Marketers, be thankful your job is more complicated than playing a game of Go.
Although humans have made great strides in artificial intelligence, applied AI has only given us smarter tools and limited-functionality assistants. As Noam Chomsky reflects in this Scientific American interview, “Even tasks mastered almost reflexively by infants are far beyond the capacities of contemporary AI.” So it’s unlikely you’ll be working for an AI overlord any time soon.
John McCarthy first coined the term “AI” in 1956 with the premiere of an academic conference on the subject. Since then it’s grown into a vast and complex subject that’s found use in an increasing variety of applications.
Chances are, you’ve been exposed to A.I. without even knowing it. If you’ve ever used SIRI, visited Facebook or made a purchase on Amazon, artificial intelligence has been there to help your journey.
In this post, we look at over two dozen branches of artificial intelligence to understand its impact on marketers, along with examples of how it is being applied to content.
Artificial creativity is an attempt to capture human-level creativity within an algorithm. Some argue that AI isn’t creative at all and that the best it can do is to mimic human creativity. Despite this, AI creativity is currently employed in various low-level ways among popular online publications.
Artificial intelligence creativity was initially seen in earnings reports and data-heavy stories about sports and election results. This type of writing follows a highly structured formula with a straightforward narrative, making it an ideal situation for the use of AI. In this context, AI augments the efforts of content creators, allowing them the opportunity to focus their time on higher-value content.
Thomson Reuters began using artificial intelligence to automate the publication of earnings reports back in 2006. Other well-known publications creatively using AI include Associated Press, Forbes, ProPublica and The Los Angeles Times.
Automated Planning and Scheduling
At its highest level, this branch of artificial intelligence deals with turning strategies into reality. However, real-world applications tend to be less abstract and more concrete in their approach to solving everyday challenges.
AI excels at rapidly analyzing vast amounts of data, which is a significant benefit to marketers. Smart software can quickly distill large buckets of information into actionable insight.
The most significant benefit for content marketers is that they’re no longer bogged down with analysis. What used to take hours, if not days, can now be done in minutes. This leaves more time for high-value activities such as critical thinking and decision making.
MarketMuse employs artificial intelligence to build content strategies that can compete for specific topic categories. It uses topic modeling to surface relevant topics that may need to be targeted with additional content or the optimization of existing pages.
Tailor Social, a social media management tool, uses artificial intelligence to help schedule social media posts. It recently came out of beta.
Automated reasoning seeks to understand the different aspects of thinking and could one day lead to “extracting aesthetic properties, sentiment, and even emotions” from content.
But we’re not there yet, in fact, we’re far from it. But when that day arrives, we’ll see even more powerful tools that plumb the depths content where marketers have neither the time nor desire to venture.
Automation has existed a lot longer than artificial intelligence, but AI has helped create smarter and better ways of increasing efficiency. Advances in this field have enabled the automation of tasks that were previously the domain of humans. As a result, the type of work that content producers perform is changing.
Take for example the web design process which typically involves numerous disciplines including design and content creation. The Grid is an AI-powered web design system that creates websites based on the content you supply.
Computer Vision / Object Recognition
Digital image processing is used in a number of areas including pattern recognition, classification and feature extraction. On a practical level, A.I. computer vision makes managing digital assets easier. MavSocial, using image recognition technology from Miro is one example.
Remove.bg uses sophisticated AI technology to detect foreground layers and separate them from the background. A difficult process that would typically require an hour or more using Photoshop now takes just five minutes.
Another way this technology can help is by automatic content moderation. Clarifai provides the ability to moderate and filter sensitive information from your platform without human intervention.
Girard & Girard, in the Online Journal of Applied Knowledge Management (PDF), state that “knowledge Management is the management process of creating, sharing and using organizational information and knowledge.” As a branch of artificial intelligence it encompasses various aspects such as concept mining, data mining, text mining, information extraction, and knowledge representation. It’s within these specific applications that we already find benefits for content marketers.
Concept mining comes from the extraction of concepts from artifacts, such as a web page or blog post. For example, Aylien offers an API with a concept extraction endpoint to find what topics are mentioned in a piece of text.
Being an API, it’s not something of which most marketers can take advantage. However, software developers can incorporate this API into their own software for use in the content marketing arena.
Refer back to the section on Artificial Creativity for some ideas on how this could be used to the benefit of marketers.
