5 questions about AI for expert Geertrui Mieke De Ketelaere

2022 01 27 134027
Ai geertrui mieke dk2 2022 01 27 134141

Q1. How can Flemish companies implement AI?

Geertrui Mieke De Ketelaere: "The task is: start from the challenge. AI can make a difference in jobs where you are dealing with the four Ds: difficult, dull, dangerous, dirty. Let’s take Gumption as an example. Dangerous or dirty are not really relevant in a business group like Gumption, but AI can help with repetitive office work, such as invoice entry."

"Or AI can help you with activities that are too difficult for the human brain, such as interpreting complex business intelligence. We, humans, can think in 2D or 3D, but five or six dimensions are beyond us. AI, on the other hand, does see correlations, but not causality. You need human ingenuity for the smart interpretation of patterns."

Q2. What should Flemish companies look out for?

Geertrui Mieke De Ketelaere: "I see some companies use AI to optimise their HR process. That means that you end up taking on profiles based on profiles you have taken on before. This can work, but if you need a change in strategy, that is actually the last thing you should be doing. So, you need to think about your strategy first before blindly implementing AI."

Q3. Can you give a good current example of how AI improves the world?

Geertrui Mieke De Ketelaere: "In the Netherlands, traffic congestion has been cut by fifteen per cent thanks to AI and the smart combination of data. Often, congestion is caused by the time it takes to get a breakdown lorry or ambulance to the scene. We combined weather data with several data sets: the type of road surface, weekday, holiday calendar, telecommunications data and past road accident locations. Based on these data sets, AI can predict where a road accident is most likely to occur. By strategically stationing breakdown lorries and ambulances near these locations, we were able to minimise congestion. AI tells you what you can’t see with the naked eye. Human input consists of interpreting the information smartly and drawing the right conclusions."

Ai geertrui mieke dk2 2022 01 27 134141

Q1. How can Flemish companies implement AI?

Geertrui Mieke De Ketelaere: "The task is: start from the challenge. AI can make a difference in jobs where you are dealing with the four Ds: difficult, dull, dangerous, dirty. Let’s take Gumption as an example. Dangerous or dirty are not really relevant in a business group like Gumption, but AI can help with repetitive office work, such as invoice entry."

"Or AI can help you with activities that are too difficult for the human brain, such as interpreting complex business intelligence. We, humans, can think in 2D or 3D, but five or six dimensions are beyond us. AI, on the other hand, does see correlations, but not causality. You need human ingenuity for the smart interpretation of patterns."

Q2. What should Flemish companies look out for?

Geertrui Mieke De Ketelaere: "I see some companies use AI to optimise their HR process. That means that you end up taking on profiles based on profiles you have taken on before. This can work, but if you need a change in strategy, that is actually the last thing you should be doing. So, you need to think about your strategy first before blindly implementing AI."

Q3. Can you give a good current example of how AI improves the world?

Geertrui Mieke De Ketelaere: "In the Netherlands, traffic congestion has been cut by fifteen per cent thanks to AI and the smart combination of data. Often, congestion is caused by the time it takes to get a breakdown lorry or ambulance to the scene. We combined weather data with several data sets: the type of road surface, weekday, holiday calendar, telecommunications data and past road accident locations. Based on these data sets, AI can predict where a road accident is most likely to occur. By strategically stationing breakdown lorries and ambulances near these locations, we were able to minimise congestion. AI tells you what you can’t see with the naked eye. Human input consists of interpreting the information smartly and drawing the right conclusions."

Graphics 3 2022 01 27 134921

Q4. Is it true that AI is mostly found in B2C, like in the algorithms that show us suggestions for films, books or products?

Geertrui Mieke De Ketelaere: "No, you will find AI just as much in B2B. For instance, we use AI to help decide where to open a new branch store. We combine the telecommunications data that tell us how busy it is, with the demography and income data of a region. There are thousands of examples like this. You can also find many AI examples based on ERP data. Think of the forecast for the energy sector. With AI, you can predict how much energy will be consumed in a specific location, but you can also see in advance who will probably no longer be able to pay their energy bill. This allows you to respond proactively by mailing or calling this person, always taking into account cost versus gain."

Q5. Will I get to see self-driving cars on the E19 in my lifetime?

Geertrui Mieke De Ketelaere: "I get asked this question a lot (laughs). AI is impressive, but there is still a lot that it can’t do. The world is too complex and is constantly changing at that. Even companies like Uber are no longer that keen on future predictions. Self-driving cars could work on long stretches of motorway, but things get more difficult in the city. It’s no coincidence that the big players have turned their attention to space these days. There is a lot less chance of a collision happening there (laughs)."

