and it can do more than generate a few sentences for marketing videos: Google’s Duplex can even make phone calls and set appointments with believable English.
If AI can do all that, it stands to reason that it can also write blog posts, news articles and other long-form content for your business – right? The answer is: not yet. While there are many ways AI can assist with those tasks and cut down on the labor required by a content strategy, it’s not good at writing and may never be.
Expectations
In 2011, IBM shocked the world when its robot Watson managed to win the gameshow Jeopardy by listening to human questions and answering them correctly. Watson represents an ongoing revolution in Natural Language Processing (NLP) that has continued ever since.
But the confidence inspired by AI breakthroughs is often unrealistic. Research firm Gartner – famous for tracking the “hype cycle” of sensationalized technologies – hyped the growth of AI itself, predicting in 2015 that 20% of business content and 85% of customer interactions would be handled by bots in 2020.
No follow-up research shows whether these benchmarks were reached, and that’s good reason to believe they haven’t been. Accordingly, Gartner has wavered on its AI stance since 2015. To date there are almost no commercially available tools for generating long-form content, and those which exist have serious limitations.
Reality
While algorithms have found success in limited domains of content generation, their ability to replicate natural language remains rudimentary at best. Brands like Coca-Cola have experimented with AI-generated advertisements, but so far there are no compelling case studies of the results, and excitement has died down significantly.
At the cutting edge of NLP research, GPT-2 has successfully demonstrated the ability to expound on a topic using passable sentences and paragraphs. When asked to answer a prompt about scientists discovering unicorns, the algorithm churned out this excerpt after several attempts:
Dr. Pérez believes that the unicorns may have originated in Argentina, where the animals were believed to be descendants of a lost race of people who lived there before the arrival of humans in those parts of South America.
While their origins are still unclear, some believe that perhaps the creatures were created when a human and a unicorn met each other in a time before human civilization. According to Pérez, ‘In South America, such incidents seem to be quite common.’
To scientists, this is exciting stuff: but for businesses and professional publications, it’s not very promising. On the one hand, GPT-2’s ability to follow conventions of the written word and tell a semi-coherent story is impressive. On the other hand, it tends to go down rabbit trails, producing what is obvious nonsense to a human reader.
What AI Is Good For
So far, it’s clear that AI won’t be generating entire blog posts, whitepapers or insightful news coverage any time soon. Nevertheless, AI’s capabilities can assist with the most repetitive parts of a content strategy, and currently that is where it shines the brightest.
When it comes to language, AI works best at converting numbers into words. Newspapers – including Washington Post, The Guardian and Forbes – have used in-house algorithms to generate articles limited to finance, local sports coverage, and other stories heavily centered on data and straightforward reporting.
For small-to-medium sized businesses, commercially available AI solutions can be used to churn out ideas, research and content summaries. While not publishable on its own, this material can easily be converted into good content with editorial input, leaving only the most interesting parts of content creation to human writers.
How You Can Use AI Right Now
Many of the tools used by major publications – including WaPo’s Heliograf and Bloomberg’s Cyborg – are proprietary, and not available for commercial use. Fortunately, there are other AI-powered solutions that businesses can use right now to assist with content creation.
Here are some examples, according to application:
For now, those who expect AI to completely take over content creation will either be disappointed or pleasantly surprised by its continued inability to do so. But – as with most applications that involve intelligence and creative thought – we can expect it to get better at assisting human labor as time goes by, and that is something which many can look forward to.
Conveyance Marketing Group is a team of bright, innovative and talented veteran marketers dedicated to big ideas, fresh insights and measurable results. We pride ourselves on taking challenging marketing issues and turning them into opportunities for our clients, on pointing brands in the right direction, and on getting our customers noticed both online and off. From branding to websites to digital marketing, and public relations, we handle all your marketing communication needs! Web Design and Development | Brand Strategy | Inbound Marketing | Social Media | SEO | PR