The Rise of Machine Learning (ML) Jobs - Does Investment in ML Guarantee Better Decisions?

In this lightning paced digital age where companies rise and fall in the blink of an eye, no one wants to get left behind by making the wrong decision. Decisions and thus strategies are now driven by data and the insights drawn from them, with instinct now taking a back seat for failing too many times.

Why Machine Learning Jobs Are Key to Better Decision Making

It is in this context that Machine Learning - and its ability via algorithms to process data, learn from the past and better predict the future even to high degrees of accuracy - has risen to such significant prominence as the Sultan of Strategy for many companies. In the US, the presence of the job title Machine Learning Engineer on Indeed in the US rose by a staggering 344% between 2015-2018, and 29% alone between May 2018 – May 2019; showing the importance companies are placing on these skills to help them understand their data, predict outcomes and react accordingly to succeed.

Is Machine Learning a Silver Bullet for Decision Making?

It is with this demand for Machine Learning expertise that Tim Heathcote, Director of Contract Business for Morgan Philips Germany & Austria, spoke to Dr. Alexandra Kirsch, Artificial Intelligence, Exploration und Prototyping Expert at Intuity Media Lab Gmbh. Intuity is a consultancy combining strategic-systemic thinking, user experience design, data science, hardware software prototyping as well as front-end and back-end development to define and develop co-creative next generation products, services and systems for clients.

Q: Is Machine Learning a silver bullet for companies to make the right data driven decisions?

Dr Kirsch: Not always, it depends on the fit of data to problem. In fact, in certain instances, for well-specified problems, such as calculating the interests for a bank account, these are still easier programmed than learned, and the results are much more reliable. ML is the tool for poorly specified problems such as "which type of service would a specific customer prefer?"

But whether ML works for a specific task or question depends on many factors such as available data, the rate at which the interesting parameters change, the requirements of the task such as robustness and understandability.

Q: Then when is it appropriate to use Machine Learning?

Dr Kirsch: Currently a lot of unused data is stored in corporate databases. It is definitely worthwhile digging in those databases and exploring whether some of the data can be utilized for better informed decisions. This should be done with an open mind. Not every question can be answered based on data. Some questions can, but the accuracy and robustness may not suffice to replace current processes, and sometimes completely new questions or even business models may turn up while playing with the data.

The most important thing to understand is that ML is a complex process, in which learning not only happens in the machine, but also in the heads of the people designing the ML task and its embedding into decision making procedures. The picture of dumping a lot of data into a magical algorithm to obtain surprising insights is fundamentally wrong. Making use of ML is hard work. Therefore I recommend an agile approach, keeping the investment small at the beginning and growing the use of ML if it proves useful in the specific context.

Q: So what makes Machine Learning a success – is it simple enough to say the more data, the better the results?

Dr Kirsch: Not necessarily. Since ML is basically statistics, you need a certain number of data to do something at all, but the sheer number is no guarantee for success, nor is a low number a guarantee for failure. It depends on variance. If you want to determine the colour of swans statistically, you can do pretty well with just one data point of a typical white swan. But you can put together as much data of European swans as you like, you will not have a black swan in your dataset. But if you take the right swans, let's say one from every continent, you have covered the possible swan colours in a handful of samples. 

Q: What would you advise to companies who want to get into Machine Learning?

Dr Kirsch: First, try with what you have got. Get out your excel files and available databases. If this is promising, then if necessary, make ML part of your business processes. Make data handling an integral part of your business to have the right data for the right task. Remember ML is a repeated process; your data will change, so must your model.

Q: What else would you recommend to companies in order to be successful with Machine Learning?

Dr Kirsch: Domain expertise! The success of ML largely depends on how you model your data and having the right data. To do this, you must understand the domain. It can also help to split a task into several ML problems or to combine ML with traditional programming. All this requires a deep understanding of the domain and requirements. Therefore if you are serious about establishing machine learning, make sure your ML expertise is connected with your domain expertise. This can be done by using a constant partner with domain expertise such as what we offer at Intuity Media Lab or better establish your own in-house team.

Q: Do companies really need a ML expert if there are so many open source products out there?

Dr Kirsch: ML is not just about getting a learning algorithm to run - anyone can do this with the available tools. It is rather about having a feeling for statistics, data preparation, modeling of the problem, and for this you need experience. But you can start today by getting out any old xls/csv-File/ Database export and plot the data. It may look trivial, but this is the first step in a ML process. Alternatively, get someone with experience, ideally in ML and your domain, and see the benefits.

Q: So what are the take home points for companies ready to make the jump into Machine Learning?

Dr Kirsch: Not all data are gold, some are pebbles. You have to find whether you are sitting on a gold mine or just a heap of pebbles. As with gold, it depends on whether it is valuable to you. If you are starving in the desert, gold will not help you much. Your path to striking gold is to start a little digging, either on your own or using someone like Intuity and see whether there is value in your data. If this is promising, go on to transform your business so that you have the right data available by making sure you systematically collect data, get an infrastructure that gives you access to your data and establish ML/data analysis as a standard method in your company.

If you are looking to help your organisation make better data driven decisions by using the possibilities of Machine Learning, and want to grow the capabilities within your team on a permanent basis please feel to reach out to us !

To get in touch with Germany-based Big Data & Analytics recruiters.

Similarly, for any UK-specific tech sector hiring needs, please get in touch with us !