The world of IT significantly advances our lives, in our business pursuits and in our personal adventures. But the vocabulary that accompanies new IT advancements can create confusion, especially when changes occur rapidly. This blog will consider the changing landscape and recent developments in what is commonly called “Cognitive Services”, one of the fastest changing and likely most impactful areas of current IT change. This can be a challenging topic to navigate, so, let’s start at the beginning.
What are Cognitive Services?
A Cognitive Service is a piece of computer software which automates the traditionally human process of acquiring knowledge and understanding through thought, experience, and the senses. In simple terms, a Cognitive Service can do what we used to believe only humans could do. For example, a Cognitive Service can “look” at a document and understand the content of that document. Or it can recognize a person by seeing that person’s facial features. Or it can predict what will happen next based on an understanding of past patterns of what happened before.
At the core of these Cognitive Services, and what makes them really revolutionary, is that they learn through experience, in much the same way that people do. In effect, they get smarter and more able to perform “cognitive” functions as they are exposed to more documents, faces, circumstances, etc. This is called “Machine Learning” and that capability lies are the heart of Cognitive Services. It’s what makes them really different to previous generations of computer software which could only do what a human had designed them to do. The “Machine Learning” that powers Cognitive Services allows those services to teach themselves. Think about that for a moment.
Cognitive Services form part of the broader “Artificial Intelligence” sphere. What’s going on inside a Cognitive Service is a type of intelligence. It’s intelligence that is not human or animal – hence the label “Artificial Intelligence”. By combining a number of Cognitive Services it is possible to mimic the action of a human performing a task or job. The result is an automated system that performs work traditionally requiring human intelligence. Commonly this is now called “intelligent automation”.
Whether these systems are truly “intelligent” is a matter of debate. Given some of the decisions that we (or our elected leaders) make, perhaps we should acknowledge that we humans do not always display intelligent behavior ourselves! At the same time, it has become clear that in many cases intelligent automation does a better job than humans can. Let’s come back to that.
Over the past few years, there has been increasing interest in “intelligent automation”. Gartner has estimated that by 2025, 50% of companies will have devised Artificial Intelligence (AI) orchestration platforms to “operationalize” AI. This is an increase from 10% in 2020. Most of these companies will use “cognitive services”.
We’ve had software containing elements of “Machine Learning” and “intelligence” for quite a few years, so why is there such a surge in AI and Cognitive Services now?
Every Cloud has a Silver Lining
Historically, creating and building software with cognitive features (processing text and language, speech, etc.) has been computationally expensive. AI systems with “learning” capabilities require large datasets and a lot of processing power. That computing power was previously so expensive that Cognitive Services were not commercially viable. It was simply too difficult to accumulate the very large datasets that “Machine Learning” systems required in order to teach themselves.
IBM’s “Deep Blue” was probably the first widely recognized AI software, rising to infamy when it defeated Gary Kasparov at chess. The method of “intelligence” which “Deep Blue” employed was to essentially project billions of moves into the future and evaluate what was the most advantageous. Significantly powerful and massively expensive computers are needed to produce that sort of “intelligence”.
DeepMind (now owned by Google) subsequently created its “AlphaGo” AI system and went one better than IBM. AlphaGo used “Machine Learning” to defeat the world Go champion Lee Sedol in this hugely complex game, widely considered more difficult than chess. AlphaGo’s victory shattered the world’s perceptions of what an automated system could do. Lee Sedol acknowledged that AI can perform better than a human can. “With the debut of AI in Go games, I’ve realized that I’m not at the top even if I become the number one through frantic efforts. Even if I become the number one, there is an entity that cannot be defeated,” he said.
Recent developments, including the widespread availability from Amazon, Microsoft, and Google of relatively inexpensive computing power, mean that businesses can now consume significant processing power “on demand”. Therefore, less investment is required to access the infrastructure needed to create Cognitive Services that use “Machine Learning” to perform work that traditionally required human intelligence. By making infrastructure available on a “pay as you go” model, cloud computing democratizes computational power to everyone, not just those with the deepest pockets. As a result, the creation and use of Cognitive Services is now feasible by just about any small startup company.
Service with a Smile (SaaS)
Right on the heels of highly-available cloud infrastructure came the “Software as a Service (SaaS)” business model. This replaced traditional methods of software purchase which had required large capital investment and expensive software installations owned, operated and maintained by the organization using them. Instead, SaaS allows buyers to pay for software based on a usage metric (such as users per month or transactions per year). The organization using the service doesn’t own the software, doesn’t own the infrastructure it runs on, and doesn’t have to maintain any of it.
