Photo by Alex Knight on Unsplash
About a month ago, Yandex (a Russian search engine) announced that AI developed by their programmers wrote a bunch of songs in Kurt Cobain’s style. Which means that AI blogging should not be that far away, right?
By hosting an active blog, businesses can increase sales leads by 67% and are thirteen times more likely to enjoy positive ROI through improved SEO, greater brand recognition and higher visitor numbers. However, according to our own data, fewer than one in twenty small businesses do have blogs.
There are many reasons why a person running a business might not run a blog. Sometimes they lack the time or the skills to produce posts. They could outsource the entire process, but paying for high-quality original content is expensive. Many small businesses simply let blogs fall by the wayside. But what if you could provide your clients with an alternative to costly and time-consuming new content? Artificial intelligence is disrupting many models of business activity that had been firmly fixed for decades, and it’s coming to the content generation as well.
We did some research to investigate how soon you could expand your service offering with this new possibility, and which startups you should look out for in this field.
What is Artificial Intelligence?Artificial Intelligence, or AI, refers to the ability of machines or software to carry out tasks and processes that are seen as requiring human intelligence, such as creative tasks, learning and responding to environmental data and problem solving. AI has been around as a concept for hundreds of years, but it’s been developed as a serious possibility only within the last decades, with major breakthroughs happening in recent years. It’s been assumed for a long time that AI cannot produce a robot that can do anything involving creativity or unpredictable information input such as writing, painting or holding a human conversation. But this is changing now.
How Advanced is Artificial Intelligence Development Today?One major test of the development of AI is called the Turing Test, which requires a human judge to determine whether he is holding a conversation with another human or with a robot. In 2014, a robot won that test for the first time (although later examples of conversation with the same robot were highly unimpressive, to say the least). A different version of the Turing Test could be writing readable original content. Natural Language Generation software, or NLG, is the name given to the branch of AI that creates original content. NLG software uses a combination of a fixed writing template that governs the tone and angle of the robot writer and a structured dataset that is used to populate the text. NLG software can create a real article that most people cannot distinguish from an article on the same topic written by human journalists.
NLG software can create quite decent content when given the template and fed structured data.
Because the template is programmable by a human, AI-generated texts don’t all come out sounding the same; they’ll vary in writing style, tone and structure similar to a human writer. NLG software is thus pretty flexible for creating new content, but it’s still not able to act independently of human involvement. A real live person needs to set the template and prepare the data for the software to use.
“NLG software in its current state will not pull unstructured data and magically generate written text without human guidance.” Automated Insights
Photo by Damian Zaleski on Unsplash
What Implications Could AI Development Have for Digital Service Providers?Robot-generated content can be produced faster, more accurately and for less money than human-authored content. Research showed that when people were asked to judge texts written by human writers and those written by machines, artificially-generated news articles were ranked as being more ‘informative,’ ‘trustworthy’ and ‘objective.’ Human-authored content was still considered to be more ‘interesting,’ ‘coherent’ and ‘pleasant to read,’ but for how long will that outweigh fast and cost-effective?
Gartner predicts that, “By 2018, 20% of all business content will be authored by machines.” How does this impact on digital service providers and their SMB clients?
Saving money and time for SMBs in the finance sector:
Because software can scan, analyze and and identify trends in huge amounts of data faster than a human, artificial intelligence is highly suited to producing the data-driven reports and articles frequently needed for finance blogs. NLG software can generate blog posts that distill the most important findings from a mass of information and convey them in readable, comprehensible prose in a fraction of the time it would take a human.
This saves human writers from a monotonous, repetitive task that few enjoy, and saves SMB owners from having to pay for the far greater number of hours that it would take for a human to complete the same work. Once they (or you as a service provider) have set up the NLG software template to their satisfaction, you will be able to scale their production of content for their blogs. This is why AP has been using Wordsmith to produce quarterly earnings reports since 2014.
