Reviewing automation
This section briefly discusses the idea of automation, why we need it, and how it applies to machine learning. We will also answer the age-old question of machine learning replacing humans in their jobs, and the role of automation in that regard.
Automation plays a huge role in the modern world, and in the past centuries it has allowed us to completely remove the human factor from dangerous and repetitive jobs. This has opened a new array of possibilities on the job market, where jobs are generally based on something that cannot be automated, at least at this point in time.
But first, we have to understand what automation is.
What is automation?
There are many syntactically different definitions out there, but they all share the same basic idea. The following one presents the idea in the simplest terms:
The essential part of the definition is the minimization of the human input. An automated process is entirely or almost entirely managed by a machine. Up to a couple of years back, machines were a great way to automate boring, routine tasks, and leave creative things to people. As you might guess, machines are not that great with creative tasks. That is, they weren't until recently.
Machine learning provides us with a mechanism to not only automate calculations, spreadsheet management, and expenses tracking, but also more cognitive tasks, such as decision making. The field evolves by the day and it's hard to say when exactly we can expect machines to take over some more creative jobs.
The concept of automation in machine learning is discussed later, but it's important to remember that machine learning can take automation to a whole other level. Not every form of automation is equal, and the generally accepted division of automation is into four levels, based on complexity:
- Basic automation: Automation of the simplest tasks. Robotic Process Automation (RPA) is the perfect example, as its goal is to use software bots to automate repetitive tasks. The end goal of this automation category is to completely remove the human factor from the equation, resulting in faster execution of repetitive tasks without error.
- Process automation: This uses and applies basic automation techniques to an entire business process. The end goal is to completely automate a business activity and leave humans to only give the final approval.
- Integration automation: This uses rules defined by humans to mimic human behavior in task completion. The end goal is to minimize human intervention in more complex business tasks.
- AI automation: The most complex form of automation. The goal is to have a machine that can learn and make decisions based on previous situations and the decisions made in those situations.
You now know what automation is, and next, we'll discuss why it is a must in the 21st century.
Why is automation needed?
Both companies and customers can benefit from automation. Automation can improve resource allocation and management, and can make the business scaling process easier. Due to automation, companies can provide a more reliable and consistent service, which results in a more consistent user experience. As the end result, customers are more likely to buy and spend more than if the service quality was not consistent.
In the long run, automation simplifies and reduces human activities and reduces costs. Further, any automated process is likely to perform better than the same process performed by humans. Machines don't get tired, don't have a bad day, and don't require a salary.
The following list shows some of the most important reasons for automation:
- Time saving: Automation simplifies daily routine tasks by making machines do them instead of humans. As the end result, humans can focus on more creative tasks right from the start.
- Reduced cost: Automation should be thought of as a long-term investment. It comes with some start-up costs, sure, but those are covered quickly if automation is implemented correctly.
- Accuracy and consistency: As mentioned before, humans are prone to errors, bad days, and inconsistencies. That's not the case with machines.
- Workflow enhancements: Due to automation, more time can be spent on important tasks, such as providing individual assistance to customers. Employees tend to be happier and deliver better results if their shift isn't made up solely of repetitive and routine tasks.
The difficult question is not "do you automate?" but rather, "when do you automate?" There are a lot of different opinions on this topic and there isn't a single right or wrong answer. Deciding when to automate depends on the budget you have available and on the opportunity cost (the decisions/investments you would be able to make if time was not an issue).
Automating anything you are good at and focusing on the areas that require improvement is a general rule of thumb for most companies. Even as an individual, there is a high probability that you are doing something on a daily or weekly basis that can be described in plain language. And if something can be described step by step, it can be automated.
But how does the concept of automation apply to machine learning? Are machine learning and automation synonymous? That's what we will discuss next.
Are machine learning and automation the same thing?
Well, no. But machine learning can take automation to a whole different level. Let's refer back to the four levels of automation discussed a few of paragraphs ago. Only the last one uses machine learning, and it is the most advanced form of automation.
Let's consider a single activity in our day as a process. If you know exactly how the process will start and end, and everything that will happen in between and in which order, then this process can be automated without machine learning.
Here's an example. For the last couple of months, you've been monitoring real-estate prices in an area you want to move to. Every morning you make yourself a cup of coffee, sit in front of a laptop, and go to a real estate website. You filter the results to see only the ads that were placed in the last 24 hours, and then enter the data, such as the location, unit price, number of rooms, and so on, into a spreadsheet.
This process takes about an hour of your day, which results in 30 hours per month. That is a lot. In 30 hours, you can easily read a book or take an online course to further develop your skills in some other area. The process described in this paragraph can be automated easily, without the need for machine learning.
Let's take a look at another example. You are spending multiple hours per day on the stock market, deciding what to buy and what to sell. This process is different from the previous one, as it involves some sort of decision making. The thing is, with all of the datasets available online, a skilled individual can use machine learning methods to automate the buy/sell decision-making process.
This is the form of automation that includes machine learning, but no, machine learning and automation are not synonymous. Each can work without the other.
The following sections discuss in great detail the role of automation in machine learning (not vice versa), and answer what we are trying to automate and how it can be achieved in the modern day and age.