This SaaS business model also provides the option of a feature set model of usage, i.e., the more you pay the more features you can use in the software. This model is so widely accepted that customers can now buy just one “feature” or function and consume it as a service. In order to get started, you need no hardware, just login credentials and a credit card.
Tech Giants Drive a Wave of Innovation
Microsoft, Google, Amazon and other technology industry giants have increasingly chosen to sell their software through the SaaS model, moving away from traditional revenue models. They have each created a vast infrastructure of cloud computing power that serves as the foundation on which computing services can be created and delivered.
The democratizing effect of cloud computing has driven a surging wave of innovation in computing services. New Cognitive Services are proliferating quickly, available in the Service Marketplaces of tech giants or from independent vendors who deliver their own Cognitive Services in a SaaS model. The result is an explosion of AI vendors, as entrepreneurs grab the opportunity to put good ideas into practice on top of the tech infrastructure now more available to them.
What Cognitive Services are The Big Three Offering?
In recent years the Cognitive Services available from Microsoft, Google and Amazon have become much more powerful. They have invested hugely in cloud computing infrastructure, SaaS business models, and the large datasets that daily feed the ravenous learning appetites of their Artificial Intelligence engines.
Microsoft provide a wide range of capability under the label “Azure Cognitive Services”. These can be used on speech, language processing (including sentiment analysis), vision (including the ever-controversial “face recognition” capabilities), and even “Decision” models which can be used to tackle content moderation.
Amazon similarly provide “Comprehend” and “Textract” for language and document analysis and “Rekognition” for image and video. They also exploit AI in their own cloud to provide several translation and speech services, such as “Translate”, “Polly” and “Transcribe”.
Google’s extensive AI services include “Vision AI”, “Translation AI”, and “Cloud Natural Language”, and Google also provides underlying infrastructure that enables the creation of AI models.
What About Everyone Else?
Other software vendors are crowding into the Cognitive Services market, creating Cognitive Services that exploit “Machine Learning” to solve specific types of business problems. ABBYY offers in its Vantage product, a Cognitive Service that understands the content of documents. In its Timeline product, ABBYY uses Artificial Intelligence to help predict future behavior based on patterns of past behavior of core business processes. Kofax likewise uses “Machine Learning” to automate the understanding of documents. UiPath does the same to enable its aptly named Document Understanding product.
In just this “document understanding” subset of the Cognitive Services ecosystem the number of new entrants making a mark is striking. This ranges from general purpose vendors such as Hyperscience, Rossum, and Infrrd, to special purpose vendors focused on narrower markets and geographies. Most notable is simply the ease with which new entrants get started. Cloud computing, SaaS business models, and more widespread understanding of Machine Learning and data science, have created fertile ground for market growth and intense competition.
Cognitive Services vs Application Vendors
Until recently the Cognitive Services vendors, including Microsoft, Google, and Amazon, have been focused on selling “bite-sized” services. They can, for example, tell you the contents of a document in an instant, but how you get that document to them and what you then do with the result of their Cognitive Service is not their concern. Technical teams of software developers use these bite-sized Cognitive Services to build their own end-to-end AI applications from a series of “building blocks”. Each service has its own distinct feature, and it’s up to you as a customer to work out the most advantageous match to suit your needs. In this way, Cognitive Services vendors can thrive on the simplicity of their offerings, as they don’t have to be concerned about the complex interactions and dependencies that enterprise software typically faces.
More recently Cognitive Services vendors have started to package these services to specific use cases, such as Invoice Processing. It’s clear that they will become more sophisticated as time passes and capable of addressing complete use cases from start to finish. Simultaneously, traditional software application vendors who directly automate entire use cases, are moving to “Cloud SaaS” models and incorporating Cognitive Services into their applications as they do so. They are also increasingly interested in monetizing small sets of features already inherent in their software.
The Rise Continues
An interesting battle between the new Cognitive Service kids on the block and the incumbent forces of the application software industry is under way. The rise of Cognitive Services is a sign of the furious pace of innovation, change and competition spurred on by advances in “Machine Learning” and AI more generally. At times, it seems as though the tectonic plates on which our IT world rests are cracking. Ultimately, we should all be the benefactors of this period of intense creativity and entrepreneurship, even if there is some disturbance along the way. In any case, it’s important to stay informed!