Photo by Markus Spiske on Unsplash
Generating content at scale for all small businesses
AI-generated content is already available and the cost of NLG software is decreasing all the time. While there is an initial outlay, once robot software is in place, it can produce content for just a few cents each time. Human input is still needed to set the template, which determines the style, voice and angle of NLG-generated content, and to send structured data to the program. Clients can then check the finished content and tweak the template so that it more accurately fits their message. Once customers are happy with the template, they can use it to produce personalized automated content at scale. Platforms like Wordsmith and Quill can be prepared, tested and produce thousands of pieces of content within a few hours..
Explaining data visualizations
Graphs, tables and infographics are all very important ways for small to medium businesses to present data to their customers, but often the topic is so complex that it can’t be easily understood from the visualization alone. Readers need accompanying text to explain what they are looking at and translate the representations before them into language that is more readily absorbed. Automated content can write these texts quickly and far more clearly than a human writer can.
The ongoing need for digital service providers in an AI world
The high capabilities of artificially generated content has made some content marketers nervous that their jobs will be the next to be taken over by robots. But this is not likely to happen any time soon. Human involvement is still needed to set the NLG templates which determine tone and story angle of AI-generated content, and to check that the content produced is in line with the company’s message. Content marketers will still be the ones to suggest what data is fed to the NLG software bots.
Furthermore, NLG-generated content is only suitable for highly data-driven fields. Blog posts that comprise financial reports, factual news reports and data visualization interpretations are ideal for robot writers, but not in-depth analysis pieces or emotive promotional blog posts. Marketing writing currently still needs human touch to produce content that moves readers and builds a relationship between consumers and the brand. For the time being, human writers will still be in demand to write emotive blog posts and in-depth research pieces, for example, while data-based articles could be handed over to the robots for cheap large-scale production. In the not-too-distant future, however, as the programs become more sophisticated, computers will be capable of producing emotional articles, a capability which many companies are already working tirelessly on producing.
Which Startups are Developing AI-Generated Content?There are already a number of startups which offer AI-generated content for business, marketing or journalism purposes.
Automated Insights develops a platform called Wordsmith. It has been in use for three years now. Wordsmith generates narratives in 20 languages using a combination of structured datasets and a pre-programmed template. Customers include Fortune 100 companies as well as media companies like AP, which have been using Wordsmith to produce their quarterly earnings reports since late 2014, increasing their coverage from 300 to 4400 articles per quarter.
Sample AI-generated Text from Wordsmith
Allergan Inc. (AGN) on Monday reported profit that climbed by 16 percent in its second quarter, and topped analysts' expectations.
The Irvine, California-based company said earnings increased to $417.2 million, or $1.37 per share, from $359.9 million, or $1.19 per share, in the same quarter a year ago.
From Automated Insights website, source
Wordsmith can generate content to fit any device, from screens to voice and text apps. Automated Insight’s public relations manager James Kotecki estimated that Wordsmith could produce 2000 articles per second. Wordsmith includes a publicly accessible API which enables clients to generate personalized content for millions of customers in real time and integrates with Zapier and Tableau.
“The standard way of creating content is, ‘I hope a million people read this.’ Our model is the inverse of that. We want to create a million pieces of content with one individual reading each copy.” Robbie Allen, Automated Insights, CEO
NLG startup Narrativa, with offices in Arab Emirates, Germany and Spain, launched their AI content generator, GabrieleAI, in 2015 to transform data sets into millions of human narratives per day via a programmed template. The intention is to finally make it possible to scale content writing. GabrieleAI uses NLG, enabling Narrativa to permit their customers to train the software to follow their writing style by customizing their example narratives, so this could be a real possibility for SMB owners needing to produce blog posts. Indeed, that is what Narrativa has in mind.
