What’s the difference between artificial intelligence, machine learning and deep learning?

The terms “artificial intelligence,” “machine learning” and “deep learning” describe a process that has built on itself over the past few decades, as the world has made enormous advances in computing power, data transfer and other technology goals.

The conversation should start with artificial intelligence, a broad term for any capability of computers or technologies to simulate human thought or brain activity. In a sense, artificial intelligence started early, with simple computer chess-playing programs and other programs that started to mimic human decision-making and thought.

Artificial intelligence continued to progress from the early days of the personal computer, to the age of the internet, and finally to the age of cloud computing, virtualization and sophisticated networks. Artificial intelligence has grown and expanded in many ways as a key technology industry.

One of the milestones in artificial intelligence is the emergence and adoption of machine learning, a particular approach to achieving artificial intelligence goals.

Machine learning uses sophisticated algorithms and programs to help computer software to get better at making certain sets of decisions in a performance environment. Instead of simply programming a computer to do one set of things over and over again, as was the case with the hand-coded programs of the 1970s and 1980s, machine learning starts to use heuristics, behavior modeling and other types of projections to allow the technology to improve its decision-making and evolve over time. Machine learning has been applied to fighting spam email, implementing artificial intelligence personalities like IBM Watson, and achieving artificial intelligence goals in other ways.

Deep learning, in turn, builds on machine learning. Experts describe deep learning as the use of algorithms to drive high-level abstractions, such as the use of artificial neural networks to train technologies on tasks. Deep learning takes machine learning to the next level by trying to model actual human brain activity and apply that to artificial decision-making or other cognitive work.

Deep learning has been demonstrated through examples such as cutting-edge supply chain optimization programs, laboratory equipment programs and other types of innovations such as the generative adversarial network, where two opposing networks, a generative and discriminative network, work against each other to model human thought processes of discrimination. This particular type of deep learning can be applied to image processing and other uses.

The reality is that deep learning drives artificial intelligence closer to what experts consider to be “strong AI,” artificial intelligence that is more or less fully capable of replicating many human thought functions. This gives rise to significant debate about how to handle these emerging technologies effectively, and how to care for a world in which computers think in some of the same ways that we do.