Unravelling AI, ML and DL
Unravelling AI, ML and DL
Technological developments are happening so fast that it is often difficult to keep up to date with the latest terminology let alone the implications. Differentiating between artificial intelligence (AI), machine learning (ML) and deep learning (DL) is the latest hurdle to be faced.
30 March 2019
What is artificial intelligence (AI)?
Artificial intelligence refers to machines that can follow a provided set of rules to solve a problem. These rules are often referred to as algorithms. To put it another way, algorithms are basically specific tasks given to computers which they carry out and which result in a desired outcome. They can be used to get machines to do specific and usually repetitive tasks such as picking boxes off an assembly line in a factory. There are two types of artificial intelligence; Narrow AI and General AI.
Narrow AI refers to intelligent systems that have been taught to carry out simple repetitive tasks, for example, factory machines packing fruit into boxes. General AI, on the other hand, refers to adaptable intelligent systems that are capable of learning and carrying out a variety of different tasks. Researchers have been able to develop many instances of Narrow AI and are working on the more complicated task of developing General AI.
It’s important to realise that AI is often used as an umbrella term encompassing machine learning and deep learning. In other words, AI = ML and DL. However, ML and DL are different kinds of processes.
What is machine learning (ML)?
ML basically involves providing machines with a limited dataset so that they can learn some facts which enable them to make some basic decisions while carrying out a set of steps. For example, machines are given information on square tins and round tins, and then asked to follow a set of rules which involves separating the square tins from the round tins. The learning involves identifying and differentiating between round and square tins based on the input data.
What is deep learning (DL)?
DL involves providing machines with larger quantities of data and more complex algorithms (sets of rules), which require them to sort through and classify the data they have received and then make their own decisions based on their own classifications. This is a more complex task, so the machines are given several layers of neural networks (or brain functionality) which provide them with the enhanced capacity to sort through and classify the data they have been given. In other words, DL is trying to replicate how the human brain works.
What’s so special about AI, ML and DL?
To put it simply, AI, ML and DL hold the key to improved data analysis and of course improved geospatial data analysis. Ever-increasing volumes of data are being captured by a range of sensors, including satellites and drones, on a daily basis and to make intelligent use of this data, we need to be able to process it. AI, via ML and specifically DL, are the solution to classifying and analysing this collected data so that humans will be able to derive benefit from it. However, we need to teach the machines how to assess and analyse the data, and in order to do this they need to learn how to be adaptable, intelligent systems. At least that’s the plan.