What is the difference between Artificial Intelligence and Machine Learning?
As an AEC domain expert, how much time have you spent on the Internet searching for a simple and non-technical definition of AI and ML? Have you felt frustrated, confused, and disappointed because you ended up in a sea of conflicting and technical jargon? Learning about Artificial Intelligent (AI) and Machine Learning (ML) can be disheartening. Even with tons of online technical resources, there is too little content written for the AEC domain experts. So, after seeing the impact of AI and ML in strengthening and transforming other industries, how do you educate yourself about the benefits of AI and ML for the AEC industry?
To simplify the basic context of AI and ML applications for the AEC industry (Architecture, Engineering, and Construction), I started a series of posts about AI-assisted technologies for the AEC domain experts.
Learning about AI and machine learning on the Internet is confusing and frustrating for AEC professionals.
What are AI & ML?
Today, most people use the terms AI and ML interchangeably; mainly because ML is the most common and powerful method of AI. As a result, the differentiation between AI and ML can be ambiguous and difficult for those who aren’t AI experts. So, what is the difference?
A straightforward way of thinking about it is all machine learning methods are AI methods, but not all AI methods are learning-based. Machine learning is a subset or a branch of AI.
All ML methods are AI-based but not all AI methods are learning-based.
Let’s think about AI as a tree; a tree with roots in philosophy, logic and mathematics, computation, psychology, cognitive science, neuroscience, and evolution. These roots provide the nutrition and water necessary for the tree to grow. The branches of this tree are various AI methods, such as machine learning, computer vision, robotics, expert systems, and natural language processing. Each branch also contains multiple sub-branches. The AI trunk of this tree stands firm, supporting the branches, and carrying nutrients from the scientific roots into the branches and sub-branches. One of the branches of this AI tree is ML with three sub-branches, namely, supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. These are the methods that enable a machine to learn to recognize or predict based on data. Even though these methods are sub-branches of machine learning, they are still part of the AI tree, and therefore, are known as AI methods.
AI is like a tree with roots in philosophy, mathematics, and other sciences, and ML is a branch of the AI tree.
But what is AI? A human acts or processes actions by acquiring knowledge and understanding through thought, experience, and the senses. To program computers to observe, analyze, learn, and process like a human, researchers use techniques that imitate human intelligence and cognitive behavior. In some cases, a system is programmed to mimic other species’ behavior, like ants or dogs. This branch of science is called Artificial Intelligence (AI), which refers to the original concept of “Can machines think?” proposed by Alan Turing, a young British polymath in 1950.
Learning is one branch of AI. Commonly, in the machine learning branch, we feed data to the computer, so it can learn the relationships and rules to make predictions or classifications. Let’s look at a tangible example.
To automate and simplify safety tracking on construction sites, several companies have been developing AI-assisted technologies that process the photos or videos captured from a construction site and report potential safety issues. These technologies are mostly based on artificial neural networks, which are ML algorithms, and can analyze the images or video feeds to determine safety risks much more quickly than a human. The way it works is straightforward. They teach or train the algorithm by providing a lot of images with the safety requirements labeled or tagged. Once the system learns to recognize the safety requirements, it can predict potential safety risks in real-time on the sites.
Summary
As we discussed, all machine learning methods are AI methods, but not all AI methods are learning-based. Machine learning is a subset, or a branch, of AI. Back to our tree metaphor, to find out which branch of this AI tree you need, you should first identify which business problems you are trying to solve using AI. Most ML methods work well when learning a simple concept with lots of data available. I’ll discuss these learning methods in the upcoming articles.
Resources
https://www.csee.umbc.edu/courses/471/papers/turing.pdf
