Artificial Intelligence, Machine Learning, and Deep Learning

[Qingquan-Li]

References:

• https://students.yourlearning.ibm.com/activity/ILB-PZXXWZEREVVQ28K2

Artificial intelligence

Artificial intelligence describes machines that exercise capabilities usually associated with human intelligence. Current research in AI focuses on learning, reasoning, problem solving, perceiving, and understanding human language. You’ll learn more in the next course about how advances in AI will help make us all smarter, communicate, solve problems for society, and change the way we work.

Machine learning

Machine learning, or ML, uses algorithms to learn from data. Given an input of data, ML can perform statistical analysis to determine an output. As with every kind of computing, the more data the machine is given (assuming that this data is valid), the more accurate is its output. ML uses human-like capabilities such as analysis, self-training, observation, and experience to learn without being explicitly programmed!

ML comes in three flavors: supervised learning, unsupervised learning, and reinforcement learning. ML relies on these three types of algorithms. The application of any one of these algorithms depends on the relation to available data you are processing, the output that you need from your model, or maybe even the possibility of retro-feeding data to improve the algorithm. If it sounds complicated, it is.

One amazing aspect of machine learning is its ability to modify itself when exposed to more data. It’s dynamic and doesn’t require human intervention to make changes! So as it’s exposed to more data, it continues to learn and improve its results.

Machine learning is what prompted the bank to contact you about what might be someone using stolen information from your credit card. But it’s not perfect. Machine learning is also what threw off your Netflix recommendations after someone in your family watched a weird comedy on your account.

Supervised learning

Supervised learning is a type of machine learning model that provides the machine with a set of highly accurate data that’s been labeled by a human. The machine uses this model to recognize related things in untrained data sets.

Unsupervised learning

Unsupervised learning is a type of machine learning model that doesn’t give the AI any labeled data. Instead it gives the AI unlabeled data, and the AI suggests various ways to cluster and organize it. This is valuable when the data is so large or complex that humans can’t identify its patterns themselves.

Reinforcement learning

Reinforcement learning is a type of machine learning model that doesn’t give the machine any data at all, labeled or unlabeled. Instead, the machine tries different actions and receives reward signals (like cookies for a dog!) when it makes correct moves. In this way the system is trained solve a problem, with no human work required.

Deep learning

Deep learning, or DL, is a subcategory of machine learning that focuses on statistical models when it solves problems. It uses an artificial neural network, made of algorithms inspired by the human brain, as it solves complex problems by performing tasks over and over again many thousand times, each time tweaking it a little to improve the outcome. DL requires big data and enormous computing power, but has tremendous potential as we move toward the goal of general AI.

Imagine the complexity of decisions to perform classification tasks directly from images, text, or sound with accuracy that might exceed human performance. How about driverless cars?

Deep learning is a key technology behind driverless cars, enabling the computer to recognize a stop sign or to distinguish a pedestrian from a street light.