What’s the difference between Machine Learning and Deep Learning?
Artificial intelligence, machine learning, robots, internet of things, smart devices… The probability of a day passing by without running across at least one of those words is not so high. You try to keep up, but it’s not so easy, especially if you’re not a software engineer. Luckily, we are here to simplify these concepts so that everyone can understand them.
Let’s try to make it simple – machine learning and deep learning are already all around us. For example, YouTube knows what videos you might like even if it’s from a channel and a topic you’ve never searched before.
Similarly, customer service agents will know what is the best solution to a customer even before opening the support request.
Let’s go a bit deeper and explain things in more detail.
What is machine learning?
If you google machine learning, the first result will be a Wikipedia article talking about algorithms that take decisions after learning from different sets of data.
A great example is an AI you’re using daily without even being aware of it – SPAM filters. Email spam filters have to use machine learning for one simple thing – humans are innovative. If we add a simple rule that every time Nigerian prince is mentioned in an email from an anonymous source it is automatically sent to the spam folder, scammers will easily trick the filter. That’s why the AI behind spam filters must learn to recognize new categories of SPAM emails from various signals and act accordingly.
Machine learning is all around us even though we’re not aware of it. Almost every industry is using some sort of AI and the applications go far beyond SPAM filters.
All things considered, machine learning is nothing more than a tool that serves its purpose made with a bunch of lines of code sticked together behind it. Translated to a more conversational language, machine learning could be described as a software that performs a function based on data with the addition that the more data is available the higher the performance of the AI will be. Don’t worry, it sounds more complicated than it actually is.
Let’s talk about how machines can learn and how deep learning works in practice.
Deep learning vs machine learning
Deep learning is a sub-discipline of machine learning. Technically, they are the same thing working in a slightly different way; with the main difference being the extent of their capabilities.
Applying Machine learning to a function means that the function provided will optimize (get better) over time, but to do so it needs to be told in what way it should improve.
On the other hand, Deep learning means that the machine can jump to its own conclusions and use the data in order to analyze whether the provided function represent the best possible option to solve the problem at hand.
Let’s consider the email filters again. It will gather data and get better at categorizing what is spam, what is important and what is promotional email. However, good copywriters know how to avoid these triggers and will try to sell you something without mentioning the sale in the email itself.
The deep learning filter will be able to distinguish the hidden scam while a “simple” machine learning application would fail.
How does deep learning work?
The deep learning model is programmed to follow a certain logic and to do so it needs to analyze tons of data. For this purpose it normally uses a network of algorithms inspired by the network of synapses found in the human brain. It might seem strange that machines with artificial neurons can learn, but the results are way better than those of the standard machine learning.
So far, it is the closest thing we have to the truly intelligent machines, and it is probably the right path to take if we ever want to achieve it.
Let’s take the example of a program created to play chess. Every time a human will challenge it, the machine will progressively get better until the moment no human will ever be able to defeat it.
The machine not only it can learn tricks from other players but it can also come up with new moves it was never programmed to perform in the first place..
Let’s list the main differences between Machine Learning and Deep Learning:
- Machine learning uses algorithms to gather data, learn from it, and make decisions based on what it has learned
- Deep learning structures those same algorithms in layers to make a “neural network” that makes autonomous decisions
- Deep learning is part of machine learning, and even though they are almost the same, deep learning is the closest we have today to true artificial intelligence.
Deep Learning in a nutshell
If you’re still staring blank at the screen trying to figure out what the heck are we talking about, don’t beat yourself up. It is a complicated subject.
In a nutshell, Deep learning allows machines to take the right decision even though they weren’t specifically programmed to do so.
An analogy to be excited about
While AI and deep learning are not magic, nor a sentient being making its own decisions, there are still amazing things that are about to happen in this field and it’s absolutely ok to be excited about it. We live in the era of the Big Data. All the technology around us provides tremendous amounts of data that can be used to train Artificial Intelligence applications.
A great analogy comes from Andrew Ng who shared on an interview with Wired Magazine. He basically compared AI to a rocket-ship. A rocket-ship needs a large engine and vast amounts of fuel. A Large ship without enough fuel cannot reach the orbit, while a ship with unlimited fuel but a small engine cannot take off.
In this analogy, deep learning is actually that big engine we need, and the fuel is all the data at its disposal. We can surely hope for major breakthroughs and innovations in the coming future as the amount of data to train AI algorithms is only going to increase.
So what do machine learning and deep learning mean for customer service?
The majority of AI tools in customer service use algorithms powered by machine learning. Their goal is to introduce self-service, increase the efficiency of agents and prioritize tasks in handling tickets. All the data coming from both customers and customer support agents means Deep Learning tools for customer service can be extremely accurate.
Jatana’s automatic replies, for example, are powered by Deep Learning and are so accurate you will prefer the machine talking to the customer rather than a real person.