When people first learn about CONCURED’s capabilities, one question we hear often is if it is an artificial intelligence (AI). Or they want to know if we’re using AI in some way. In this article, we’re going to lay it all out there and explain what an AI is and the different AI technologies CONCURED uses to drive content marketing for our customers.
To understand AI, we have to look at its history. AI research and business use have gone through several boom-bust cycles. By understanding the history and the struggles, we can get a clear picture of just what an artificial intelligence is and isn’t.
As a discipline, the field was formed at a conference at Dartmouth College in 1956. After noticing similarities in how digital signals in computers and the brain were similar, it was hypothesized that humans could build an artificial brain that worked like human brains.
Such a brain would have what is called “general intelligence”. It would be able to reason about many different kinds of problems. Such an AI is now called a “strong AI” or an “artificial general intelligence”. This is the sort of AI that you see in science fiction books and movies and is the sort of AI that the general public knows.
Unfortunately for the researchers, they ran into many problems and the funding dried up by the mid-70s. A lot of valuable research was created, but the products were little more than toys. What went wrong?
One problem was the level of computing power needed. The computers of the time were woefully inadequate. Storage, memory, and processing power weren’t enough to really start tackling the problems of a general intelligence.
There were also several theory problems and paradoxes that showed that an artificial general intelligence would be incredibly hard to make. We’ll just highlight two of the issues here. The first is our common sense reasoning is built on a foundation of an astounding amount of information. If we wanted to make a computer that could see or interpret language, we would need a way to gather, store, and analyze much more information than was possible back then.
The second issue is that, paradoxically, what we consider to be a hard problem is easy for a computer, but things we find simple are incredibly difficult for a computer to do. Our abilities to recognize faces, answer questions correctly, and manipulate ourselves and our space were far more complicated than we realized.
There were also political rifts within the community and between disciplines about the feasibility of creating a general AI and how these problems would best be solved. Funding dried up and dreams of an artificial general intelligence were shelved. But there were some gems found from all the research. Researchers and businesses asked a new question.
What if instead of a general intelligence, we made one that was smart about a specific problem or domain of knowledge? Experts would program all of the rules they used to make decisions into the program, then feed the program lots of information. After a process of refinement, the program would make decisions just like an expert would, but much faster.
This was the creation of expert systems and created a second AI boom in the 1980s. They were the first AI product to be used by business. These systems saved companies millions of dollars and made it clear that if you wanted an AI to answer a problem you would have to give it a lot of information. The success of these systems led to more funding and further research into artificial intelligence, including things like speech recognition and optical character recognition.
However, by 1993, the expert systems industry collapsed for economic reasons. The specialized hardware used wasn’t as powerful as the then-new PCs coming out. Also, while expert systems were clearly useful they had strong limitations. They were difficult to change and handled outliers poorly.
Unlike the last time AI interest collapsed, researchers continued to get funding to explore new approaches. This led to further research, but it also created a lot of fracturing within the field as they competed for funding. Many were reluctant to call their research as artificial intelligence because they felt it could damage their chances to get funding. It’s one of the reasons there is such a confusion of terms today.
Still, advances were made and some were used in business, though often weren’t called AI. In fact, a funny process began to happen. When AI research found something useful, those solutions were pulled into other products and called something else. The solutions were just seen as a clever computation and not “real” intelligence. Every time a milestone was reached, it was discounted. This has been called the AI effect.
So what changed, and why AI now a positive term for business?
As this article from Wired explains, three technologies came around that made AI much more attractive to businesses. First was the discovery that graphics processing units were phenomenal at parallel computation. That gave AI the horsepower it needed. The second was the set of technologies that allow us to capture, store, and analyze vast data sets known as “Big Data”. That gave AI makers ways to give their programs massive amounts of information to work with. Finally, there was the discovery of deep learning algorithms which optimized the way information is processed through these programs. These came together around 2011 and the field hasn’t been the same since.
Researchers now have all of the pieces they need for AI to take the next leap up in utility for businesses and to push the boundaries of research. Technical terms from AI are now business buzzwords, and the general public’s knowledge about AI is growing. Some even think we’re on the edge of reaching the holy grail of AI research, the creation of a strong AI. Considering the history so far this is unlikely, but it is true that AI has reached a new level of usefulness and acceptance.
So, to answer the original question at the start of this piece, artificial intelligence is best seen as a collection of technologies that simulate different domains of intelligence to provides useful answers to human questions.
So if AI is a collection of different technologies and approaches to solving difficult problems, which ones does CONCURED use?
The two core technologies are natural language processing (NLP) and topic modeling. We’ve written about NLP in the past, but not so much about topic modeling. Topic modeling is a technique used in NLP to model what the topics of a particular piece of text is about based on the words used, their proximity to each other, and other factors. We also use speech recognition to make transcripts of videos that feed into our NLP system.
From there, we use machine learning to help our system pull in new pieces, analyze them, and put them into context. Machine learning is a technology that lets a program learn from its previous results to make a better guess next time. You see this all the time when Google tries to guess what you want to know based on your previous search results.
The purpose of all this is to discover the semantic relationships between pieces of content. Just as Google uses what you click on and your queries to guess what you want, we use what’s posted on your site and by your competitors to make an educated guess on what people want to know about next in the moment. As the number of categories grows and our system gets more experience analyzing different niches, we hope to be able to release trending keyword predictions for what people will want to read a week in advance.
This will involve new applications of deep learning and additional proprietary research. CONCURED will keep trying to push the envelope of what AI technologies are capable of to solve the marketing problems of our time to reach our holy grail of content marketing, knowing the perfect thing to write about next that will yield maximum engagement from any niche.
So, is CONCURED an AI like you see in science fiction? No. But it does use the latest advancements in AI and builds upon them to create our unique product. We are looking forward to releasing several new features very soon that will demonstrate how far our processes have come. If you are interested in a demo of our technology, contact us through this page.
Further your understanding of AI with our article "Why AI is better than A/B Testing"
Subscribe to our Clever Content Club below! We'll send you all you need to know about AI and Content Marketing
Learn more about Content Marketing, Content Intelligence, Content Marketing Artificial Intelligence on our blog.
276 Fifth Avenue, Suite 704 New York, NY 10001, USA
Canalot Studios, 222 Kensal Road London, W10 5BN, United Kingdom
3 Place Ville Marie, Montreal, QC H3B 2E3, Canada