AI explained
09/05/2018

AI explained: an overview for busy people

A machine is considered “intelligent” if it can independently take actions to maximize its success within the environment it perceives. However, what is considered intelligent now will not necessarily be so in ten years. Intelligent systems today not only act on their own, but also learn how to do so based on contextual information and other data.

Why is AI so popular today?

Cheaper computing power and data resources and the support of big players like Microsoft, Google, Amazon and more led to the development of new, effective algorithms. Much of this innovation has occurred in the field of machine learning with the development of deep neural networks coining the term ‘deep learning’. Training these deep networks requires significantly more computing power and data, which is now available to players of all sizes.

Most of these breakthrough algorithms and corresponding models are open source – enabling companies to reuse existing work to develop their own cases.


Any player that sees the opportunity to improve processes and products, reduce costs or boost revenue should have their sights set on AI.

Who is using it?

Big players with specialized research labs (Microsoft Research AI, Facebook AI Research, Google DeepMind, Google Research, etc.) have the resources needed to drive these new developments. DeepMind came with a USD 500 million price tag and was initially used to cut electricity consumption at Google datacenters – an investment that will pay for itself in 10 years through one application alone.

Small companies also use AI. For example, Lee & Ally is a chatbot developed by DeJuristen that provides legal advice. Swarovski created an intelligent visual search system to identify pieces of jewelry, and Rolls Royce uses predictive maintenance to reduce errors and cut fuel costs. These are just a few of many examples of successful applications.

Any player that sees the opportunity to improve processes and products, reduce costs or boost revenue should have their sights set on AI.

What are they doing with it?

Key application areas include computer vision, natural language processing, forecasting, classification, recommendation engines and clustering. They are all driven by data, of course.

  • Computer vision: the intelligent gleaning of information from images, such as quality, object detection and location, and similarities.
  • Natural language processing: used to power smart chatbots – matching questions to commands – and automatic document tagging, among others.
  • Forecasting: predicting numerical values to gauge, for example, product demand.
  • Classification: answers yes/no questions, e.g. “will this customer churn?”
  • Recommendation engines: connecting the right products with the right customers.
  • Clustering: grouping similar customers together to approach them in more personalized ways based on their actions, preferences or behaviors.

How do you implement it?

Classification, forecasting and recommendation engines aren’t new, but they are useful for improving processes. Computer vision and natural language processing are tougher, but new deep learning techniques have made these solutions more accessible. Models have been trained using large datasets (ImageNet, COCO) and released for free use.

Take an example: an application that identifies whether a picture contains a hotdog or not. To create this tool, you’ll need a dataset of hotdog pictures and not-hotdog pictures – and lots of them. However, you can drastically reduce the volume of training images needed by applying transfer learning: chopping off the ‘head’ of an open source neural network and retraining that head using hotdog images.

The same technique can be applied to natural language processing. Pretrained word embeddings capture similar word meanings they contain. The word ‘bus’ will be more closely related to ‘train’ than ‘chihuahua’. Word embeddings will capture synonyms by default, reducing your implementation workload.



Choose diverse profiles to ensure varied approaches and a greater range of knowledge. But most of all, invest in curious people with a strong willingness to learn.

Setting up your dream AI team

You’re definitely going to need skilled people to tackle AI challenges, but while it might take some searching, it’s possible to find the right profiles. These include:

  • Data analysts: capable of transforming data into useful formats for decision-making though ETL work, visualization and statistical analysis.
  • Data scientists: expand on the work of the data analyst by applying the correct statistical and machine learning techniques, translating a business question into an experiment and delivering the resulting insights. Statistics experts.
  • Data engineers: tackle big data problems such as unstructured data, continuous data or huge volumes of data from a software development approach.
  • Machine learning engineers: specialized in deep learning, grow the depth of data models to improve the accuracy (or other metrics) of a simple model. Strong in calculus and linear algebra.
Choose diverse profiles to ensure varied approaches and a greater range of knowledge. But most of all, invest in curious people with a strong willingness to learn. This will boost the quality of the data research, drive continuous improvement, and ensures that your team can keep up with the trends.

In search of an AI innovation partner that thinks alongside your team to develop value-adding applications unique to your business?