Data mining seeks to discover patterns in large sets of data. From a content marketing perspective, this branch of AI can help determine the best content to present at the right time in the buyer journey
By analyzing cardholder spending, American Express can present customized offers that attract and retain customers. Then it uses this targeted marketing to match merchants with the right customers, meaning those who tend to spend more than the average consumer.
British Airways mines its data assets to uniquely ID every single customer and personalize their email marketing and other customer communications.
Text mining is a type of data mining explicitly focused on extracting information from text using pattern recognition and other approaches via natural language processing (NLP). Advanced grammar checkers like Grammarly are one type of tool that has been created from this technology. There is no doubt marketers appreciate the benefits!
On a more advanced level, Acrolinx platform offers a linguistic analytics engine that guides writers style, grammar, terminology, and tone to ensure all your content remains on-brand. This addresses a significant challenge faced by all enterprise content creators; keeping content consistent and reflective of the brand, irrespective of the content creator.
Analyzing event log data to identify trends and patterns is known as process mining. Livejourney has applied this to track the customer journey in real-time. By monitoring all customer touchpoints and analyzing journeys from a customer’s perspective, you can address any inefficiencies and increase customer satisfaction.
E-mail Spam Filtering
Fight spam with spam. Rescam is an artificially intelligent email bot that replies to scam emails. The chatbot uses one of its human personalities to continue the conversation with a would-be scammer. Wasting their time with the bot gives scammers less time to pursue real victims. To date, Rescam has wasted over five years of scammers’ time.
This AI branch frequently focuses on processing human language text using NLP. Real-world applications range from the relatively simple such as automatically extracting text from emails (Parseur) to the more complex task of summarizing a document (frase). A direct use case of this technology would be providing automatic text summaries for daily new roundups or for use in content curation.
The semantic web was a term coined by Tim Berners-Lee, inventor of the world wide web, back in 2001 (pdf). The idea is that all data on the internet, content, links, and transaction can be analyzed by machine. But the concept struggles to gain traction. Although the number of sites using semantic web markup continues to increase, they are still in the minority.
One of the more notable uses of semantic web technology is the BBS which used it to power their entire World Cup website in 2010. Other significant uses of semantic web technology include Time Inc., Elsevier, and the Library of Congress.
However, formatting content for the semantic web is far more complicated than current processes. The added time and expense may account for its slow adoption among marketers.
Machine learning is a field of artificial intelligence that gets computers to “learn” from data, without being programmed for the task. Recommendation engines like those used in Curata, Scoop.it and Zetaare machine learning in action.
The benefit for systems like these is that recommendations improve over time as the system learns what constitutes a good suggestion. However, marketers may not have the patience to wait for those improvements!
Persado uses “AI generated language for email subject lines and social media paid ads” optimized for conversions. Think of this as automatic testing for conversions.
You don’t need to be a conversion rate optimization specialist, nor regularly conduct tests. The AI behind the software takes care of all that work, continually learning and presenting the marketer with better choices.
Imagine offering real-time recommendations with content personalized for each user. LiftIgniter uses machine learning to achieve that objective. It’s continually learning and optimizing based on a specific goal you’ve set.
Machine learning can do more than just help make appropriate recommendations. It can even be used to create, or more accurately curate, content.
IBM Watson helped create a trailer for a Hollywood suspense/horror film.
First, it was taught how to identify frightening scenes and horrifying music and analyze the composition of movie scenes to determine what made them scary. Then it watched the full-length version of the film and selected ten moments that would be the best contenders for the trailer.
It was left to human ingenuity to perform the actual editing task. Here’s the result:
Deep learning is an aspect of machine learning based on data representation instead of using task-specific algorithms. It has been applied to various fields including natural language processing, machine translation, and computer vision.
At the 2018 Game Developers Conference, NVIDIA demonstrated “recent research into ways that Deep Learning networks can be used to generate realistic looking human animation,” and how they’ve applied deep learning to texture synthesis.
In a more practical application Envision uses artificial intelligence to help marketers choose the best video thumbnail, segment, or picture. It also uses deep learning to automatically find and add the best hashtags for your Instagram posts.
Natural Language Processing
For marketers dealing with text content, natural language processing is an essential field of artificial intelligence. This branch of AI is concerned with enabling computers to understand language.
The significant challenges in this field are speech recognition, natural language understanding, and natural language generation. However, considerable progress is being made as seen in the proliferation of chatbots, machine translation, and the use of natural language user interface like SIRI.
Here are a couple of interesting examples to note.
