Artificial Intelligence (AI) and Machine Learning (ML)

AI is changing the way we work and live and we see a lot of examples of how AI is affecting the society and impacting our lifestyle. But what really is AI? Let us understand that first before we delve into how Machine Learning (ML) is used in AI.

You've probably seen news articles about how much value AI is creating. According to a study by McKinsey Global Institute, AI is estimated to create an additional 13 trillion US dollars of economic output annually by the year 2030, boosting global GDP by about 1.2 percent a year. Even though AI is already creating tremendous amounts of value into software industry, a lot of the value to be created in the future lies outside the software industry. In sectors such as retail, travel, transportation, automotive, materials, manufacturing and so on. I should have a hard time thinking of an industry that I don't think AI will have a huge impact on in the next several years. 

As per Wikipedia, Artificial Intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and animals. A more vivid definition will be any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Colloquially, the term "artificial intelligence" is often used to describe machines (or computers) that mimic "cognitive" functions that humans associate with the human mind, such as "learning" and "problem solving".

There are 2 types of Artificial Intelligence:

-      Artificial Narrow Intelligence (ANI)

-      Artificial General Intelligence (AGI)

Almost all the progress we are seeing in the AI today is Artificial Narrow Intelligence. These are AIs that do one thing such as a smart speaker or a self-driving car or AI to do web search or AI applications in farming or in a factory. These types of AI are one trick ponies but when you find the appropriate trick, this can be incredibly valuable. 

Unfortunately, AI also refers to a second concept of AGI or artificial general intelligence. That is the goal to build AI. They can do anything a human can do or maybe even be super intelligent and do even more things than any human can. 

We can see tons of progress in ANI, artificial narrow intelligence and almost no progress to what AGI or artificial general intelligence. Both of these are worthy goals and unfortunately the rapid progress in ANI which is incredibly valuable, that has caused people to conclude that there's a lot of progress in AI, which is true. 

But that has caused people to falsely think that there might be a lot of progress in AGI as well which is leading to some irrational fears about evil clever robots coming over to take over humanity anytime now. 

I think AGI is an exciting goal for researchers to work on, but it'll take most for technological breakthroughs before we get there and it may be decades or hundreds of years or even thousands of years away. Given how far away AGI is, I think there is no need to unduly worry about it. 

The rise of AI has been largely driven by one tool in AI called Machine Learning (ML). So what is Machine Learning? As per Wikipedia, Machine learning is the study of computer algorithms that improve automatically through experience. It is seen as a subset of artificial intelligence. Machine learning algorithms build a mathematical model based on sample data, known as "training data", in order to make predictions or decisions without being explicitly programmed to do so.

The most commonly used type of machine learning is a type of AI that learns A to B, 

or input to output mappings. This is called supervised learning. 

Let's see some examples. If the input A is an email and the output B is whether email is spam or not (0 or 1), then this is the core piece of AI used to build a spam filter. Another example is if you want to input English and have it output a different language, Chinese, Spanish, or something else, then this is machine translation. The most lucrative form of supervised learning, of this type of machine learning, maybe be online advertising, where all the large online ad platforms have a piece of AI that inputs some information about an ad, and some information about you, and tries to figure out, will you click on this ad or not? By showing you the ads you're most likely to click on, this turns out to be very lucrative. Maybe not the most inspiring application, but certainly having a huge economic impact today. 

A larger example could be if you want to build a self-driving car, one of the key pieces of AI is in the AI that takes as input an image, and some information from their radar, or from other sensors, and output the position of other cars, so your self-driving car can avoid the other cars. 

In manufacturing, you take as input a picture of something you've just manufactured, such as a picture of a cell phone coming off the assembly line and you want to output, is there a scratch, or is there a dent, or some other defects on this thing you've just manufactured? This is visual inspection which is helping manufacturers to reduce or prevent defects in the things that they're making. 

This set of AI called supervised learning, just learns input to output, or A to B mappings. On one hand, input to output, A to B it seems quite limiting. But when you find a right application scenario, this can be incredibly valuable. 

The most important idea in AI has been machine learning, has basically supervised learning, which means A to B, or input to output mappings. What enables it to work really well is data. The more data you have the more accurate predictions can be done and the performance will be better. In other words, AI will be able to predict more accurate results if we use a larger set of “training” data to make the system act on any scenario. The more scenarios are added, the better the performance. When you add neural networks and deep learning to machine learning, the results will be even better.  

AI is currently being used in a lot of day to day applications like speech recognition, online advertising, building self-driving car, where having a high-performance, highly accurate is important. The researchers are working on making this technology fool proof and it is not far away when we will see self driven cars moving around with us on the roads and perhaps driving better than humans!