"I’ll give you a few examples. What we, as a child aged two or three, understand is not that easy for AI to comprehend. A child is able to tell the difference between a person wearing a snowman suit and an actual snowman made of snow. A smart camera can’t do this. AI can’t recognise that a cat’s tail sticking out from behind a cabinet means that there the cat is still there, because during training, the system mostly saw pictures of whole cats."

Graphics 3 2022 01 27 134921

Q4. Is it true that AI is mostly found in B2C, like in the algorithms that show us suggestions for films, books or products?

Geertrui Mieke De Ketelaere: "No, you will find AI just as much in B2B. For instance, we use AI to help decide where to open a new branch store. We combine the telecommunications data that tell us how busy it is, with the demography and income data of a region. There are thousands of examples like this. You can also find many AI examples based on ERP data. Think of the forecast for the energy sector. With AI, you can predict how much energy will be consumed in a specific location, but you can also see in advance who will probably no longer be able to pay their energy bill. This allows you to respond proactively by mailing or calling this person, always taking into account cost versus gain."

Q5. Will I get to see self-driving cars on the E19 in my lifetime?

Geertrui Mieke De Ketelaere: "I get asked this question a lot (laughs). AI is impressive, but there is still a lot that it can’t do. The world is too complex and is constantly changing at that. Even companies like Uber are no longer that keen on future predictions. Self-driving cars could work on long stretches of motorway, but things get more difficult in the city. It’s no coincidence that the big players have turned their attention to space these days. There is a lot less chance of a collision happening there (laughs)."

"I’ll give you a few examples. What we, as a child aged two or three, understand is not that easy for AI to comprehend. A child is able to tell the difference between a person wearing a snowman suit and an actual snowman made of snow. A smart camera can’t do this. AI can’t recognise that a cat’s tail sticking out from behind a cabinet means that there the cat is still there, because during training, the system mostly saw pictures of whole cats."

P90379558 bmw i3 urban suite artwork 12 2019 2408px resized 2022 01 27 134950

"The CAPTCHAs you solve to prove that you are not a robot are objects of which AI doesn’t yet completely understand the concept: zebra crossings, traffic lights etc. For instance, AI doesn’t see the difference between the letters of a place name on a sign and the same lettering on a lorry. To us, these differences are obvious. By submitting your answer, you are training the system to better understand our complex world."

"Suppose a lorry with a white roof has overturned on a clear day. A smart camera won’t be able to recognise this white surface from the air. That’s dangerous and extra sensors such as radar are needed to correctly assess the situation. Also, what happens when a robot and a self-driving car meet? At the moment, they are not trained to interact with each other and so, there are no right-of-way rules. They don’t recognise each other and will collide. Or, they do recognise each other and they both come to a stop and remain stationary."

"Here, too, you have to ask yourself what problem the self-driving car is meant to solve. Smart systems in the car should mainly increase our intelligence. For instance, they should help us in reduced visibility. AI has to expand my intelligence, not replace it. Things that we can do ourselves, we should most definitely continue to do ourselves."

Read a long interview with Geertrui Mieke De Ketelaere in Gumption Snapshot magazine! Order your copy at https://gumption-snapshot.eu

P90379558 bmw i3 urban suite artwork 12 2019 2408px resized 2022 01 27 134950

"The CAPTCHAs you solve to prove that you are not a robot are objects of which AI doesn’t yet completely understand the concept: zebra crossings, traffic lights etc. For instance, AI doesn’t see the difference between the letters of a place name on a sign and the same lettering on a lorry. To us, these differences are obvious. By submitting your answer, you are training the system to better understand our complex world."

"Suppose a lorry with a white roof has overturned on a clear day. A smart camera won’t be able to recognise this white surface from the air. That’s dangerous and extra sensors such as radar are needed to correctly assess the situation. Also, what happens when a robot and a self-driving car meet? At the moment, they are not trained to interact with each other and so, there are no right-of-way rules. They don’t recognise each other and will collide. Or, they do recognise each other and they both come to a stop and remain stationary."

"Here, too, you have to ask yourself what problem the self-driving car is meant to solve. Smart systems in the car should mainly increase our intelligence. For instance, they should help us in reduced visibility. AI has to expand my intelligence, not replace it. Things that we can do ourselves, we should most definitely continue to do ourselves."

Read a long interview with Geertrui Mieke De Ketelaere in Gumption Snapshot magazine! Order your copy at https://gumption-snapshot.eu

Menu

ENNL