Artificial Intelligence Explained in Simple Terms

• Read original article here

There are many great articles about Artificial Intelligence (AI) and its benefits for business and society. However, many of these articles are too technical for the average reader. I love reading about AI, but I sometimes think to myself, ‘Gee, I wish the author had explained this in simple English.’

I will try and explain AI and its related technologies in simple terms, using real-life examples, as though I were talking to someone at a party. Your colleagues or your (close) friends may tolerate your endless and complex ramblings, but I guarantee you that people at parties are far less forgiving.

After reading this, you will be able to talk about AI at a party andentertain your fellow guests, instead of making them hate the host for inviting you in the first place.

Artificial Intelligence (AI) involves using computers to do things that traditionally require human intelligence. This means creating algorithms to classify, analyze, and draw predictions from data. It also involves acting on data, learningfrom new data, and improving over time. Just like a tiny human child growing up into a (sometimes) smarter human adult. And like humans, AI is not perfect. Yet.

The difference between AI and regular programing? Regular programs define all possible scenarios and only operate within those defined scenarios. AI ‘trains’ a program for a specific task and allows it to explore and improve on its own. A good AI ‘figures out’ what to do when met with unfamiliar situations. Microsoft Word cannot improve on its own, but facial recognition software can get better at recognizing faces the longer it runs.

To apply AI, you need data. Lots of it. AI algorithms are trained using large datasets so that they can identify patterns, make predictions and recommend actions, much like a human would, just faster and better.

We interact with AI every day in our professional and personal lives:

Still, even the best AI today cannot match up to the human brain in some respects. While some AI is designed to mimic the human brain, AI today is only good at a relatively narrow range of tasks. AI can applymassive computing power to a narrow set of data and methods. The brain, on the other hand, applies medium computing power to a much wider set of data and methods.

Put differently, we can apply our brains to almost anything, while AI specializes in certain things.

Let’s briefly look at machine learning, deep learning, neural networks, evolutionary algorithms, and some real-world applications. Keep in mind that many real-world applications use more than one AI technology.

Machine learning algorithms identify patterns and/or predict outcomes. Many organizations sit on huge data sets related to customers, business operations, or financials. Human analysts have limited time and brainpower to process and analyze this data. Therefore, machine learning can be used to:

Many AI methodologies including neural networks, deep learning, and evolutionary algorithms, are related to machine learning.

Netflix says they “invest heavily in machine learning to continually improve our member experience and optimize the Netflix service end-to-end.”

Netflix applies machine learning to your viewing history to personalize the movie TV show recommendations you see. Netflix also analyzes what you and people with similar preferences watched in the past, and even auto generates personalized thumbnails and artwork for movie titles, to entice you to click on a title that you’d otherwise ignore.

All to ensure that you stay glued to the screen while your brain melts.

A neural network tries to replicate the human brain’s approach to analyzing data. They can identify, classify and analyze diverse data, deal with many variables, and find patterns that are too complex for human brains to see.

Deep learning is a subset of machine learning. When applied to a neural network, it allows the network to learn without human supervision from unstructured data (data that isn’t classified or labeled). This is perfect for analyzing ‘big data’ that organizations collect. These big data sets include different data formats such as text, images, video and sound.

Neural networks are frequently combined with machine learning, deep learning, and computer vision (training computers to derive meaning from pictures). That’s why people talk about ‘deep neural networks,’ which is basically a neural network with more than 2 layers. More layers = more analytical power.

Deep neural networks can be trained to identify and classify objects. A cool use is facial recognition — identifying unique faces in photos and videos. Neural networks also learn over time. For instance, they get better at classifying objects and identifying faces as they are fed more data.

China is doing a lot with facial recognition. Which makes sense, because there are cameras everywhere in China. Many cameras mean plenty of data for deep neural networks to use. In the interest of time, here are three examples.

A university in Eastern China has implemented an AI-powered attendance system, with cameras that constantly observe students in class.

Naturally, it scans faces to check that the student actually turns up to class. More importantly, it also analyzes facial expressions in real time, and can judge whether students are paying attention. It can apparently recognize people are sleeping or playing on their phones. I’ll bet you’re thinking ‘I’m glad they didn’t have this tech at my university…’

While people are still talking about mobile payments using WeChat Pay and Alipay, these companies are already moving on to the next phase: face-based payments. Why bother with your phone when you can pay by looking into a camera?

Alipay’s Dragonfly facial recognition system has expanded to over 300 cities in China. WeChat Pay also has a similar system. Businesses from bakeries to supermarkets have adopted these systems to speed up customer payments. After a face is scanned, money is deducted from the customer’s Alipay or WeChat Pay account. Businesses benefit from spending less on cashier staff.

Face-based payments also benefit less tech-savvy citizens such as elderly people. Also, only 60% of China’s 1.4bn population is connected to the internet and only 40% of the population pays with smartphones. In future, we can even expect these unconnected citizens to use face-based payments if their face can be linked to a bank account.

AI firms in China are using facial recognition to help law enforcement track suspects.

State-backed Cloudwalk’s facial recognition tech has helped police make over 10,000 arrests and is used in 29 Chinese provinces. It makes over 1 billion facial comparisons daily and has amassed over 100 billion data points.

SenseTime made the news recently for using facial recognition to identify one suspect in a crowd of 50,000 concertgoers in China.

Watrix, another computer vision firm, has apparently decided that facial recognition is old news. Their tech helps police track people by the way they walk and move, even if their face is hidden. Also, you can’t fool their tech by adding a limp or altering your gait.

A subset of machine learning, evolutionary algorithms self-improve over time. They create a population of algorithms and preserve the ones most successful at predicting outcomes. Applying the ‘survival of the fittest’ principle, the best algorithms are kept alive and the losers are discarded. Sections of code from the winning algorithms are used to create a new population of algorithms, and the selection process repeats.

Evolutionary algorithms are well suited to optimization tasks where there are a lot of variables and a dynamic environment. Basically: find a way to the best possible result.

Evolutionary algorithms can be built into neural network models to pick stocks and identify trades. Trading rules are set up as parameters and the algorithm works to maximize trading profit. Small changes are introduced into the model over time, and the changes that have the largest desirable impact are kept for the next generation. The model improves with time.

These trading models are popular across institutional quantitative traders. Individuals can also access these models through software packages on the market.