Artificial Intelligence. The word itself says something about what it does. It tries to create an intelligence, artificially. But how? What exactly is it? Can it improve your everyday life? Can it improve businesses? What is Machine Learning in this picture?

This blog post is for you, who is interested in understanding Machine Learning and Artificial Intelligence, not as a programmer on a detailed level, but on a more general level, where you know what it does, and what is needed to get started on your company’s first ML project.

Picture of a robot

First, we need to understand what Artificial Intelligence and Machine Learning is. Artificial Intelligence is according to the Dictionary:

The theory and development of making a computer system perform a certain task that normally requires human intelligence, such as visual perception, speech recognition, decision-making, and translation between languagesDictionary


The theory of Artificial Intelligence dates back more than 50 years. In Artificial Intelligence, you make a computer system learn a specific task. You do this by providing examples the system uses to learn different patterns or you can make thousands of “if-else”-statements describing these patterns by hand. The pattern recognition is key in both Artificial Intelligence and Machine Learning, and is the basis for their existence. The idea of Artificial Intelligence is, that the system learns by experience (Trial and Error).

When explaining what machine learning is, we always uses Arthur Samuel’s definition:

Machine Learning is a field of study that gives computers the ability to learn without being programmedSamuel, A. (1959)


Machine Learning and Artificial Intelligence are connected because you often use Machine Learning, to get Artificial Intelligence. However, you can have machine learning that is not Artificial Intelligence, but instead an advanced statistical model. This all sounds a bit confusing, but just remember, when people say Artificial Intelligence, you can link it to Machine Learning. That is why, from now on, we will only be talking about Machine Learning, as it is the key component to the “magic”.

In Machine Learning you write some code which can consume vast amounts of data and then produce a model, based on this data, which can say something about the data it has seen. An example of this is something called Predictive Maintenance where if we have data of e.g. a water pump running correctly we can train a model to say if the water pump is running correctly now and will be running correctly in the nearest future. However, this is only one way to solve this task.

Machine Learning consist of three different learning strategies. These defines how the algorithm learns while training on a very specific dataset. Under each learning method, a set of algorithms are defined. Therefore, machine learning is an umbrella of a specific set of algorithms that all have one specific purpose to learn to detect certain patterns from data. In a world full of patterns everywhere, patterns we do not even think about exists or is sometimes too complex for us to see, machine learning has the ability to connect the dots. That is, what machine learning do - find complex patterns. The values (data) for when your machine is running well differs from the values in any other machine. With machine learning it is possible to detect threshold levels dynamic. That is why, when training a machine learning algorithm, it must be on your data.

We asked a newly educated Bachelor of Technology Management and Marine Engineering what he would like to know about Machine Learning. The following will answer these questions!

How it benefits something you already do

Your life is already influenced by machine learning without you even knows it.

The police’s ANPG (automatic driving license recognition) programme, usesto investigate whether a given car is listen in a hotlist. This is a specific area of Machine Learning called computer vision where you learn the computer system to see patterns in images. Additionally, if you have passed through Storebæltsbroen recently and used the PaybyPlate paying method, your license plate is used to give you access to the bridge.

If you have used sentences such as “Siri” or “Ok Google” then you have also been in touch with machine learning as it is used to understand speech and make speech responses back to you.

Have you searched for something on google today? Then all the results you have seen has been improved by Machine Learning to give you the results you most likely want to find.

Finally, if you use Youtube or Spotify, you have been introduced to Machine learning algorithms that try to predict what you want to hear/see next!

As you can see Machine Learning is everywhere if you use technology.

How Machine Learning can improve your everyday life

We believe that machine learning is going to be the most life-changing technology in our life-time even bigger than electricity or the internet. It has the potential to improve all industries - both the way we work, but also the amount of information we have on certain things, and the quality of decision making. We have already discussed how it can be used to predict credit card fraud in the finance sector. They can additionally use machine learning to automating processes in paperwork.

For the health sector, you can use image recognition to detect breaks in bones, benign/malignant tumors, and to detect possible tumors in mammograms. Finally, the health sector can use something called predictive maintenance to predict future breakdowns on their machines to reduce mean-time-between-failure, and mean-time-to-repair - both improving the total up-time.

Manufacturing companies uses machine learning to automate processes, improve decision-making, and predict future events. We help companies use machine learning to get better forecast (with many sources like weather), on which they can better plan production setting, purchase of raw material etc. Additionally, manufacturing can use machine learning for reducing unplanned downtime due to breakdowns. By implementing predictive maintenance, you have monitoring on how production or a single machine is working today - and in the future. Thereby, you can get a warning in time to take action. For some companies, breakdowns can cause severe safety challenges. For other companies, the Work-in-Progress must be discarded when a breakdown occurs. Some have such high utilization of their production that a breakdown can mean not delivering in time to customers. Finally, the manufacturing companies can use image recognition for quality inspections.

Finally, the utility sector can use machine learning for precise forecasting of the consumption of water and heat. Machine learning can include many sources such as weather predictions as well as seasonality and holiday increases and decreases in demand. The utility sector can additionally use predictive maintenance to improve safety, reduce breakdowns, and mean time to repair (MTTR).

Besides pattern recognition which runs real-time machine learning gives you the opportunity to have real-time and future insight of production. You can use this new knowledge for better decision making, and planning one step ahead. Additionally, you can reduce total costs of operation. It does not just have the potential to give you a competitive advantage - it gives you one right off the bat.

What you need to get started

Now the police and finance industry are already working with machine learning solutions. Even for the agri industry we have seen intelligent products that e.g. can adjust the amount of water the given sort is getting based on how dry it looks (image recognition). However, it seems that only a few of the manufacturing-, health-, and utility companies are working with machine learning solutions.

A survey from 2017 suggest that only 5% of companies in Denmark uses machine learning and artificial intelligence. To compare - 13% of the companies in the Information and Communication sector used these technologies back in 2017. A lot has happened since 2017, so the number might be a bit higher today. Specific manufacturing companies must start utilizing the possibilities of software.

But how do you get started on your company’s first Machine Learning project? We have already introduced it a couple of times in previous blog posts about data strategy, and big data.

In neurospace we live by:

Think big, Start small, and create value fastneurospace


To start a Machine Learning project, you must be creative, and be able to detect issues in your everyday work, that is based on patterns. When you know where you want to use Machine Learning, you do not have to start gathering data for years before you can leverage the potential. You start at a small defined area, of which you gather enough data to prove the hypothesis that machine learning can create value. These are called pilot-projects, and can be performed within a few months. This additionally helps creating value fast, and reduces the total cost of the project.

However, to be able to deliver machine learning to production, you need people with the right set of skills. This includes, amongst others, statistics, machine learning, cloud computing, security, sensors (sometimes called IoT and Industrial IoT), and kubernetes.

Conclusion

Artificial Inteligence is when a computer system learns to perform a certain task, that normally requires human intelligence. Machine Learning is one approach to achieve Artificial Intelligence. Machine Learning is an umbrella of specific algorithms that all have the purpose of learning to detect patterns in a specific problem. It detects these patterns dynamic, and gives you real-time insight on e.g. how your machine is operating now, and in the future.

To get started you must be able to detect these patterns in your everyday life, and begin pilot projects to seek the potential.

// Maria Jensen, ML Engineer @ neurospace