AI today: superior processing power and a wealth of data give rise to machine learning and deep learning
This is the second of a NewtonX three-part series on AI for senior executives. Part 1 examined the history of AI and what led to the breakthroughs of the past decade. In this second part we take a look at the two factors that enable machine learning and deep learning, and what each of these terms really means.
The explosion of data in the mid aughts gave us the necessary training material for AI, while the birth of cloud computing for storing and processing that data made AI truly feasible (the graph below demonstrates the massive price drop off in data storage). These two factors — processing power and data explosion — allowed for the first commercial applications of AI, including Google’s search algorithm, the Roomba, and the Netflix predictive recommendations engine.
This entry of AI into everyday life exerted pressure on executives to invest in the technology. The two type of learning that executives identified as the most important to be knowledgeable about are machine learning and deep learning — this is an overview of the terms and their applications.
Machine learning describes a type of algorithm that learns over time without being specifically programmed. One common type of machine learning is predictive algorithms, which are widely implemented in our daily lives: the weather app on smartphones utilizes predictive analytics; Spotify Discover playlists use predictive; even schools use predictive analytics to identify at-risk students.
Machine learning algorithms incorporate new data in real-time to improve accuracy over time. So for instance, AI-powered customer service platforms will say that they utilize machine learning — which means that their predictive algorithm for responding to customer queries via chatbot or for auto classifying tickets learns over time without receiving explicit programming instruction.
Deep learning is a particular form of AI in which, as described above, the algorithm can learn on its own through an artificial neural network. Deep learning is significantly more effective at accurately producing results in image recognition (as in the dog example above) and voice recognition. It is utilized in everyday applications such as Facebook’s facial recognition tool for tagging and Alexa’s speech recognition technology. This form of AI is particularly useful in scenarios in which you want the machine to be able to recognize previously unknown input, and be able to classify or identify it correctly. For instance, if the machine sees a breed of dog it’s never encountered before, it should still be able to identify the image as “dog”.
One of the biggest use cases for deep learning is identifying defective products on a production line through image recognition and classification. Deep learning has also been employed as a tool for image-based diagnoses in medical contexts.
Now that you have a comprehensive understanding of AI, in Part 3 of this series we will delve into the costs and limitations of implementation in enterprise settings.