“Artificial intelligence” is not only a term used in sci-fi movies, but also one of the hottest buzzwords in today’s business world. In recent years we have seen quite some AI-related news such as:
- AI-powered Alpha Go beats the best human chess player;
- AI robots help surgeons carry out surgeries much easier;
- Google’s AI Assistant sounds like a human;
… Ok hold on, we all know AI is the trend, but… what is AI?
Non-technical professionals keen to understand AI might google this question but find it hard to obtain a simple and clear answer. That is why we decided to write this post, which breaks down the concept of AI, what it can do, and its typical applications, in the most non-technician-friendly way.
What is AI?
In short, AI is used to describe any machine that can mimic human behaviors. By processing large amounts of data and recognizing patterns, machines can perform specific tasks, which were usually done by humans like speaking, seeing, moving around, etc. Based on the type of human ability to imitate, AI can be roughly classified into two genres.
- Humans can see through their eyes and interpret what they see. The equivalence of this in AI is called computer vision.
By imitate human’s vision system, such technology enables machines to process and identify objects in images and videos, just like we do.
Depending on which part of the vision capability machines try to mimic, computer vision can be further divided into certain subcategories. Some of the most important ones are:
- Object recognition – What is in the picture?
- Object classification – Which category does the object in the picture belong to?
E.g. Is the animal in the picture a dog?
- Object detection – Where is the object in the picture?
E.g. Where is the dog in the picture?
- Object tracking– how the object in the video has moved?
E.g. Where did the dog run to in the video? … Some classic applications of computer vision include:
- Facial recognition – Banks use facial recognition to verify customer’s identities and therefore prevent fraud.
- Search by image – E-commerce giants like Taobao allows users to search for wanted products by photos.
- Self-driving cars – By identifying different objects in the environment, the car can navigate the right path on its own.
- Humans can use language to communicate thoughts and understand each other. The similar ability of AI is achieved by a technology called Natural Language Processing.
With this technology AI can talk with a human back and forth, to deal with our queries or perform a task we require, like we can ask the virtual assistant in the phone what the weather is tomorrow, or open an app. Some of the subcategories of natural language processing include:
- Automatic Speech Recognition, which helps machines to transcribe spoken words into text. We can think of it as the human’s ear.
- Natural Language Understanding, which enables machines to process and understand the meaning of words. It is similar to the part of our brain that interprets what other people say.
- Natural Language Generation, used by machines to transfer data into natural language. It is equivalent to the process when humans turn thoughts into speech or writing.
- Sentiment Analysis, based on which machines can figure out what topics have been discussed relating to a certain issue, and whether they are negative or positive.
Some common usage of Natural Language Processing is:
- Voice assistant – For instance, Google Assistant , Alexa, Siri, etc.
- Chatbot – Today many companies use chatbot to deal with customer inquires, promote products, etc. Compared with human agents chatbot is scalable, 24/7 available, and much cheaper.
- Voice AI agents – Similar to chatbot, voice AI agents can also be used to automate customer service, product promotion, verifying information, etc.
The difference is that while chatbot sends customers messages on the screen, voice AI “talks” with customers directly over the phone. Therefore, besides responding “to” customers immediately, voice AI can also get instant response “from” customers. This makes the voice AI one step ahead compared with chatbot when getting customer feedback is crucial.
AI vs Machine Learning vs Deep Learning
“Machine Learning” and “Deep Learning” are two terms that often appear together with “AI”. Sometimes the three are even used interchangeably, but what is the difference among these popular concepts?
Simply put, Machine learning is the foundation of AI. The essence of AI is to have machines performing human tasks. While humans need to learn before carrying out a task, so do machines. That is where machine learning comes from.
Before asking machines to make predictions and finish a task, we need to train it (by an algorithm) with large amounts of data derived from historical context. For instance, for voice AI to talk with customers on a certain topic fluently, it needs to learn from perhaps millions of similar real-life conversations beforehand. The more (high-quality) data the machine is exposed to, the better it learns, and the better quality decision it can make.
That is also the key that separates AI from other traditional softwares– while the former improves its own performance as more data comes in, the latter is more “static” and can only be updated by engineers.
Deep learning is simply a type of “advanced” machine learning. It is inspired by the way human’s brain functions. With something called neural networks, it improves its performance as time passes by. Object recognition, for instance, is achieved by deep learning.
- AI is the science that allows machines to replicate human ability.
- The two biggest genres of AI are Computer Vision and Natural Language Processing, the former resembles human’s vision system while the latter mimics human’s language ability.
- Machine learning lies the foundation of AI, it is the process in which machines learn from previous data and make predictions in a new environment. The more high-quality data the machine is exposed to, the better it performs.
- Deep Learning is a subset of Machine Learning that is inspired by how human’s brain works.