Intelligence demonstrated by machines

AI is defined as intelligence demonstrated by machines, as opposed to the natural intelligence displayed by humans. From that perspective, an optimization model is an example of AI and is used in many advanced decision-support systems.

More recent AI-techniques, like machine learning and deep learning, focus on learning from historical data, which has strong applications in forecasting. These techniques can also be used to continuously improve data, variables and restrictions.

In today’s context, the term AI is frequently used to refer to these more modern techniques – “if your algorithm doesn’t include machine learning, then it’s outdated.” However, it’s good to understand there is more to AI.

AI adds value in many domains

AI is well-known by now in recognizing relationships, images, or learning about your favorite Netflix preferences. But it can also be applied in various domains of optimization and decision-making, across different industries.

Forecasting in uncertainty

Thanks to new AI techniques and technologies, we are now better able to deal with uncertainty. Uncertainty in the weather forecast, uncertainty about future stock, uncertainty in predicted maintenance … it’s now possible to create robust and resilient schedules and reliable forecasts, depending on the level of uncertainty in the input data. For example, in workforce scheduling, forecasting is used to predict the expected amount of work per anticipated available employees for a certain moment in time.

Integrated Approaches

Due to the availability and circulation of data, and the increased power of computation and mathematics, more integrated approaches are possible to solve complex problems. The same applies in business: advanced algorithms enable organizations to integrate all sorts of business processes into one. For instance, combining forecasting algorithms for inventory management with routing optimization, or combining supply chain optimization with revenue management and pricing.

Network Effects

Network Effects is a phenomenon that makes a product increase in value when the number of customers using it increases. More users means more data, more data means more learning, and more learning means new possibilities.

Take for example AI for routing optimization. Real-time data sources in algorithms, like traffic density in map data, increase the accuracy of suggested routes. Accuracy receives an even bigger boost when you consider additional data, like the moment in the week. If an increasing number of users actively provide feedback on suggested routes, accuracy can be improved even more. Combine this with some machine learning, continuously comparing and benchmarking the data to improve the optimizers, and you can imagine how network effects quickly result in improved routes. In a similar way, we can tune and script optimization to learn from past usage.

Personalization

One of the major AI trends which emerged in the past 10 years is personalized services. In this case, AI “optimizes” the offering of personalized services. For example, AI can be used in e-commerce to offer the right personalized time slots per request. Here, a trade-off is required between customer satisfaction (offering the most popular ones) and efficiency (offering the most efficient ones).

Another example is creating workforce schedules for drivers, nurses, service employees, and so on. Here, AI can be used to learn from historical rosters (planned data) and the changing preferences of the workforce (actual data). For example, desired shift lengths. The optimizer combines multiple aspects, resulting in a more personalized roster that drives employee engagement and efficiency. It can also learn how attractive a roster is by monitoring the amount of personal changes made after publication.

These examples show that AI is gaining momentum as a mean to improve business decision-making.

The Artificial Intelligence momentum – 5 reasons

Not only is AI omnipresent, even the discussion about AI - and Explainable AI – takes place all around us. Why now? I can think of 5 reasons:

1. Increasing computer power. Back in 1970, Dr. Moore claimed that computers’ processing power would double every two years. This brazen prediction, known as Moore’s law, was proven true for decades - until now. In the 1980’s, it took computers hours to calculate a simple task. These days, they can calculate unimaginably large tasks in a split second.

2. Enormous amounts of data. Thanks to online videos, sensor data, social media and other platforms, the amount of data is exploding day by day. Currently, this is expressed in zettabytes: 10 to the power 21. How is that helping AI? Take, for example, a general practitioner who needs to diagnose a sneezing patient. The more observations (data) we add, like increase in temperature, the more precise and valid the diagnosis will be. The more data, the higher the quality of the decision.

3. Algorithm availability. Thanks to mathematicians, algorithm improvements are going even faster than Moore’s Law predicted! Why? Because you get an exponential outcome when you steadily improve algorithms – doing something ten times faster or better.

Since I’m a teacher, let me give one example used in school: sorting a random set of numbers. In the early days, our algorithm would just go through the whole list to find the smallest number, and then again, repeatedly. So, 100 searches for 100 numbers. Mathematicians created an algorithm that only requires 10 searches. That is a big improvement. Imagine if that algorithm is made ten times faster. And then imagine machines using many algorithms at the same time …

4. Specialized hardware. Nowadays, special hardware is available to fully integrate AI, like AI accelerators, robotics, and parallel machines, and speed up the technology for specific applications. For example, the AI in your phone’s camera recognizes faces or signals suspicious behavior. This AI works even without an internet connection.

5. Proliferating use cases. Increasingly practical examples are now available in all kind of industries, like health, gaming, logistics, HR, and others. Most social media platforms operate using AI. In other words, AI is finding its way in everyone’s daily life, both personal and professional.

These reasons are all driving business awareness. Both commercial and non-commercial organizations, as well as governments, are convinced they must invest in AI, given its success and clear use cases. That said, I am convinced that AI should not be an end goal. It’s “just” one of many tools we have at our disposal to improve decision-making for organizations and society at large.

About the author

Goos Kant (1967) is a full-time professor of Logistic Optimization at Tilburg University. He is involved in the master program of Business Analytics and Operations Research, as well as in the master program Data Science & Entrepreneurship. He is the project leader of a large R&D project on horizontal collaboration in logistics. Goos is also a managing partner at ORTEC, with global responsibility for all solutions in the logistics industry. His primary area of interest lies in the 3PL-industry in optimizing their planning processes in the end-to-end supply chain. Goos is involved in courses from MBA-schools TIAS and Nyenrode. He holds both an MSc and a PhD in Computer Science.

Connect with Goos
ORTEC