Published on July 2nd, 2019 | by Emergent Enterprise0
Wrapping Your Head Around AI
Set aside the doomsday theories of AI taking over the world and see how actual businesses use it. Toby McClean at Forbes gives a straightforward explanation of AI and its promise and potential. As companies large and small implement AI solutions, there will be less handwringing and more handshaking – with our machine workmates.
2001: A Space Star Terminator, TheEx Machina
But let’s step back and look at what today’s AI really is and decide whether it is really something to fear. I propose that today’s AI is nothing more than a new paradigm for developing computer applications. Rather than a software developer sitting down and writing a set of rules that map inputs into outputs, the rules for mapping inputs to outputs are discovered as patterns in a large amount of training data.
What Does AI Really Look Like?
Digging deeper, today’s artificial intelligence has three key ingredients: data, recipes and optimizations. The data is the foundation and the basis for learning. Just like we as humans learn from books, movies, teachers, etc. — all of which are sources of data — a computer needs data from which to learn in order to become intelligent. And with the proliferation of sensors —sensors attached to refrigerators, sensors attached to pipelines, sensors attached to production lines and even sensors attached to cows — there is more data than ever before. The volume and quality of data have a critical impact on the success of an artificial intelligence system.
The recipe of an AI system provides guidance on how the computer should discover patterns in data to produce outputs. The recipe helps the computer to create rules for classifications, segmentation and predictions based on the data it is learning from. In the case of machine learning, this is a network that consists of several layers that guide the learning process to solve a specific problem. Just like in the preparation of the perfect meal for dinner guests, we sometimes have to experiment to learn which recipe yields the best result for the ingredients (data) we have at hand.
Finally, optimizations measure and provide feedback to the computer regarding success in finding patterns to map the inputs to outputs in the data. Think back to your days in school when you spent time learning a new subject — the culmination of your learning was a test to measure how well you grasped the subject matter. Optimizations in artificial intelligence are nothing more than this. In the context of AI in a business environment, this optimization is typically tied to a specific business problem that you are trying to solve. And more often than not, it takes a human to provide guidance to the machine as to how exactly to measure that success.
According to a recent survey by Infosys, organizations that had made more progress on AI initiatives also showed faster revenue growth, clearly linking AI maturity with a company’s bottom line.