From autonomous vehicles, predictive analytics applications, facial recognition, to chatbots, virtual assistants, cognitive automation, and fraud detection, the use cases for AI are many. However, regardless of the application of AI, there is commonality to all these applications. Those who have implemented hundreds or even thousands of AI projects realize that despite all this diversity in application, AI use cases fall into one or more of seven common patterns. The seven patterns are: hyperpersonalization, autonomous systems, predictive analytics and decision support, conversational/human interactions, patterns and anomalies, recognition systems, and goal-driven systems.
Any customized approach to AI is going to require its own programming and pattern, but no matter what combination these trends are used in, they all follow their own pretty standard set of rules. These seven patterns are then applied individually or in various combinations depending on the specific solution to which AI Is being applied.
(Disclosure: I’m a principal analyst with Cognilytica)
The hyperpersonalization pattern: treat each customer as an individual
The hyperpersonalization pattern is defined as using machine learning to develop a profile of each individual, and then having that profile learn and adapt over time for a wide variety of purposes including displaying relevant content, recommend relevant products, provide personalized recommendations and so on. The objective of this pattern is to treat each individual as an individual.
Implementations of the hyperpersonalization particular pattern include creating personalized recommendations based off of browsing patterns and searches. A company that is currently utilizing a technology of this nature would be Netflix. They use AI to help recommend shows and movies to viewers based off of personal choices. Another example is Starbucks, who is using hyperpersonalization to connect with their customer base.
The implementation of hyperpersonalization isn’t only limited to the marketing industry. It’s also popping up in industries such as finance, healthcare, or personalized fitness and wellness applications. For example, one common application of where hyperpersonalization can make a huge impact is in finance and loans. In the US the FICO credit score is used to lump individuals together who might otherwise have vastly different amounts of credit worthiness and penalizes groups of individuals who lack credit history. By moving away from using the traditional FICO score into something that treats each individual as an individual we might get more accurate pictures of individuals to see just how likely they are to pay back loans.
Autonomous systems pattern: reducing the need for manual labor
Autonomous systems are physical and virtual software and hardware systems that are able to accomplish a task, reach a goal, interact with their surroundings, and achieve an objective with minimal human involvement. Where the primary objective of hyper-personalization is to treat people as individuals, the goal of autonomous systems is to streamline things with as little human interaction as possible. The autonomous pattern requires machine learning capability that can independently perceive the outside world, predict the future behavior of external elements, and plan for how to deal with those changes.
Obvious applications of this pattern include autonomous machines and vehicles of all sorts includes cars, boats, trains, airplanes, and more. However this pattern also includes autonomous systems including autonomous documentation and knowledge generation, autonomous business processes, and cognitive autonomation. These include systems that can operate in close proximity to humans, including preferential decision making.
AI powered predictive analytics
Another pattern of AI is predictive analytics and decision support. This is defined as using machine learning and other cognitive approaches to understand how past or existing behaviors can help predict future outcomes or help humans make decisions about future outcomes based on these patterns. The objective of this pattern is helping humans make better decisions.
Some uses of this pattern include assisted search and retrieval, predicting some future value for data, predicting behavior, predicting failure, assisted problem resolution, identifying and selecting best fit, identifying matches in data, optimization activities, giving advice, and intelligent navigation. The idea is that it helps to make better decisions, providing augmented intelligence capabilities. Machine learning is what is helping to make the decision, adapting over time to provide better results.
The conversational pattern: machines that can communicate as humans do
Another pattern of AI is the conversational/human interaction pattern. This is defined as machines and humans interacting with each other through conversational forms of interaction and content across a variety of methods including voice, text, and image forms. This includes machine to human, human to machine, and back and forth human and machine interaction. The objective of this pattern is enabling machines to interact with humans how humans interact with each other.
The most obvious examples of this include pattern include chatbots, voice assistants, and sentiment, mood and intent analysis. The point is that it is trying to understand the intent behind human interactions. It can also be used to facilitate human to human interaction through translation. The big thing to remember is that this pattern is used to create an easier way for humans to interact with each other and machines through methods that are natural or comfortable for humans.
Identifying patterns and anomalies with AI
Machine learning is particularly good at identifying patterns and finding anomalies or outliers. The “pattern-matching pattern” is one of the repeating approaches to AI projects that has seen wide and increasing adoption. The goal of the Patterns and Anomalies pattern of AI is to use machine learning and other cognitive approaches to learn patterns in the data and learn higher order connections between data points to see if it fits an existing pattern or if it is an outlier or anomaly. The object of this pattern is to find what fits with existing data and what doesn’t.
Applications of this pattern include fraud and risk detection to see if things are out of the ordinary or expectations are happening. Another application is finding patterns among data, and helping to minimize or fix human mistakes. This pattern also includes predictive text, where it can analyze patterns in speech and grammar to help suggest words to choose to speed up the writing process.
Machines that can recognize the world: the recognition pattern
One of the big advancements in machine learning is the use of deep learning to greatly improve the accuracy of recognition-related tasks such as image, video, audio, and object recognition, classification, and identification. The recognition pattern is defined as using machine learning and other cognitive approaches to identify and determine objects or other desired things to be identified within image, video, audio, text, or other primarily unstructured data formarily. The objective of this pattern is to have machines identify and understand things.
Examples include image and object recognition, facial recognition, audio and sound recognition, handwriting and text recognition, and gesture detection. This is a well developed pattern that computers are really good at and is fairly widely used. There are many companies who are investing heavily in recognition systems. In fact, one of the most well funded AI companies, Sensetime, is focused on facial recognition applications and the Chinese government is investing heavily in the use and adoption of this pattern.
Solving the puzzle: the goal-driven systems pattern
Machines have proven to be particularly adept at learning the rules of games, and beating humans at their own games. In the past, machines have easily conquered the games of checkers, chess, and finding solutions to mazes. Through the power of reinforcement learning and much more advanced computation capabilities, machines are now able to win at Go, multi-player games such as DoTA, and much more complicated games. Alpha Go and Alpha Zero were created by Google’s DeepMind division under the theory that through goals, computers could learn anything through game play. Games are just the beginning to solutions that could potentially even lead to breakthroughs in solving long-hoped for goals in Artificial General Intelligence (AGI).
Games are not the only possibility for goal-driven systems. With the power of reinforcement learning and other machine learning techniques, organizations can apply machine learning and other cognitive approaches to give their systems the ability to learn through trial and error. This is useful for any situation where you want to have the system find the optimal solution to a problem. The main learning approach for this pattern is through reinforcement learning. Examples in the pattern can include game playing, resource optimization, iterative problem solving, and bidding and real-time auctions. While the goal-driven systems pattern is not yet as widely implemented as some of the other patterns, it is gaining rapid adoption.
Combining patterns for AI project success
While these might seem like discrete patterns that are implemented individually in typical AI projects, in reality, we have seen organizations combine one or more of these seven patterns to realize their goals. By companies thinking of AI projects in terms of these patterns it will help them better approach, plan, and executate AI projects. In fact, emerging methodologies are focusing on the use of these seven patterns as a way to expedite AI project planning. Once you know that you’re doing a recognition pattern, for example, you can gain insight into a wide range of solutions that have been applied to that problem, insights into the data that’s needed to power the pattern, use cases and examples of applications of the pattern, algorithm and model development tips, and other insights that can help speed up the delivery of high quality AI projects.
While AI is still in the early majority phase of adoption, it’s clear that the identification and use of these patterns will help organizations realize their AI project goals more quickly, with less re-inventing of the wheel, and with much better chances of success.