There’s no doubt. The machines are taking over.

When I say “machines”, what I really mean is “machine learning” and by “taking over”, I mean “dramatically increasing in popularity”.

It’s true though. Google Trends shows that the frequency of searches for “machine learning” have nearly tripled over the last three years, with similar trends for related terms such as “deep learning” and “deep neural network”.

The rise in popularity of these terms has seen them become buzzwords, synonymous with companies and projects that are branded as “leading”, “cutting edge” and “revolutionary”. As a direct consequence of this, there has been a huge market increase in the number of available AI-based solutions such as data mining and trend forecasting. Going beyond all the marketing hype, there is a lot of technical complexity being developed and refined, which will be crucial to the long-term usefulness of AI, so the takeover is affecting all aspects of the commercial sector.

The widespread nature of the machine takeover often makes it difficult to understand the impact and usefulness of everything that’s going on and being discussed. As a way of making it clearer, let’s start bridging the gap between the technical and functional by asking some key questions. What is AI? What problems can it solve? Most importantly, how will it affect our industry? 

What does AI mean to you?

For some people it will conjure thoughts of futuristic robots taking over humanity. For others it might mean computers capable of beating humans at chess and for some it might be a worrying thought about our jobs becoming obsolete. Hollywood has done an effective job of selling AI as a nearly unlimited force of self-sustaining intelligence and sentience. However, like most things in Hollywood, very little of it is based in truth. Do we really need to fear robots taking over factories and stealing our jobs? No. Because at the very core of its purpose, AI is designed for one thing. To solve problems.

AI is essentially a big toolbox for helping us solve specific problems. One of the tools we have in our AI toolbox is machine learning. Machine learning is just like a screwdriver… if screwdrivers could automatically predict whether you need a Phillips/slotted type and how big the screw head will be. There are other tools in our AI toolbox too, like genetic algorithms and artificial neural networks, but we won’t worry about those. For now, it’s important to understand that each tool is great at solving a specific type of problem, and not so great at everything else. Yes, you can hammer a nail in using a screwdriver, but it’s much easier to use a hammer.

What problems can machine learning solve then?

In the broadest view, it’s about looking for patterns in large amounts of data. We usually “train” AI on small amounts of data at first, so that we can manually check that it’s finding useful patterns, but it is never taught how it should look for patterns. When we think it’s ready, we can provide it with full-size data sets and ask questions of it, which it will then answer based on the patterns it has discovered.

At this point, the key thing to understand is that AI must be taught before it can be useful. We’re never going to develop the all-powerful, free-thinking AI you see in movies, because we have no way of teaching the concept of free thought. As we’ve already discussed, AI is good for solving very specific, complex problems. We are nowhere near the stage of creating AI which can solve the types of general problems that our brains are so good at.

A great example of AI solving a specific problem is self-driving cars. We take driving for granted, having learned and refined the skill set since we were in our teenage years, but driving is essentially just a problem-solving skill. I want to get from point A to point B. There are several factors that determine how successful you are at solving that problem, like how long it takes you to get there, how many pedestrians you pass, and which radio station you choose. Self-driving cars have learnt to take all these factors into account as well as an understanding of the importance of each factor. What’s more important? Getting there faster, or getting there safely? What if we could take roads that had no pedestrians where we could drive faster? Oh, but what if those roads were slower at certain times of the day? These are all decisions that we make without thinking but are so critical to the performance of the AI algorithms used in self-driving cars.

So if AI is already having real-world impact, how will it affect our industry?

To answer that, we need to take a quick step back and ask “what are the problems we need to solve?” Perhaps more importantly, “what things are most important to us?” From a human-impact perspective, we strive to continue building products that are safe, reliable, and affordable. From a business perspective, we aim to reduce waste, maximize efficiency, and maintain a high standard of quality.

One of the first areas AI will most likely impact is a focus on the efficiency and quality of manufacturing machinery. The problems that AI will be able to solve include answering key questions around how to best optimize workflows to maximize production and when to schedule maintenance so as to get the most out of the lifetime of the equipment. These complex problems involve a huge number of variables and factors, which is where the power of machine learning starts to come alive.

Another area that may benefit from the introduction of AI is in the optimization of storage and transportation solutions. Manufacturing and warehousing areas are generally organised based on human thinking patterns, but machine learning algorithms can often find solutions which may seem random and disorganized on the surface but turn out to be far more efficient in the long-term. Using machine learning to redesign the way we store and transport both raw materials and finished products doesn’t make the human workforce obsolete, it just allows us to become more productive in ways we couldn’t have previously predicted.

The last major area where we may start to see the influence of AI is the creation of product design solutions. However, this is the area that will be met with the most resistance because this requires us to make the biggest leap towards trusting machines with our lives. Self-driving cars still struggle to prove themselves to general society, despite having a proven safety record far greater than most human drivers, so we’re already seeing the challenges. Businesses will be wary of adopting a machine learning approach, because nobody wants to cause an industrial accident as the result of misguided AI. However, we can take baby steps by using AI to assist with product design, instead of giving it sole control. We have already seen a significant shift towards the use of computer-based design and production, particularly with the increased use of CNC machines, as the industry begins to trust Computer-Aided Design and Manufacturing (CAD/CAM).

Where to from here then?

Well, we won’t be bowing down to any AI overlords, but we should change our mindset about how technology can be utilized. Instead of fearing the changes that may come, we should see AI as a tool to making our lives easier, more productive and with less risk. Hollywood does get some things right; AI doesn’t get tired, it doesn’t get distracted and it can solve some problems faster than we can, so we should take those advantages and work alongside our new artificial colleagues.

It’s an exciting time for software development and the ability to innovate with new products and designs is a major factor for success. AI is all about making predictions, so as a prediction for the commercial industry: the companies which come out on top will be those who strike the perfect balance between AI-based innovation, safety, and reliability.