“We use Structured Data, so for example sports stats, product specifications, weather forecast data, KPIs, etc.,” said David Llorente, CEO of Narrativa, in a conversation with Camilyo. “We indeed use Narrativa to write blog posts... I see Natural Language Generation technologies contributing to content generation but not substituting totally the content generated by humans.”
Currently, they mainly produce news reports, product descriptions and financial earnings reports, but Narrativa’s long-term vision is to disrupt the digital content industry by providing content of every type and style at scale.
Sample of AI-generated text from Narrativa:
"Sevilla won with a resounding 2-0 against Athletic in Nervión, which sums up eight straight wins at home. Gameiro scored the first one for the locals and closed the scoreboard by converting a penalty kick after Kychowiak was fouled. Athletic was unlucky despite controlling ball possession, and wasn’t able to finish any of the numerous chances that they had.”
GabrieleAI promise. From Narrativa’s website, source
Quill was launched in 2014 by Narrative Science, a Chicago-based startup that grew out of journalism research at Northwestern University. Quill also uses advanced NLG to scan structured datasets to learn the important details hidden within and turn them into readable content. Forbes is just one major company that has been using Quill to produce their quarterly earnings reports. Using Quill permits companies of all sizes to scale their production of blog posts and other data-driven content that’s based on financial reports, news stories and investment research, as well as mining big data for relevant and actionable information points for blog posts.
“We found an unlimited need to help organizations take their data, figure out what is interesting and important, and then communicate that. “
“Quill transforms data into automated, human-sounding Intelligent Narratives that empower your people with insights to improve every aspect of your business.”
Sample of AI-generated text from Quill
“Over the past three months, the consensus estimate has sagged from $1.25. For the fiscal year, analysts are expecting earnings of $5.75 per share. A year after being $1.37 billion, analysts expect revenue to fall 1% year-over-year to $1.35 billion for the quarter. For the year, revenue is expected to come in at $5.93 billion.
A year-over-year drop in revenue in the fourth quarter broke a three-quarter streak of revenue increases.”
Quill at work from Narrative Science website,source
Founded in 2007 in France, Yseop offers two self-service AI software suites that enable enterprises to automate the process of transforming data into written content. Yseop’s NLG offering is the only self-service on-premise NLG software on the market. Their artificial intelligence NLG programs can create comprehensible content out of big data in English, French, Spanish, German and Japanese in real-time.
Yseop offers two separate AI platforms for content creation: Yseop Compose converts data into written reports that highlight implications and trends in finance, business intelligence and marketing sectors both for customers and for in-house study.
Yseop Savvy launched in 2016 as a self-service plug-in that integrates with Qlik, Excel and other dashboards or data visualization software to create accompanying text explanations that help make sense out of tables and graphs. The texts generated by both of these plugins could be used as blog posts for small businesses, particularly those in the finance sector.
“We believe that human beings don’t speak data, we speak English (or whatever language). In order for businesses to be able to make data driven decisions at every level, they need data explained to them in written form, in their native language, and in real time.” Yseop, About Us
For now, Yseop offers its help to marketers with data analysis and reports. Is content marketing the next move? From Yseop’s website, source
What Does Google Say About AI-Generated Content?Since 2012, Google has made it clear that they will penalize websites that use auto-generated content. Since we’ve had this rule well drummed into us for over five years, it’s not surprising that one of our biggest concerns as marketers with artificially generated content is that it might be penalized in the same way.
But the reasons that Google gave for penalizing automated content in 2012 - that it is unoriginal, low value and nonsensical - do not apply to today’s NLG-generated content. Automated Insights, creators of Wordsmith, reassure clients that robot-authored content that is original, high value and written for humans is going to be indistinguishable from content that was written by humans.
“As long as the content is as good as human produced content - then Google will accept it without penalty.” - Automated Insights
Talia Klein Perez is a small business owner who specializes in small business marketing services, ranging from content writing to social media marketing. Talia also operates two websites of her own, so her expertise comes from both the side of the owner and the side of the marketer.