Narrative Science automates personalization at scale using natural language generation to automatically generate descriptions and provide personalized communications.
Conversica relies on AI to help nurture inbound leads until their interest change to an intent to buy. More specifically, they use artificial intelligence to interpret email responses, sending unique and natural-sounding responses as a follow-up.
Chatbots are everywhere! They’re typically used to qualify leads or employed as a content delivery system.
Two basic types of chatbots are ones that understand limited commands and those that use natural language processing.
Limited command type chatbots are useful when task options are limited, and you can guide the interaction down very well defined paths. NLP chatbots allow for a more conversational flow.
Chatbot platforms such as Flow XO remove the need for coding skills, but they can still be challenging to implement for a few reasons:
- Marketers need to know the tasks their visitors are looking to accomplish.
- They need a deep understanding of the journeys taken by their audience.
- The type of content used in this context can be radically different.
- They may lack a rich content strategy.
Language identification is a critical component of machine translation and natural language processing. Although it doesn’t have a direct impact on marketers, content strategists working at the enterprise level often deal with language translation issues and its effect on content.
Google AI is deeply involved in all things related to natural language technologies, including language identification. You may have experienced this when visiting a page in a foreign language.
For content strategists with zero budget for translation, this can be a lifesaver.
Natural Language User Interface
Prototype natural language user interfaces first appeared in the late 1960’s and only recently have become popularized. If you’ve ever used SIRI or Amazon Echo, you’ve experienced a natural language interface.
For content marketers, this potentially changes the way we think and create content. Search engines remain keyword-based and content creators often create long pieces of content in the hope of ranking well for many keywords.
But the use of a natural language interface, especially one that’s voice-controlled, creates a new context. Users are searching for targeted answers to specific questions, something for which a 10,000-word manifesto is ill-suited. Although we’ve seen a tendency towards more extended content, this trend could soon change.
Natural Language Understanding
Natural language understanding is a sub-topic of natural language processing dating back to 1964 with the publication of Daniel Bobrow’s Ph.D. dissertation. There’s a great deal of interest in natural language understanding due to its relevance to large-scale content analysis, question answering and text categorization.
Real world applications, at least for marketing use, requires fairly sophisticated comprehension capabilities in order to grasp the concepts within a document. OneSpot is one such company that uses a combination of machine learning and natural language to process site visitors’ content consumption and deliver individualized content at scale.
Another sub-branch of natural language processing is machine translation. If you’ve ever used Google translate, you’ve scratched the surface of machine translation.
As translation accuracy continues to increase, so do opportunities for translation in B2B and consumer markets.
Content strategists working at global corporations leverage machine translation to translate content at scale in a cost-effective manner. Business domains like technology, finance, legal, healthcare, etc. have their own nuances and terms, which is why domain-specific machine translation has been the preferred route.
However, multi-domain machine translation, like that offered by Kantan is beginning to gain ground.
Question answering is a subset of natural language processing that designs systems to answer questions posed in a human language automatically. The START Natural Language Question Answering System claims to be the “world’s first Web-based question answering system, has been on-line and continuously operating since December 1993.”
Did you know that Facebook is training AI to answer questions? They’re teaching AI to analyze images and respond to questions asked about those pictures. Right now the AI can only provide simple answers. But the idea is to enable it to offer more elaborate responses just like a human would.
Semantics is a branch of linguistics concerned with interpreting the meaning of words and their structure. Semantic translation aims to retain the meaning of a document when translating into another language.
Content strategists working for global business are challenged with content governance across multiple languages. The use of semantic translation can help increase translation efficiency and mitigate the risk of incorrect interpretation. PROMT is one example of an application incorporating semantic information.
The Future of Artificial Intelligence in Content Marketing
Elon Musk may fear an A.I. apocalypse but, from what I’ve seen, we’re still struggling to create viable applications for many branches of artificial intelligence. That said, we are starting to see smarter software, limited in scope, that substantially improves our lives.
The recommendation engines powering giant companies like Netflix and Amazon is one example. Now you can bring that sophistication to your own site with software like Bibblio. Unlike the simple approach provided many “related-post” apps, its AI-powered algorithm offers semantically relevant content suggestions.
Unless you run a large site, software such as this may not be a big hit. However, the field of artificial intelligence is making real quantifiable steps that impact marketers right now, today. Just don’t expect AI to perform your job for you any time soon.
Feature image vector designed by Freepik