How much do you actually know about Artificial Intelligence (AI)?
We often think that AI will be taking over the world, replacing us at our jobs and making us obsolete. Or when we think about AI, we imagine a world with robots everywhere.
But other than our imaginations and what movies often display, what is AI really all about?
In this episode, we discuss a few introductory topics to what AI is:
- Will AI really take over all our jobs?
- How difficult it is to learn about AI?
- Artificial Intelligence versus Machine Learning
- How can businesses adopt AI to make their work more efficient?
What is Artificial Intelligence (AI) all about?
Dr Lau Cher Han: So, let me ask you a question, what makes you think that AI will be taking over the world? For example, before we had cars, what did people use to commute?
Dr Lau Cher Han: Horses! Before we had horses, what did we use?
Reuben: We walked?
Dr Lau Cher Han: We walked, exactly. When humans first discovered horses, do you think they were worried about horses taking over the world?
Reuben: Probably not.
Dr Lau Cher Han: Exactly, probably not. Then why up till today, when we discovered that AI can help us do a lot of mundane tasks in our daily lives, why are we so worried about it taking over the world?
Reuben: I think it’s more because of the media, things we read online, things we read on facebook etc; We think it in the way where AI is coming, our jobs are going to be gone, bankers are going to be gone, lawyers are going to be gone and things like that.
So I guess these are the things that attribute to people thinking that AI will wipe out the human workforce one day.
Dr Lau Cher Han: In most of the jobs, there are two parts. There are technical parts and also the creative parts. AI can help us with the technical parts. This means that AI can help us write articles, but it cannot help us to write beautiful, quality articles that have the same style.
It can copy the way a person write the article, but it will not be able to copy his style or come up with a better style, at least at this stage, where there is a limited knowledge on AI application.
Reuben: Do you think one fine day, in the future, that day will come?
Dr Lau Cher Han: Yes, but that is given on an assumption that humans are not advancing at the same time.
Let’s say, back to our previous example, horses help us save time on commuting, and with that, we can use that time to do something else.
As AI advances, we are advancing together at the same time. For example, in the law and legal industry, there are a lot of processes and tasks that may seem mundane or tiresome, not to make fun of my lawyer friends out there. But how AI comes in to help us, is by automating these mundane tasks, while still keeping the interesting and fun parts for humans to do.
Reuben: So what we should worry about is not having the knowledge of AI.
Dr Lau Cher Han: Correct. That is a very good point. Many people fear the use of AI because they don’t have much knowledge about AI applications.
Reuben: This reminds me of another example. Fear comes when we do not know something. Like when you go for a big presentation in front of a crowd, when you do not know certain domains, you will be more afraid. So fear comes from not knowing something.
What is the best way for someone to break into AI?
Reuben: I think it is really important for people to know and learn about AI. So for a layman’s perspective, for someone who does not know anything, what is the best way to get into AI?
Dr Lau Cher Han: So if you refer to previous episodes of Data Crunch, I always talked about used cases or applications. No matter where I go or when I am giving workshops, I always tell people that you must tie a particular technology to real-life application.
So when you want to start AI, look for the boring bits in your industry.
Let’s not make fun of lawyers anymore, but instead, look at other industries. Let’s say, digital marketing for you, Reuben. I believe there are many boring bits that you don’t really like to do, or that it is so repetitive that you hope one day the machine can take over for you. Is there any?
Reuben: Of course! Definitely! Like creating creative copies that are repetitive, or copy and pasting tasks etc.
Dr Lau Cher Han: I will give you an example, let’s say you are doing something related to cooking. What if I come up with an AI bot that can help you to collect 10,000 swipe found that is related to the information you put in. Wouldn’t that be helpful?
Reuben: Easily helpful!
Dr Lau Cher Han: Yeah, then that will be a good starting point.
AI is in many other industries, I am sure you have seen it a lot in gaming, like Super Mario, Pacman or even the recent Assassin Creed or Car Racing games, AI is everywhere.
Reuben: So it has been around for some time, but in recent years, people started talking about it and it becomes a hot, trendy topic again.
Speaking about AI, that reminds me of the movie that you mentioned earlier, The Imitation Game.
Dr Lau Cher Han: So I wouldn’t go into too many details of the movie, but some people say that Apple’s logo was inspired by Alan Turing.
Alan Turing was one of the first or earliest in the data science field, and he was considered the father of artificial intelligence with his famous Turing Test. If the system passed the Turing Test, then it is considered as an AI machine.
Reuben: So the Turing Test is used until today?
Dr Lau Cher Han: Yes, it’s still very famous and being used until today.
Reuben: If you haven’t watched the movie, you really should. Watch it and get inspired!
Dr Lau Cher Han: Yes, for those people who don’t really understand what is AI and machine learning, it gives you an insight or a very gentle introduction about it. This will begin a new chapter in your journey as well.
What is the difference between AI and Machine Learning?
Reuben: AI versus Machine Learning, what is the difference? Many people get confused by this two terms. So what is the difference?
Dr Lau Cher Han: I will show you a chart, it will show you the difference between AI and machine learning and also deep learning.
Dr Lau Cher Han: Have you ever wondered, why out of a sudden, people are starting to talk about machine learning, big data etc. Maybe not many people talk about machine learning, but a lot of people talk about big data, deep learning and AI.
You probably have heard these terms quite often, especially in the past two to three years. So have you ever thought about why is it so?
Reuben: Never thought about it really, I always think its the media that is pushing all these.
Dr Lau Cher Han: Stop always pushing the blame to media, it’s always the media, point taken, haha!
So the media has been pushing a lot of big data. So to give you a brief introduction, big data usually refers to the data that the current systems, algorithms, software applications are unable to process them, because of the structure, speed, volume, etc.
In layman’s term, we call it Big Data. Machine learning is a subset of AI. I am giving you a very detailed explanation here, AI is the application of machine learning algorithms. Machine learning is a set of algorithms and techniques that we use.
So I’ll give you an example. Today if I can teach a machine how to recognise my face, that is the machine’s learning technique or machine learning algorithms. So the machine can recognise my face and emotions.
If I use it in a shopping mall, I can use it to determine my shoppers whether they are happy or not. But, if this machine is used in primary school or secondary school, the teachers would use it to take attendance. So that is the difference.
The attendance taking tool and the emotion sentiment analysis based on your facial recognition, that itself is called an AI application. That is where AI is.
Anything that we teach the machine about our physical world, is machine learning. And then the use of those machine learning algorithms and techniques, it will be considered as AI applications.
If I have zero knowledge, how can I build an AI application?
Reuben: So do tell it to me like I am someone who does not know anything about technology at all. How hard is it for me to do something like that? Is there readily made algorithms that can be plucked out from the internet? Say, for example, to do something like what you said, to recognise the faces of my students in a class.
Dr Lau Cher Han: So there are a lot of algorithms, libraries, frameworks out there which is ready to download since they are open sourced and free. You can just download those source codes and run it on your own.
Number one, the good thing is that it does not require a lot of understanding about mathematics, or hardcore programming, statistics knowledge just to get it to work. You just need to understand the application you want to build.
Number two, you don’t have to go for those shiny tools. There are a lot of things that talk about Tenser Flow, Kara, Torch etc, those are deep learning. Those are very advanced things that can go very far, such as building a machine that can play chess with humans etc. Yes, you can go to that level, but as a beginner, you do not have to. You should start with very small applications and learn how it works in real life.
Reuben: Ok, so speaking about that, those are some applications that we can use to detect faces, even on your phones as well, when you unlock your phones with facial recognition, that is also AI right?
Dr Lau Cher Han: Yes, that is also one of the applications of facial recognition as well.
How are some of the Malaysian businesses using AI?
Reuben: We talk about bringing AI to Malaysian real-life cases, such as businesses in Malaysia. How are some Malaysian businesses using AI to power their business?
Dr Lau Cher Han: I think in the past two to three years, we start to see a lot of companies and corporations adopt this thing called chatbots. A chatbot is a very interesting use of machine learning, plus NLP, which is Natural Language Processing.
Everything that we want to process, whether we want to do machine learning or data science, it has to be in the form of numbers. So in order to convert tax information into numbers, we need something called NLP.
Once we convert text into numbers, there are a lot of interesting things we can do. We can teach the machine how to understand the sentiments, so you can understand how things work.
Politicians like to use this tool a lot. They use it to measure the interestingness of a particular topic and whether the public is paying enough attention to certain things, be it in education, hospitals, public transport etc. That is one of the applications.
Now we take it to the next level, where the chatbots come in. At the moment we have a lot of failed cases because they only use chatbot as a one-way tool, or as another channel or they do not train their chatbots enough.
Reuben: But I’ve seen chatbots in Malaysia, especially by banks.
Dr Lau Cher Han: I think Hong Leong is one of a good example.
Reuben: Yeah, and I’ve seen some of the chatbots that are able to understand our local slangs. So when you type in things like, “where is my account lah?”, they know what you’re talking about. So all those involves machine learning?
Dr Lau Cher Han: Yes, you are teaching the machine to understand our local slangs.
Reuben: Companies that do that charge tons of money for that. It must be something complex to do?
Dr Lau Cher Han: Actually it’s not really true, because it is unknown to many people, so we will think “oh it must be very expensive”.
But in fact, you can set up a chatbot for free, but it takes a long time for you to gather those data, such as the previous chat records, what are the good data that you’re going to fit in, in order to train the model.
The thing in the machine learning model is, “garbage in garbage out”. If you don’t feed it with good quality data, it will not work. For example, like you give your car good petrol, good fuel, your car can run and perform very well.
But if in machine learning, you give the data and the data is rubbish or messy, then you’re in trouble.
If I were to build a chatbot, where is the easiest way to get some data?
Reuben: One very quick question before we end the session today. When it comes to chatbots, in regards to learning and feeding all the data in, what is one very good technique for getting that in?
Let’s say I want to build a chatbot today, where do I get all this data of people’s languages, the way they talk online, or the way they type online? So what is one source where I can take that?
Dr Lau Cher Han: Twitter!
Twitter would be your best friend because if you follow let’s say 500 to 1,000 twitter active Twitter users, especially if they are from Malaysia, then you can learn the way that they speak.
So you would have a standard English dictionary that they write in, but at the same time you would have a lot of words that are outside the dictionary, and this is where you can pick up these things.
Like how when we type in Malay, we type “Xde” or “tak ade”, and there are a lot of variations to it. With machine learning techniques, combined with NLP techniques, you are able to pick them up, and you can use that as training data for your chatbots. That would be a very good starting point.
Reuben: So if I want to build a chatbot today, I would be going on to Twitter and mine the text out?
Dr Lau Cher Han: Correct. Those are publicly available information true API, so you can easily get them.
We hope this has answered some of your questions on artificial intelligence, as well as give you a better understanding of what it is all about. If you have other questions about AI, be it from a beginner’s perspective, or if you are someone who is considering applying AI into your work and not sure how to do so, you can just leave a comment in the comment section down below.
Do subscribe to us on Youtube, where we come up with content like this every single week, and also go and like our facebook page for more helpful discussions about AI, Big Data, Cloud Computing and Data Science.